WebSVN – HelenOS – Diff – /trunk/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(int, int, unsigned long, 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(int fs_handle, int 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)
{
    int 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 = 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.
     */
    int 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(int fs_handle, int dev_handle, unsigned long 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 2763	Rev 2766
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 <atomic.h>	-
54	#include <vfs/canonify.h>	53	#include <vfs/canonify.h>
55		54
56	/* Forward declarations of static functions. */	55	/* Forward declarations of static functions. */
57	static int vfs_truncate_internal(int, int, unsigned long, size_t);	56	static int vfs_truncate_internal(int, int, unsigned long, size_t);
58		57
59	/**	58	/**
60	* 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
61	* concurrent VFS operation which modifies the file system namespace.	60	* concurrent VFS operation which modifies the file system namespace.
62	*/	61	*/
63	RWLOCK_INITIALIZE(namespace_rwlock);	62	RWLOCK_INITIALIZE(namespace_rwlock);
64		63
65	atomic_t rootfs_futex = FUTEX_INITIALIZER;	64	futex_t rootfs_futex = FUTEX_INITIALIZER;
66	vfs_triplet_t rootfs = {	65	vfs_triplet_t rootfs = {
67	.fs_handle = 0,	66	.fs_handle = 0,
68	.dev_handle = 0,	67	.dev_handle = 0,
69	.index = 0,	68	.index = 0,
70	};	69	};
71		70
72	static int lookup_root(int fs_handle, int dev_handle, vfs_lookup_res_t *result)	71	static int lookup_root(int fs_handle, int dev_handle, vfs_lookup_res_t *result)
73	{	72	{
74	vfs_pair_t altroot = {	73	vfs_pair_t altroot = {
75	.fs_handle = fs_handle,	74	.fs_handle = fs_handle,
76	.dev_handle = dev_handle,	75	.dev_handle = dev_handle,
77	};	76	};
78		77
79	return vfs_lookup_internal("/", L_DIRECTORY, result, &altroot);	78	return vfs_lookup_internal("/", L_DIRECTORY, result, &altroot);
80	}	79	}
81		80
82	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)	81	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
83	{	82	{
84	int dev_handle;	83	int dev_handle;
85	vfs_node_t *mp_node = NULL;	84	vfs_node_t *mp_node = NULL;
86		85
87	/*	86	/*
88	* We expect the library to do the device-name to device-handle	87	* We expect the library to do the device-name to device-handle
89	* translation for us, thus the device handle will arrive as ARG1	88	* translation for us, thus the device handle will arrive as ARG1
90	* in the request.	89	* in the request.
91	*/	90	*/
92	dev_handle = IPC_GET_ARG1(*request);	91	dev_handle = IPC_GET_ARG1(*request);
93		92
94	/*	93	/*
95	* For now, don't make use of ARG2 and ARG3, but they can be used to	94	* For now, don't make use of ARG2 and ARG3, but they can be used to
96	* carry mount options in the future.	95	* carry mount options in the future.
97	*/	96	*/
98		97
99	ipc_callid_t callid;	98	ipc_callid_t callid;
100	size_t size;	99	size_t size;
101		100
102	/*	101	/*
103	* Now, we expect the client to send us data with the name of the file	102	* Now, we expect the client to send us data with the name of the file
104	* system.	103	* system.
105	*/	104	*/
106	if (!ipc_data_write_receive(&callid, &size)) {	105	if (!ipc_data_write_receive(&callid, &size)) {
107	ipc_answer_0(callid, EINVAL);	106	ipc_answer_0(callid, EINVAL);
108	ipc_answer_0(rid, EINVAL);	107	ipc_answer_0(rid, EINVAL);
109	return;	108	return;
110	}	109	}
111		110
112	/*	111	/*
113	* Don't receive more than is necessary for storing a full file system	112	* Don't receive more than is necessary for storing a full file system
114	* name.	113	* name.
115	*/	114	*/
116	if (size < 1 \|\| size > FS_NAME_MAXLEN) {	115	if (size < 1 \|\| size > FS_NAME_MAXLEN) {
117	ipc_answer_0(callid, EINVAL);	116	ipc_answer_0(callid, EINVAL);
118	ipc_answer_0(rid, EINVAL);	117	ipc_answer_0(rid, EINVAL);
119	return;	118	return;
120	}	119	}
121		120
122	/* Deliver the file system name. */	121	/* Deliver the file system name. */
123	char fs_name[FS_NAME_MAXLEN + 1];	122	char fs_name[FS_NAME_MAXLEN + 1];
124	(void) ipc_data_write_finalize(callid, fs_name, size);	123	(void) ipc_data_write_finalize(callid, fs_name, size);
125	fs_name[size] = '\0';	124	fs_name[size] = '\0';
126		125
127	/*	126	/*
128	* Check if we know a file system with the same name as is in fs_name.	127	* Check if we know a file system with the same name as is in fs_name.
