WebSVN – HelenOS – Diff – /tags/0.3.0/uspace/srv/vfs/vfs_ops.c


/*
 * 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    vfs_ops.c
 * @brief   Operations that VFS offers to its clients.
 */
 
#include "vfs.h"
#include <ipc/ipc.h>
#include <async.h>
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <bool.h>
#include <futex.h>
#include <rwlock.h>
#include <libadt/list.h>
#include <unistd.h>
#include <ctype.h>
#include <fcntl.h>
#include <assert.h>
#include <vfs/canonify.h>
 
/* Forward declarations of static functions. */
static int vfs_truncate_internal(fs_handle_t, dev_handle_t, fs_index_t, size_t);
 
/**
 * This rwlock prevents the race between a triplet-to-VFS-node resolution and a
 * concurrent VFS operation which modifies the file system namespace.
 */
RWLOCK_INITIALIZE(namespace_rwlock);
 
futex_t rootfs_futex = FUTEX_INITIALIZER;
vfs_triplet_t rootfs = {
    .fs_handle = 0,
    .dev_handle = 0,
    .index = 0,
};
 
static int
lookup_root(fs_handle_t fs_handle, dev_handle_t dev_handle,
    vfs_lookup_res_t *result)
{
    vfs_pair_t altroot = {
        .fs_handle = fs_handle,
        .dev_handle = dev_handle,
    };
 
    return vfs_lookup_internal("/", L_DIRECTORY, result, &altroot);
}
 
void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
{
    dev_handle_t dev_handle;
    vfs_node_t *mp_node = NULL;
 
    /*
     * We expect the library to do the device-name to device-handle
     * translation for us, thus the device handle will arrive as ARG1
     * in the request.
     */
    dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);
 
    /*
     * For now, don't make use of ARG2 and ARG3, but they can be used to
     * carry mount options in the future.
     */
 
    ipc_callid_t callid;
    size_t size;
 
    /*
     * Now, we expect the client to send us data with the name of the file
     * system.
     */
    if (!ipc_data_write_receive(&callid, &size)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
 
    /*
     * Don't receive more than is necessary for storing a full file system
     * name.
     */
    if (size < 1 || size > FS_NAME_MAXLEN) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
 
    /* Deliver the file system name. */
    char fs_name[FS_NAME_MAXLEN + 1];
    (void) ipc_data_write_finalize(callid, fs_name, size);
    fs_name[size] = '\0';
   
    /*
     * Check if we know a file system with the same name as is in fs_name.
     * This will also give us its file system handle.
     */
    fs_handle_t fs_handle = fs_name_to_handle(fs_name, true);
    if (!fs_handle) {
        ipc_answer_0(rid, ENOENT);
        return;
    }
 
    /* Now, we want the client to send us the mount point. */
    if (!ipc_data_write_receive(&callid, &size)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
 
    /* Check whether size is reasonable wrt. the mount point. */
    if (size < 1 || size > MAX_PATH_LEN) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
    /* Allocate buffer for the mount point data being received. */
    uint8_t *buf;
    buf = malloc(size + 1);
    if (!buf) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
 
    /* Deliver the mount point. */
    (void) ipc_data_write_finalize(callid, buf, size);
    buf[size] = '\0';
 
    /*
     * Lookup the root node of the filesystem being mounted.
     * In this case, we don't need to take the namespace_futex as the root
     * node cannot be removed. However, we do take a reference to it so
     * that we can track how many times it has been mounted.
     */
    int rc;
    vfs_lookup_res_t mr_res;
    rc = lookup_root(fs_handle, dev_handle, &mr_res);
    if (rc != EOK) {
        free(buf);
        ipc_answer_0(rid, rc);
        return;
    }
    vfs_node_t *mr_node = vfs_node_get(&mr_res);
    if (!mr_node) {
        free(buf);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
 
    /* Finally, we need to resolve the path to the mountpoint. */
    vfs_lookup_res_t mp_res;
    futex_down(&rootfs_futex);
    if (rootfs.fs_handle) {
        /* We already have the root FS. */
        rwlock_write_lock(&namespace_rwlock);
        rc = vfs_lookup_internal(buf, L_DIRECTORY, &mp_res, NULL);
        if (rc != EOK) {
            /* The lookup failed for some reason. */
            rwlock_write_unlock(&namespace_rwlock);
            futex_up(&rootfs_futex);
            vfs_node_put(mr_node);  /* failed -> drop reference */
            free(buf);
            ipc_answer_0(rid, rc);
            return;
        }
        mp_node = vfs_node_get(&mp_res);
        if (!mp_node) {
            rwlock_write_unlock(&namespace_rwlock);
            futex_up(&rootfs_futex);
            vfs_node_put(mr_node);  /* failed -> drop reference */
            free(buf);
            ipc_answer_0(rid, ENOMEM);
            return;
        }
        /*
         * Now we hold a reference to mp_node.
         * It will be dropped upon the corresponding VFS_UNMOUNT.
         * This prevents the mount point from being deleted.
         */
        rwlock_write_unlock(&namespace_rwlock);
    } else {
        /* We still don't have the root file system mounted. */
        if ((size == 1) && (buf[0] == '/')) {
            /* For this simple, but important case, we are done. */
            rootfs = mr_res.triplet;
            futex_up(&rootfs_futex);
            free(buf);
            ipc_answer_0(rid, EOK);
            return;
        } else {
            /*
             * We can't resolve this without the root filesystem
             * being mounted first.
             */
            futex_up(&rootfs_futex);
            free(buf);
            vfs_node_put(mr_node);  /* failed -> drop reference */
            ipc_answer_0(rid, ENOENT);
            return;
        }
    }
    futex_up(&rootfs_futex);
   
    free(buf);  /* The buffer is not needed anymore. */
   
    /*
     * At this point, we have all necessary pieces: file system and device
     * handles, and we know the mount point VFS node and also the root node
     * of the file system being mounted.
     */
 
    int phone = vfs_grab_phone(mp_res.triplet.fs_handle);
    /* Later we can use ARG3 to pass mode/flags. */
    aid_t req1 = async_send_3(phone, VFS_MOUNT,
        (ipcarg_t) mp_res.triplet.dev_handle,
        (ipcarg_t) mp_res.triplet.index, 0, NULL);
    /* The second call uses the same method. */
    aid_t req2 = async_send_3(phone, VFS_MOUNT,
        (ipcarg_t) mr_res.triplet.fs_handle,
        (ipcarg_t) mr_res.triplet.dev_handle,
        (ipcarg_t) mr_res.triplet.index, NULL);
    vfs_release_phone(phone);
 
    ipcarg_t rc1;
    ipcarg_t rc2;
    async_wait_for(req1, &rc1);
    async_wait_for(req2, &rc2);
 
    if ((rc1 != EOK) || (rc2 != EOK)) {
        /* Mount failed, drop references to mr_node and mp_node. */
        vfs_node_put(mr_node);
        if (mp_node)
            vfs_node_put(mp_node);
    }
   
    if (rc2 == EOK)
        ipc_answer_0(rid, rc1);
    else if (rc1 == EOK)
        ipc_answer_0(rid, rc2);
    else
        ipc_answer_0(rid, rc1);
}
 
void vfs_open(ipc_callid_t rid, ipc_call_t *request)
{
    if (!vfs_files_init()) {
        ipc_answer_0(rid, ENOMEM);
        return;
    }
 
    /*
     * The POSIX interface is open(path, oflag, mode).
     * We can receive oflags and mode along with the VFS_OPEN call; the path
     * will need to arrive in another call.
     *
     * We also receive one private, non-POSIX set of flags called lflag
     * used to pass information to vfs_lookup_internal().
     */
    int lflag = IPC_GET_ARG1(*request);
    int oflag = IPC_GET_ARG2(*request);
    int mode = IPC_GET_ARG3(*request);
    size_t len;
 
    if (oflag & O_CREAT)
        lflag |= L_CREATE;
    if (oflag & O_EXCL)
        lflag |= L_EXCLUSIVE;
 
    ipc_callid_t callid;
 
    if (!ipc_data_write_receive(&callid, &len)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
    char *path = malloc(len + 1);
    if (!path) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
    int rc;
    if ((rc = ipc_data_write_finalize(callid, path, len))) {
        ipc_answer_0(rid, rc);
        free(path);
        return;
    }
    path[len] = '\0';
   
    /*
     * Avoid the race condition in which the file can be deleted before we
     * find/create-and-lock the VFS node corresponding to the looked-up
     * triplet.
     */
    if (lflag & L_CREATE)
        rwlock_write_lock(&namespace_rwlock);
    else
        rwlock_read_lock(&namespace_rwlock);
 
    /* The path is now populated and we can call vfs_lookup_internal(). */
    vfs_lookup_res_t lr;
    rc = vfs_lookup_internal(path, lflag, &lr, NULL);
    if (rc) {
        if (lflag & L_CREATE)
            rwlock_write_unlock(&namespace_rwlock);
        else
            rwlock_read_unlock(&namespace_rwlock);
        ipc_answer_0(rid, rc);
        free(path);
        return;
    }
 
    /* Path is no longer needed. */
    free(path);
 
    vfs_node_t *node = vfs_node_get(&lr);
    if (lflag & L_CREATE)
        rwlock_write_unlock(&namespace_rwlock);
    else
        rwlock_read_unlock(&namespace_rwlock);
 
    /* Truncate the file if requested and if necessary. */
    if (oflag & O_TRUNC) {
        rwlock_write_lock(&node->contents_rwlock);
        if (node->size) {
            rc = vfs_truncate_internal(node->fs_handle,
                node->dev_handle, node->index, 0);
            if (rc) {
                rwlock_write_unlock(&node->contents_rwlock);
                vfs_node_put(node);
                ipc_answer_0(rid, rc);
                return;
            }
            node->size = 0;
        }
        rwlock_write_unlock(&node->contents_rwlock);
    }
 
    /*
     * Get ourselves a file descriptor and the corresponding vfs_file_t
     * structure.
     */
    int fd = vfs_fd_alloc();
    if (fd < 0) {
        vfs_node_put(node);
        ipc_answer_0(rid, fd);
        return;
    }
    vfs_file_t *file = vfs_file_get(fd);
    file->node = node;
    if (oflag & O_APPEND)
        file->append = true;
 
    /*
     * The following increase in reference count is for the fact that the
     * file is being opened and that a file structure is pointing to it.
     * It is necessary so that the file will not disappear when
     * vfs_node_put() is called. The reference will be dropped by the
     * respective VFS_CLOSE.
     */
    vfs_node_addref(node);
    vfs_node_put(node);
 
    /* Success! Return the new file descriptor to the client. */
    ipc_answer_1(rid, EOK, fd);
}
 
void vfs_close(ipc_callid_t rid, ipc_call_t *request)
{
    int fd = IPC_GET_ARG1(*request);
    if (fd >= MAX_OPEN_FILES) {
        ipc_answer_0(rid, EBADF);
        return;
    }
    vfs_fd_free(fd);
    ipc_answer_0(rid, EOK);
}
 
static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
{
 
    /*
     * The following code strongly depends on the fact that the files data
     * structure can be only accessed by a single fibril and all file
     * operations are serialized (i.e. the reads and writes cannot
     * interleave and a file cannot be closed while it is being read).
     *
     * Additional synchronization needs to be added once the table of
     * open files supports parallel access!
     */
 
    int fd = IPC_GET_ARG1(*request);
 
    /* Lookup the file structure corresponding to the file descriptor. */
    vfs_file_t *file = vfs_file_get(fd);
    if (!file) {
        ipc_answer_0(rid, ENOENT);
        return;
    }
 
    /*
     * Now we need to receive a call with client's
     * IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
     */
    ipc_callid_t callid;
    int res;
    if (read)
        res = ipc_data_read_receive(&callid, NULL);
    else
        res = ipc_data_write_receive(&callid, NULL);
    if (!res) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
 
    /*
     * Lock the open file structure so that no other thread can manipulate
     * the same open file at a time.
     */
    futex_down(&file->lock);
 
    /*
     * Lock the file's node so that no other client can read/write to it at
     * the same time.
     */
    if (read)
        rwlock_read_lock(&file->node->contents_rwlock);
    else
        rwlock_write_lock(&file->node->contents_rwlock);
 
    int fs_phone = vfs_grab_phone(file->node->fs_handle);  
   
    /* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
    aid_t msg;
    ipc_call_t answer;
    if (!read && file->append)
        file->pos = file->node->size;
    msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
        file->node->dev_handle, file->node->index, file->pos, &answer);
   
    /*
     * Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
     * destination FS server. The call will be routed as if sent by
     * ourselves. Note that call arguments are immutable in this case so we
     * don't have to bother.
     */
    ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
 
    vfs_release_phone(fs_phone);
 
    /* Wait for reply from the FS server. */
    ipcarg_t rc;
    async_wait_for(msg, &rc);
    size_t bytes = IPC_GET_ARG1(answer);
 
    /* Unlock the VFS node. */
    if (read)
        rwlock_read_unlock(&file->node->contents_rwlock);
    else {
        /* Update the cached version of node's size. */
        if (rc == EOK)
            file->node->size = IPC_GET_ARG2(answer);
        rwlock_write_unlock(&file->node->contents_rwlock);
    }
 
    /* Update the position pointer and unlock the open file. */
    if (rc == EOK)
        file->pos += bytes;
    futex_up(&file->lock);
 
    /*
     * FS server's reply is the final result of the whole operation we
     * return to the client.
     */
    ipc_answer_1(rid, rc, bytes);
}
 
void vfs_read(ipc_callid_t rid, ipc_call_t *request)
{
    vfs_rdwr(rid, request, true);
}
 
void vfs_write(ipc_callid_t rid, ipc_call_t *request)
{
    vfs_rdwr(rid, request, false);
}
 
void vfs_seek(ipc_callid_t rid, ipc_call_t *request)
{
    int fd = (int) IPC_GET_ARG1(*request);
    off_t off = (off_t) IPC_GET_ARG2(*request);
    int whence = (int) IPC_GET_ARG3(*request);
 
 
    /* Lookup the file structure corresponding to the file descriptor. */
    vfs_file_t *file = vfs_file_get(fd);
    if (!file) {
        ipc_answer_0(rid, ENOENT);
        return;
    }
 
    off_t newpos;
    futex_down(&file->lock);
    if (whence == SEEK_SET) {
        file->pos = off;
        futex_up(&file->lock);
        ipc_answer_1(rid, EOK, off);
        return;
    }
    if (whence == SEEK_CUR) {
        if (file->pos + off < file->pos) {
            futex_up(&file->lock);
            ipc_answer_0(rid, EOVERFLOW);
            return;
        }
        file->pos += off;
        newpos = file->pos;
        futex_up(&file->lock);
        ipc_answer_1(rid, EOK, newpos);
        return;
    }
    if (whence == SEEK_END) {
        rwlock_read_lock(&file->node->contents_rwlock);
        size_t size = file->node->size;
        rwlock_read_unlock(&file->node->contents_rwlock);
        if (size + off < size) {
            futex_up(&file->lock);
            ipc_answer_0(rid, EOVERFLOW);
            return;
        }
        newpos = size + off;
        futex_up(&file->lock);
        ipc_answer_1(rid, EOK, newpos);
        return;
    }
    futex_up(&file->lock);
    ipc_answer_0(rid, EINVAL);
}
 
int
vfs_truncate_internal(fs_handle_t fs_handle, dev_handle_t dev_handle,
    fs_index_t index, size_t size)
{
    ipcarg_t rc;
    int fs_phone;
   
    fs_phone = vfs_grab_phone(fs_handle);
    rc = async_req_3_0(fs_phone, VFS_TRUNCATE, (ipcarg_t)dev_handle,
        (ipcarg_t)index, (ipcarg_t)size);
    vfs_release_phone(fs_phone);
    return (int)rc;
}
 
void vfs_truncate(ipc_callid_t rid, ipc_call_t *request)
{
    int fd = IPC_GET_ARG1(*request);
    size_t size = IPC_GET_ARG2(*request);
    int rc;
 
    vfs_file_t *file = vfs_file_get(fd);
    if (!file) {
        ipc_answer_0(rid, ENOENT);
        return;
    }
    futex_down(&file->lock);
 
    rwlock_write_lock(&file->node->contents_rwlock);
    rc = vfs_truncate_internal(file->node->fs_handle,
        file->node->dev_handle, file->node->index, size);
    if (rc == EOK)
        file->node->size = size;
    rwlock_write_unlock(&file->node->contents_rwlock);
 
    futex_up(&file->lock);
    ipc_answer_0(rid, (ipcarg_t)rc);
}
 
void vfs_mkdir(ipc_callid_t rid, ipc_call_t *request)
{
    int mode = IPC_GET_ARG1(*request);
 
    size_t len;
    ipc_callid_t callid;
 
    if (!ipc_data_write_receive(&callid, &len)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
    char *path = malloc(len + 1);
    if (!path) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
    int rc;
    if ((rc = ipc_data_write_finalize(callid, path, len))) {
        ipc_answer_0(rid, rc);
        free(path);
        return;
    }
    path[len] = '\0';
   
    rwlock_write_lock(&namespace_rwlock);
    int lflag = L_DIRECTORY | L_CREATE | L_EXCLUSIVE;
    rc = vfs_lookup_internal(path, lflag, NULL, NULL);
    rwlock_write_unlock(&namespace_rwlock);
    free(path);
    ipc_answer_0(rid, rc);
}
 
void vfs_unlink(ipc_callid_t rid, ipc_call_t *request)
{
    int lflag = IPC_GET_ARG1(*request);
 
    size_t len;
    ipc_callid_t callid;
 
    if (!ipc_data_write_receive(&callid, &len)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
    char *path = malloc(len + 1);
    if (!path) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
    int rc;
    if ((rc = ipc_data_write_finalize(callid, path, len))) {
        ipc_answer_0(rid, rc);
        free(path);
        return;
    }
    path[len] = '\0';
   
    rwlock_write_lock(&namespace_rwlock);
    lflag &= L_DIRECTORY;   /* sanitize lflag */
    vfs_lookup_res_t lr;
    rc = vfs_lookup_internal(path, lflag | L_UNLINK, &lr, NULL);
    free(path);
    if (rc != EOK) {
        rwlock_write_unlock(&namespace_rwlock);
        ipc_answer_0(rid, rc);
        return;
    }
 
    /*
     * The name has already been unlinked by vfs_lookup_internal().
     * We have to get and put the VFS node to ensure that it is
     * VFS_DESTROY'ed after the last reference to it is dropped.
     */
    vfs_node_t *node = vfs_node_get(&lr);
    futex_down(&nodes_futex);
    node->lnkcnt--;
    futex_up(&nodes_futex);
    rwlock_write_unlock(&namespace_rwlock);
    vfs_node_put(node);
    ipc_answer_0(rid, EOK);
}
 
void vfs_rename(ipc_callid_t rid, ipc_call_t *request)
{
    size_t len;
    ipc_callid_t callid;
    int rc;
 
    /* Retrieve the old path. */
    if (!ipc_data_write_receive(&callid, &len)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
    char *old = malloc(len + 1);
    if (!old) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
    if ((rc = ipc_data_write_finalize(callid, old, len))) {
        ipc_answer_0(rid, rc);
        free(old);
        return;
    }
    old[len] = '\0';
   
    /* Retrieve the new path. */
    if (!ipc_data_write_receive(&callid, &len)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        free(old);
        return;
    }
    char *new = malloc(len + 1);
    if (!new) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        free(old);
        return;
    }
    if ((rc = ipc_data_write_finalize(callid, new, len))) {
        ipc_answer_0(rid, rc);
        free(old);
        free(new);
        return;
    }
    new[len] = '\0';
 
    char *oldc = canonify(old, &len);
    char *newc = canonify(new, NULL);
    if (!oldc || !newc) {
        ipc_answer_0(rid, EINVAL);
        free(old);
        free(new);
        return;
    }
    if (!strncmp(newc, oldc, len)) {
        /* oldc is a prefix of newc */
        ipc_answer_0(rid, EINVAL);
        free(old);
        free(new);
        return;
    }
   
    vfs_lookup_res_t old_lr;
    vfs_lookup_res_t new_lr;
    vfs_lookup_res_t new_par_lr;
    rwlock_write_lock(&namespace_rwlock);
    /* Lookup the node belonging to the old file name. */
    rc = vfs_lookup_internal(oldc, L_NONE, &old_lr, NULL);
    if (rc != EOK) {
        rwlock_write_unlock(&namespace_rwlock);
        ipc_answer_0(rid, rc);
        free(old);
        free(new);
        return;
    }
    vfs_node_t *old_node = vfs_node_get(&old_lr);
    if (!old_node) {
        rwlock_write_unlock(&namespace_rwlock);
        ipc_answer_0(rid, ENOMEM);
        free(old);
        free(new);
        return;
    }
    /* Lookup parent of the new file name. */
    rc = vfs_lookup_internal(newc, L_PARENT, &new_par_lr, NULL);
    if (rc != EOK) {
        rwlock_write_unlock(&namespace_rwlock);
        ipc_answer_0(rid, rc);
        free(old);
        free(new);
        return;
    }
    /* Check whether linking to the same file system instance. */
    if ((old_node->fs_handle != new_par_lr.triplet.fs_handle) ||
        (old_node->dev_handle != new_par_lr.triplet.dev_handle)) {
        rwlock_write_unlock(&namespace_rwlock);
        ipc_answer_0(rid, EXDEV);   /* different file systems */
        free(old);
        free(new);
        return;
    }
    /* Destroy the old link for the new name. */
    vfs_node_t *new_node = NULL;
    rc = vfs_lookup_internal(newc, L_UNLINK, &new_lr, NULL);
    switch (rc) {
    case ENOENT:
        /* simply not in our way */
        break;
    case EOK:
        new_node = vfs_node_get(&new_lr);
        if (!new_node) {
            rwlock_write_unlock(&namespace_rwlock);
            ipc_answer_0(rid, ENOMEM);
            free(old);
            free(new);
            return;
        }
        futex_down(&nodes_futex);
        new_node->lnkcnt--;
        futex_up(&nodes_futex);
        break;
    default:
        rwlock_write_unlock(&namespace_rwlock);
        ipc_answer_0(rid, ENOTEMPTY);
        free(old);
        free(new);
        return;
    }
    /* Create the new link for the new name. */
    rc = vfs_lookup_internal(newc, L_LINK, NULL, NULL, old_node->index);
    if (rc != EOK) {
        rwlock_write_unlock(&namespace_rwlock);
        if (new_node)
            vfs_node_put(new_node);
        ipc_answer_0(rid, rc);
        free(old);
        free(new);
        return;
    }
    futex_down(&nodes_futex);
    old_node->lnkcnt++;
    futex_up(&nodes_futex);
    /* Destroy the link for the old name. */
    rc = vfs_lookup_internal(oldc, L_UNLINK, NULL, NULL);
    if (rc != EOK) {
        rwlock_write_unlock(&namespace_rwlock);
        vfs_node_put(old_node);
        if (new_node)
            vfs_node_put(new_node);
        ipc_answer_0(rid, rc);
        free(old);
        free(new);
        return;
    }
    futex_down(&nodes_futex);
    old_node->lnkcnt--;
    futex_up(&nodes_futex);
    rwlock_write_unlock(&namespace_rwlock);
    vfs_node_put(old_node);
    if (new_node)
        vfs_node_put(new_node);
    free(old);
    free(new);
    ipc_answer_0(rid, EOK);
}
 
/**
 * @}
 */
 

Rev 2770	Rev 2782
1	/*	1	/*
2	* Copyright (c) 2008 Jakub Jermar	2	* Copyright (c) 2008 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 fs	29	/** @addtogroup fs
30	* @{	30	* @{
31	*/	31	*/
32		32
33	/**	33	/**
34	* @file vfs_ops.c	34	* @file vfs_ops.c
35	* @brief Operations that VFS offers to its clients.	35	* @brief Operations that VFS offers to its clients.
36	*/	36	*/
37		37
38	#include "vfs.h"	38	#include "vfs.h"
39	#include <ipc/ipc.h>	39	#include <ipc/ipc.h>
40	#include <async.h>	40	#include <async.h>
41	#include <errno.h>	41	#include <errno.h>
42	#include <stdio.h>	42	#include <stdio.h>
43	#include <stdlib.h>	43	#include <stdlib.h>
44	#include <string.h>	44	#include <string.h>
45	#include <bool.h>	45	#include <bool.h>
46	#include <futex.h>	46	#include <futex.h>
47	#include <rwlock.h>	47	#include <rwlock.h>
48	#include <libadt/list.h>	48	#include <libadt/list.h>
49	#include <unistd.h>	49	#include <unistd.h>
50	#include <ctype.h>	50	#include <ctype.h>
51	#include <fcntl.h>	51	#include <fcntl.h>
52	#include <assert.h>	52	#include <assert.h>
53	#include <vfs/canonify.h>	53	#include <vfs/canonify.h>
54		54
55	/* Forward declarations of static functions. */	55	/* Forward declarations of static functions. */
56	static int vfs_truncate_internal(fs_handle_t, dev_handle_t, fs_index_t, size_t);	56	static int vfs_truncate_internal(fs_handle_t, dev_handle_t, fs_index_t, size_t);
57		57
58	/**	58	/**
59	* This rwlock prevents the race between a triplet-to-VFS-node resolution and a	59	* This rwlock prevents the race between a triplet-to-VFS-node resolution and a
60	* concurrent VFS operation which modifies the file system namespace.	60	* concurrent VFS operation which modifies the file system namespace.
61	*/	61	*/
62	RWLOCK_INITIALIZE(namespace_rwlock);	62	RWLOCK_INITIALIZE(namespace_rwlock);
63		63
64	futex_t rootfs_futex = FUTEX_INITIALIZER;	64	futex_t rootfs_futex = FUTEX_INITIALIZER;
65	vfs_triplet_t rootfs = {	65	vfs_triplet_t rootfs = {
66	.fs_handle = 0,	66	.fs_handle = 0,
67	.dev_handle = 0,	67	.dev_handle = 0,
68	.index = 0,	68	.index = 0,
69	};	69	};
70		70
71	static int	71	static int
72	lookup_root(fs_handle_t fs_handle, dev_handle_t dev_handle,	72	lookup_root(fs_handle_t fs_handle, dev_handle_t dev_handle,
73	vfs_lookup_res_t *result)	73	vfs_lookup_res_t *result)
74	{	74	{
75	vfs_pair_t altroot = {	75	vfs_pair_t altroot = {
76	.fs_handle = fs_handle,	76	.fs_handle = fs_handle,
77	.dev_handle = dev_handle,	77	.dev_handle = dev_handle,
78	};	78	};
79		79
80	return vfs_lookup_internal("/", L_DIRECTORY, result, &altroot);	80	return vfs_lookup_internal("/", L_DIRECTORY, result, &altroot);
81	}	81	}
82		82
83	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)	83	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
84	{	84	{
85	dev_handle_t dev_handle;	85	dev_handle_t dev_handle;
86	vfs_node_t *mp_node = NULL;	86	vfs_node_t *mp_node = NULL;
87		87
88	/*	88	/*
89	* We expect the library to do the device-name to device-handle	89	* We expect the library to do the device-name to device-handle
90	* translation for us, thus the device handle will arrive as ARG1	90	* translation for us, thus the device handle will arrive as ARG1
91	* in the request.	91	* in the request.
92	*/	92	*/
93	dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);	93	dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);
94		94
95	/*	95	/*
96	* For now, don't make use of ARG2 and ARG3, but they can be used to	96	* For now, don't make use of ARG2 and ARG3, but they can be used to
97	* carry mount options in the future.	97	* carry mount options in the future.
98	*/	98	*/
99		99
100	ipc_callid_t callid;	100	ipc_callid_t callid;
101	size_t size;	101	size_t size;
102		102
103	/*	103	/*
104	* Now, we expect the client to send us data with the name of the file	104	* Now, we expect the client to send us data with the name of the file
105	* system.	105	* system.
106	*/	106	*/
107	if (!ipc_data_write_receive(&callid, &size)) {	107	if (!ipc_data_write_receive(&callid, &size)) {
108	ipc_answer_0(callid, EINVAL);	108	ipc_answer_0(callid, EINVAL);
109	ipc_answer_0(rid, EINVAL);	109	ipc_answer_0(rid, EINVAL);
110	return;	110	return;
111	}	111	}
112		112
113	/*	113	/*
114	* Don't receive more than is necessary for storing a full file system	114	* Don't receive more than is necessary for storing a full file system
115	* name.	115	* name.
116	*/	116	*/
117	if (size < 1 \|\| size > FS_NAME_MAXLEN) {	117	if (size < 1 \|\| size > FS_NAME_MAXLEN) {
118	ipc_answer_0(callid, EINVAL);	118	ipc_answer_0(callid, EINVAL);
119	ipc_answer_0(rid, EINVAL);	119	ipc_answer_0(rid, EINVAL);
120	return;	120	return;
121	}	121	}
122		122
123	/* Deliver the file system name. */	123	/* Deliver the file system name. */
124	char fs_name[FS_NAME_MAXLEN + 1];	124	char fs_name[FS_NAME_MAXLEN + 1];
125	(void) ipc_data_write_finalize(callid, fs_name, size);	125	(void) ipc_data_write_finalize(callid, fs_name, size);
126	fs_name[size] = '\0';	126	fs_name[size] = '\0';
127		127
128	/*	128	/*
129	* Check if we know a file system with the same name as is in fs_name.	129	* Check if we know a file system with the same name as is in fs_name.
130	* This will also give us its file system handle.	130	* This will also give us its file system handle.
131	*/	131	*/
132	fs_handle_t fs_handle = fs_name_to_handle(fs_name, true);	132	fs_handle_t fs_handle = fs_name_to_handle(fs_name, true);
133	if (!fs_handle) {	133	if (!fs_handle) {
134	ipc_answer_0(rid, ENOENT);	134	ipc_answer_0(rid, ENOENT);
135	return;	135	return;
136	}	136	}
137		137
138	/* Now, we want the client to send us the mount point. */	138	/* Now, we want the client to send us the mount point. */
139	if (!ipc_data_write_receive(&callid, &size)) {	139	if (!ipc_data_write_receive(&callid, &size)) {
140	ipc_answer_0(callid, EINVAL);	140	ipc_answer_0(callid, EINVAL);
141	ipc_answer_0(rid, EINVAL);	141	ipc_answer_0(rid, EINVAL);
142	return;	142	return;
143	}	143	}
144		144
145	/* Check whether size is reasonable wrt. the mount point. */	145	/* Check whether size is reasonable wrt. the mount point. */
146	if (size < 1 \|\| size > MAX_PATH_LEN) {	146	if (size < 1 \|\| size > MAX_PATH_LEN) {
147	ipc_answer_0(callid, EINVAL);	147	ipc_answer_0(callid, EINVAL);
148	ipc_answer_0(rid, EINVAL);	148	ipc_answer_0(rid, EINVAL);
149	return;	149	return;
150	}	150	}
151	/* Allocate buffer for the mount point data being received. */	151	/* Allocate buffer for the mount point data being received. */
152	uint8_t *buf;	152	uint8_t *buf;
153	buf = malloc(size + 1);	153	buf = malloc(size + 1);
154	if (!buf) {	154	if (!buf) {
155	ipc_answer_0(callid, ENOMEM);	155	ipc_answer_0(callid, ENOMEM);
156	ipc_answer_0(rid, ENOMEM);	156	ipc_answer_0(rid, ENOMEM);
157	return;	157	return;
158	}	158	}
159		159
160	/* Deliver the mount point. */	160	/* Deliver the mount point. */
161	(void) ipc_data_write_finalize(callid, buf, size);	161	(void) ipc_data_write_finalize(callid, buf, size);
162	buf[size] = '\0';	162	buf[size] = '\0';
163		163
164	/*	164	/*
165	* Lookup the root node of the filesystem being mounted.	165	* Lookup the root node of the filesystem being mounted.
166	* In this case, we don't need to take the namespace_futex as the root	166	* In this case, we don't need to take the namespace_futex as the root
167	* node cannot be removed. However, we do take a reference to it so	167	* node cannot be removed. However, we do take a reference to it so
168	* that we can track how many times it has been mounted.	168	* that we can track how many times it has been mounted.
169	*/	169	*/
170	int rc;	170	int rc;
171	vfs_lookup_res_t mr_res;	171	vfs_lookup_res_t mr_res;
172	rc = lookup_root(fs_handle, dev_handle, &mr_res);	172	rc = lookup_root(fs_handle, dev_handle, &mr_res);
173	if (rc != EOK) {	173	if (rc != EOK) {
174	free(buf);	174	free(buf);
175	ipc_answer_0(rid, rc);	175	ipc_answer_0(rid, rc);
176	return;	176	return;
177	}	177	}
178	vfs_node_t *mr_node = vfs_node_get(&mr_res);	178	vfs_node_t *mr_node = vfs_node_get(&mr_res);
179	if (!mr_node) {	179	if (!mr_node) {
180	free(buf);	180	free(buf);
181	ipc_answer_0(rid, ENOMEM);	181	ipc_answer_0(rid, ENOMEM);
182	return;	182	return;
183	}	183	}
184		184
185	/* Finally, we need to resolve the path to the mountpoint. */	185	/* Finally, we need to resolve the path to the mountpoint. */
186	vfs_lookup_res_t mp_res;	186	vfs_lookup_res_t mp_res;
187	futex_down(&rootfs_futex);	187	futex_down(&rootfs_futex);
188	if (rootfs.fs_handle) {	188	if (rootfs.fs_handle) {
189	/* We already have the root FS. */	189	/* We already have the root FS. */
190	rwlock_write_lock(&namespace_rwlock);	190	rwlock_write_lock(&namespace_rwlock);
191	rc = vfs_lookup_internal(buf, L_DIRECTORY, &mp_res, NULL);	191	rc = vfs_lookup_internal(buf, L_DIRECTORY, &mp_res, NULL);
192	if (rc != EOK) {	192	if (rc != EOK) {
193	/* The lookup failed for some reason. */	193	/* The lookup failed for some reason. */
194	rwlock_write_unlock(&namespace_rwlock);	194	rwlock_write_unlock(&namespace_rwlock);
195	futex_up(&rootfs_futex);	195	futex_up(&rootfs_futex);
196	vfs_node_put(mr_node); /* failed -> drop reference */	196	vfs_node_put(mr_node); /* failed -> drop reference */
197	free(buf);	197	free(buf);
198	ipc_answer_0(rid, rc);	198	ipc_answer_0(rid, rc);
199	return;	199	return;
200	}	200	}
201	mp_node = vfs_node_get(&mp_res);	201	mp_node = vfs_node_get(&mp_res);
202	if (!mp_node) {	202	if (!mp_node) {
203	rwlock_write_unlock(&namespace_rwlock);	203	rwlock_write_unlock(&namespace_rwlock);
204	futex_up(&rootfs_futex);	204	futex_up(&rootfs_futex);
205	vfs_node_put(mr_node); /* failed -> drop reference */	205	vfs_node_put(mr_node); /* failed -> drop reference */
206	free(buf);	206	free(buf);
207	ipc_answer_0(rid, ENOMEM);	207	ipc_answer_0(rid, ENOMEM);
208	return;	208	return;
209	}	209	}
210	/*	210	/*
211	* Now we hold a reference to mp_node.	211	* Now we hold a reference to mp_node.
212	* It will be dropped upon the corresponding VFS_UNMOUNT.	212	* It will be dropped upon the corresponding VFS_UNMOUNT.
213	* This prevents the mount point from being deleted.	213	* This prevents the mount point from being deleted.
214	*/	214	*/
215	rwlock_write_unlock(&namespace_rwlock);	215	rwlock_write_unlock(&namespace_rwlock);
216	} else {	216	} else {
217	/* We still don't have the root file system mounted. */	217	/* We still don't have the root file system mounted. */
218	if ((size == 1) && (buf[0] == '/')) {	218	if ((size == 1) && (buf[0] == '/')) {
219	/* For this simple, but important case, we are done. */	219	/* For this simple, but important case, we are done. */
220	rootfs = mr_res.triplet;	220	rootfs = mr_res.triplet;
221	futex_up(&rootfs_futex);	221	futex_up(&rootfs_futex);
222	free(buf);	222	free(buf);
223	ipc_answer_0(rid, EOK);	223	ipc_answer_0(rid, EOK);
224	return;	224	return;
225	} else {	225	} else {
226	/*	226	/*
227	* We can't resolve this without the root filesystem	227	* We can't resolve this without the root filesystem
228	* being mounted first.	228	* being mounted first.
229	*/	229	*/
230	futex_up(&rootfs_futex);	230	futex_up(&rootfs_futex);
231	free(buf);	231	free(buf);
232	vfs_node_put(mr_node); /* failed -> drop reference */	232	vfs_node_put(mr_node); /* failed -> drop reference */
233	ipc_answer_0(rid, ENOENT);	233	ipc_answer_0(rid, ENOENT);
234	return;	234	return;
235	}	235	}
236	}	236	}
237	futex_up(&rootfs_futex);	237	futex_up(&rootfs_futex);
238		238
239	free(buf); /* The buffer is not needed anymore. */	239	free(buf); /* The buffer is not needed anymore. */
240		240
241	/*	241	/*
242	* At this point, we have all necessary pieces: file system and device	242	* At this point, we have all necessary pieces: file system and device
243	* handles, and we know the mount point VFS node and also the root node	243	* handles, and we know the mount point VFS node and also the root node
244	* of the file system being mounted.	244	* of the file system being mounted.
245	*/	245	*/
246		246
247	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);	247	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);
248	/* Later we can use ARG3 to pass mode/flags. */	248	/* Later we can use ARG3 to pass mode/flags. */
249	aid_t req1 = async_send_3(phone, VFS_MOUNT,	249	aid_t req1 = async_send_3(phone, VFS_MOUNT,
250	(ipcarg_t) mp_res.triplet.dev_handle,	250	(ipcarg_t) mp_res.triplet.dev_handle,
251	(ipcarg_t) mp_res.triplet.index, 0, NULL);	251	(ipcarg_t) mp_res.triplet.index, 0, NULL);
252	/* The second call uses the same method. */	252	/* The second call uses the same method. */
253	aid_t req2 = async_send_3(phone, VFS_MOUNT,	253	aid_t req2 = async_send_3(phone, VFS_MOUNT,
254	(ipcarg_t) mr_res.triplet.fs_handle,	254	(ipcarg_t) mr_res.triplet.fs_handle,
255	(ipcarg_t) mr_res.triplet.dev_handle,	255	(ipcarg_t) mr_res.triplet.dev_handle,
256	(ipcarg_t) mr_res.triplet.index, NULL);	256	(ipcarg_t) mr_res.triplet.index, NULL);
257	vfs_release_phone(phone);	257	vfs_release_phone(phone);
258		258
259	ipcarg_t rc1;	259	ipcarg_t rc1;
260	ipcarg_t rc2;	260	ipcarg_t rc2;
261	async_wait_for(req1, &rc1);	261	async_wait_for(req1, &rc1);
262	async_wait_for(req2, &rc2);	262	async_wait_for(req2, &rc2);
263		263
264	if ((rc1 != EOK) \|\| (rc2 != EOK)) {	264	if ((rc1 != EOK) \|\| (rc2 != EOK)) {
265	/* Mount failed, drop references to mr_node and mp_node. */	265	/* Mount failed, drop references to mr_node and mp_node. */
266	vfs_node_put(mr_node);	266	vfs_node_put(mr_node);
267	if (mp_node)	267	if (mp_node)
268	vfs_node_put(mp_node);	268	vfs_node_put(mp_node);
269	}	269	}
270		270
271	if (rc2 == EOK)	271	if (rc2 == EOK)
272	ipc_answer_0(rid, rc1);	272	ipc_answer_0(rid, rc1);
273	else if (rc1 == EOK)	273	else if (rc1 == EOK)
274	ipc_answer_0(rid, rc2);	274	ipc_answer_0(rid, rc2);
275	else	275	else
276	ipc_answer_0(rid, rc1);	276	ipc_answer_0(rid, rc1);
277	}	277	}
278		278
279	void vfs_open(ipc_callid_t rid, ipc_call_t *request)	279	void vfs_open(ipc_callid_t rid, ipc_call_t *request)
280	{	280	{
281	if (!vfs_files_init()) {	281	if (!vfs_files_init()) {
282	ipc_answer_0(rid, ENOMEM);	282	ipc_answer_0(rid, ENOMEM);
283	return;	283	return;
284	}	284	}
285		285
286	/*	286	/*
287	* The POSIX interface is open(path, oflag, mode).	287	* The POSIX interface is open(path, oflag, mode).
288	* We can receive oflags and mode along with the VFS_OPEN call; the path	288	* We can receive oflags and mode along with the VFS_OPEN call; the path
289	* will need to arrive in another call.	289	* will need to arrive in another call.
290	*	290	*
291	* We also receive one private, non-POSIX set of flags called lflag	291	* We also receive one private, non-POSIX set of flags called lflag
292	* used to pass information to vfs_lookup_internal().	292	* used to pass information to vfs_lookup_internal().
293	*/	293	*/
294	int lflag = IPC_GET_ARG1(*request);	294	int lflag = IPC_GET_ARG1(*request);
295	int oflag = IPC_GET_ARG2(*request);	295	int oflag = IPC_GET_ARG2(*request);
296	int mode = IPC_GET_ARG3(*request);	296	int mode = IPC_GET_ARG3(*request);
297	size_t len;	297	size_t len;
298		298
299	if (oflag & O_CREAT)	299	if (oflag & O_CREAT)
300	lflag \|= L_CREATE;	300	lflag \|= L_CREATE;
301	if (oflag & O_EXCL)	301	if (oflag & O_EXCL)
302	lflag \|= L_EXCLUSIVE;	302	lflag \|= L_EXCLUSIVE;
303		303
304	ipc_callid_t callid;	304	ipc_callid_t callid;
305		305
306	if (!ipc_data_write_receive(&callid, &len)) {	306	if (!ipc_data_write_receive(&callid, &len)) {
307	ipc_answer_0(callid, EINVAL);	307	ipc_answer_0(callid, EINVAL);
308	ipc_answer_0(rid, EINVAL);	308	ipc_answer_0(rid, EINVAL);
309	return;	309	return;
310	}	310	}
311	char *path = malloc(len + 1);	311	char *path = malloc(len + 1);
312	if (!path) {	312	if (!path) {
313	ipc_answer_0(callid, ENOMEM);	313	ipc_answer_0(callid, ENOMEM);
314	ipc_answer_0(rid, ENOMEM);	314	ipc_answer_0(rid, ENOMEM);
315	return;	315	return;
316	}	316	}
317	int rc;	317	int rc;
318	if ((rc = ipc_data_write_finalize(callid, path, len))) {	318	if ((rc = ipc_data_write_finalize(callid, path, len))) {
319	ipc_answer_0(rid, rc);	319	ipc_answer_0(rid, rc);
320	free(path);	320	free(path);
321	return;	321	return;
322	}	322	}
323	path[len] = '\0';	323	path[len] = '\0';
324		324
325	/*	325	/*
326	* Avoid the race condition in which the file can be deleted before we	326	* Avoid the race condition in which the file can be deleted before we
327	* find/create-and-lock the VFS node corresponding to the looked-up	327	* find/create-and-lock the VFS node corresponding to the looked-up
328	* triplet.	328	* triplet.
329	*/	329	*/
330	if (lflag & L_CREATE)	330	if (lflag & L_CREATE)
331	rwlock_write_lock(&namespace_rwlock);	331	rwlock_write_lock(&namespace_rwlock);
332	else	332	else
333	rwlock_read_lock(&namespace_rwlock);	333	rwlock_read_lock(&namespace_rwlock);
334		334
335	/* The path is now populated and we can call vfs_lookup_internal(). */	335	/* The path is now populated and we can call vfs_lookup_internal(). */
336	vfs_lookup_res_t lr;	336	vfs_lookup_res_t lr;
337	rc = vfs_lookup_internal(path, lflag, &lr, NULL);	337	rc = vfs_lookup_internal(path, lflag, &lr, NULL);
338	if (rc) {	338	if (rc) {
339	if (lflag & L_CREATE)	339	if (lflag & L_CREATE)
340	rwlock_write_unlock(&namespace_rwlock);	340	rwlock_write_unlock(&namespace_rwlock);
341	else	341	else
342	rwlock_read_unlock(&namespace_rwlock);	342	rwlock_read_unlock(&namespace_rwlock);
343	ipc_answer_0(rid, rc);	343	ipc_answer_0(rid, rc);
344	free(path);	344	free(path);
345	return;	345	return;
346	}	346	}
347		347
348	/* Path is no longer needed. */	348	/* Path is no longer needed. */
349	free(path);	349	free(path);
350		350
351	vfs_node_t *node = vfs_node_get(&lr);	351	vfs_node_t *node = vfs_node_get(&lr);
352	if (lflag & L_CREATE)	352	if (lflag & L_CREATE)
353	rwlock_write_unlock(&namespace_rwlock);	353	rwlock_write_unlock(&namespace_rwlock);
354	else	354	else
355	rwlock_read_unlock(&namespace_rwlock);	355	rwlock_read_unlock(&namespace_rwlock);
356		356
357	/* Truncate the file if requested and if necessary. */	357	/* Truncate the file if requested and if necessary. */
358	if (oflag & O_TRUNC) {	358	if (oflag & O_TRUNC) {
359	rwlock_write_lock(&node->contents_rwlock);	359	rwlock_write_lock(&node->contents_rwlock);
360	if (node->size) {	360	if (node->size) {
361	rc = vfs_truncate_internal(node->fs_handle,	361	rc = vfs_truncate_internal(node->fs_handle,
362	node->dev_handle, node->index, 0);	362	node->dev_handle, node->index, 0);
363	if (rc) {	363	if (rc) {
364	rwlock_write_unlock(&node->contents_rwlock);	364	rwlock_write_unlock(&node->contents_rwlock);
365	vfs_node_put(node);	365	vfs_node_put(node);
366	ipc_answer_0(rid, rc);	366	ipc_answer_0(rid, rc);
367	return;	367	return;
368	}	368	}
369	node->size = 0;	369	node->size = 0;
370	}	370	}
371	rwlock_write_unlock(&node->contents_rwlock);	371	rwlock_write_unlock(&node->contents_rwlock);
372	}	372	}
373		373
374	/*	374	/*
375	* Get ourselves a file descriptor and the corresponding vfs_file_t	375	* Get ourselves a file descriptor and the corresponding vfs_file_t
376	* structure.	376	* structure.
377	*/	377	*/
378	int fd = vfs_fd_alloc();	378	int fd = vfs_fd_alloc();
379	if (fd < 0) {	379	if (fd < 0) {
380	vfs_node_put(node);	380	vfs_node_put(node);
381	ipc_answer_0(rid, fd);	381	ipc_answer_0(rid, fd);
382	return;	382	return;
383	}	383	}
384	vfs_file_t *file = vfs_file_get(fd);	384	vfs_file_t *file = vfs_file_get(fd);
385	file->node = node;	385	file->node = node;
386	if (oflag & O_APPEND)	386	if (oflag & O_APPEND)
387	file->append = true;	387	file->append = true;
388		388
389	/*	389	/*
390	* The following increase in reference count is for the fact that the	390	* The following increase in reference count is for the fact that the
391	* file is being opened and that a file structure is pointing to it.	391	* file is being opened and that a file structure is pointing to it.
392	* It is necessary so that the file will not disappear when	392	* It is necessary so that the file will not disappear when
393	* vfs_node_put() is called. The reference will be dropped by the	393	* vfs_node_put() is called. The reference will be dropped by the
394	* respective VFS_CLOSE.	394	* respective VFS_CLOSE.
395	*/	395	*/
396	vfs_node_addref(node);	396	vfs_node_addref(node);
397	vfs_node_put(node);	397	vfs_node_put(node);
398		398
399	/* Success! Return the new file descriptor to the client. */	399	/* Success! Return the new file descriptor to the client. */
400	ipc_answer_1(rid, EOK, fd);	400	ipc_answer_1(rid, EOK, fd);
401	}	401	}
402		402
403	void vfs_close(ipc_callid_t rid, ipc_call_t *request)	403	void vfs_close(ipc_callid_t rid, ipc_call_t *request)
404	{	404	{
405	int fd = IPC_GET_ARG1(*request);	405	int fd = IPC_GET_ARG1(*request);
406	if (fd >= MAX_OPEN_FILES) {	406	if (fd >= MAX_OPEN_FILES) {
407	ipc_answer_0(rid, EBADF);	407	ipc_answer_0(rid, EBADF);
408	return;	408	return;
409	}	409	}
410	vfs_fd_free(fd);	410	vfs_fd_free(fd);
411	ipc_answer_0(rid, EOK);	411	ipc_answer_0(rid, EOK);
412	}	412	}
413		413
414	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)	414	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
415	{	415	{
416		416
417	/*	417	/*
418	* The following code strongly depends on the fact that the files data	418	* The following code strongly depends on the fact that the files data
419	* structure can be only accessed by a single fibril and all file	419	* structure can be only accessed by a single fibril and all file
420	* operations are serialized (i.e. the reads and writes cannot	420	* operations are serialized (i.e. the reads and writes cannot
421	* interleave and a file cannot be closed while it is being read).	421	* interleave and a file cannot be closed while it is being read).
422	*	422	*
423	* Additional synchronization needs to be added once the table of	423	* Additional synchronization needs to be added once the table of
424	* open files supports parallel access!	424	* open files supports parallel access!
425	*/	425	*/
426		426
427	int fd = IPC_GET_ARG1(*request);	427	int fd = IPC_GET_ARG1(*request);
428		428
429	/* Lookup the file structure corresponding to the file descriptor. */	429	/* Lookup the file structure corresponding to the file descriptor. */
430	vfs_file_t *file = vfs_file_get(fd);	430	vfs_file_t *file = vfs_file_get(fd);
431	if (!file) {	431	if (!file) {
432	ipc_answer_0(rid, ENOENT);	432	ipc_answer_0(rid, ENOENT);
433	return;	433	return;
434	}	434	}
435		435
436	/*	436	/*
437	* Now we need to receive a call with client's	437	* Now we need to receive a call with client's
438	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.	438	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
439	*/	439	*/
440	ipc_callid_t callid;	440	ipc_callid_t callid;
441	int res;	441	int res;
442	if (read)	442	if (read)
443	res = ipc_data_read_receive(&callid, NULL);	443	res = ipc_data_read_receive(&callid, NULL);
444	else	444	else
445	res = ipc_data_write_receive(&callid, NULL);	445	res = ipc_data_write_receive(&callid, NULL);
446	if (!res) {	446	if (!res) {
447	ipc_answer_0(callid, EINVAL);	447	ipc_answer_0(callid, EINVAL);
448	ipc_answer_0(rid, EINVAL);	448	ipc_answer_0(rid, EINVAL);
449	return;	449	return;
450	}	450	}
451		451
452	/*	452	/*
453	* Lock the open file structure so that no other thread can manipulate	453	* Lock the open file structure so that no other thread can manipulate
454	* the same open file at a time.	454	* the same open file at a time.
455	*/	455	*/
456	futex_down(&file->lock);	456	futex_down(&file->lock);
457		457
458	/*	458	/*
459	* Lock the file's node so that no other client can read/write to it at	459	* Lock the file's node so that no other client can read/write to it at
460	* the same time.	460	* the same time.
461	*/	461	*/
462	if (read)	462	if (read)
463	rwlock_read_lock(&file->node->contents_rwlock);	463	rwlock_read_lock(&file->node->contents_rwlock);
464	else	464	else
465	rwlock_write_lock(&file->node->contents_rwlock);	465	rwlock_write_lock(&file->node->contents_rwlock);
466		466
467	int fs_phone = vfs_grab_phone(file->node->fs_handle);	467	int fs_phone = vfs_grab_phone(file->node->fs_handle);
468		468
469	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */	469	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
470	aid_t msg;	470	aid_t msg;
471	ipc_call_t answer;	471	ipc_call_t answer;
472	if (!read && file->append)	472	if (!read && file->append)
473	file->pos = file->node->size;	473	file->pos = file->node->size;
474	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),	474	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
475	file->node->dev_handle, file->node->index, file->pos, &answer);	475	file->node->dev_handle, file->node->index, file->pos, &answer);
476		476
477	/*	477	/*
478	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the	478	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
479	* destination FS server. The call will be routed as if sent by	479	* destination FS server. The call will be routed as if sent by
480	* ourselves. Note that call arguments are immutable in this case so we	480	* ourselves. Note that call arguments are immutable in this case so we
481	* don't have to bother.	481	* don't have to bother.
482	*/	482	*/
483	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);	483	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
484		484
485	vfs_release_phone(fs_phone);	485	vfs_release_phone(fs_phone);
486		486
487	/* Wait for reply from the FS server. */	487	/* Wait for reply from the FS server. */
488	ipcarg_t rc;	488	ipcarg_t rc;
489	async_wait_for(msg, &rc);	489	async_wait_for(msg, &rc);
490	size_t bytes = IPC_GET_ARG1(answer);	490	size_t bytes = IPC_GET_ARG1(answer);
491		491
492	/* Unlock the VFS node. */	492	/* Unlock the VFS node. */
493	if (read)	493	if (read)
494	rwlock_read_unlock(&file->node->contents_rwlock);	494	rwlock_read_unlock(&file->node->contents_rwlock);
495	else {	495	else {
496	/* Update the cached version of node's size. */	496	/* Update the cached version of node's size. */
497	if (rc == EOK)	497	if (rc == EOK)
498	file->node->size = IPC_GET_ARG2(answer);	498	file->node->size = IPC_GET_ARG2(answer);
499	rwlock_write_unlock(&file->node->contents_rwlock);	499	rwlock_write_unlock(&file->node->contents_rwlock);
500	}	500	}

Subversion Repositories HelenOS

(root)/tags/0.3.0/uspace/srv/vfs/vfs_ops.c – Rev 2770 → 2782