129	* This will also give us its file system handle.	128	* This will also give us its file system handle.
130	*/	129	*/
131	int fs_handle = fs_name_to_handle(fs_name, true);	130	int fs_handle = fs_name_to_handle(fs_name, true);
132	if (!fs_handle) {	131	if (!fs_handle) {
133	ipc_answer_0(rid, ENOENT);	132	ipc_answer_0(rid, ENOENT);
134	return;	133	return;
135	}	134	}
136		135
137	/* Now, we want the client to send us the mount point. */	136	/* Now, we want the client to send us the mount point. */
138	if (!ipc_data_write_receive(&callid, &size)) {	137	if (!ipc_data_write_receive(&callid, &size)) {
139	ipc_answer_0(callid, EINVAL);	138	ipc_answer_0(callid, EINVAL);
140	ipc_answer_0(rid, EINVAL);	139	ipc_answer_0(rid, EINVAL);
141	return;	140	return;
142	}	141	}
143		142
144	/* Check whether size is reasonable wrt. the mount point. */	143	/* Check whether size is reasonable wrt. the mount point. */
145	if (size < 1 \|\| size > MAX_PATH_LEN) {	144	if (size < 1 \|\| size > MAX_PATH_LEN) {
146	ipc_answer_0(callid, EINVAL);	145	ipc_answer_0(callid, EINVAL);
147	ipc_answer_0(rid, EINVAL);	146	ipc_answer_0(rid, EINVAL);
148	return;	147	return;
149	}	148	}
150	/* Allocate buffer for the mount point data being received. */	149	/* Allocate buffer for the mount point data being received. */
151	uint8_t *buf;	150	uint8_t *buf;
152	buf = malloc(size + 1);	151	buf = malloc(size + 1);
153	if (!buf) {	152	if (!buf) {
154	ipc_answer_0(callid, ENOMEM);	153	ipc_answer_0(callid, ENOMEM);
155	ipc_answer_0(rid, ENOMEM);	154	ipc_answer_0(rid, ENOMEM);
156	return;	155	return;
157	}	156	}
158		157
159	/* Deliver the mount point. */	158	/* Deliver the mount point. */
160	(void) ipc_data_write_finalize(callid, buf, size);	159	(void) ipc_data_write_finalize(callid, buf, size);
161	buf[size] = '\0';	160	buf[size] = '\0';
162		161
163	/*	162	/*
164	* Lookup the root node of the filesystem being mounted.	163	* Lookup the root node of the filesystem being mounted.
165	* In this case, we don't need to take the namespace_futex as the root	164	* In this case, we don't need to take the namespace_futex as the root
166	* node cannot be removed. However, we do take a reference to it so	165	* node cannot be removed. However, we do take a reference to it so
167	* that we can track how many times it has been mounted.	166	* that we can track how many times it has been mounted.
168	*/	167	*/
169	int rc;	168	int rc;
170	vfs_lookup_res_t mr_res;	169	vfs_lookup_res_t mr_res;
171	rc = lookup_root(fs_handle, dev_handle, &mr_res);	170	rc = lookup_root(fs_handle, dev_handle, &mr_res);
172	if (rc != EOK) {	171	if (rc != EOK) {
173	free(buf);	172	free(buf);
174	ipc_answer_0(rid, rc);	173	ipc_answer_0(rid, rc);
175	return;	174	return;
176	}	175	}
177	vfs_node_t *mr_node = vfs_node_get(&mr_res);	176	vfs_node_t *mr_node = vfs_node_get(&mr_res);
178	if (!mr_node) {	177	if (!mr_node) {
179	free(buf);	178	free(buf);
180	ipc_answer_0(rid, ENOMEM);	179	ipc_answer_0(rid, ENOMEM);
181	return;	180	return;
182	}	181	}
183		182
184	/* Finally, we need to resolve the path to the mountpoint. */	183	/* Finally, we need to resolve the path to the mountpoint. */
185	vfs_lookup_res_t mp_res;	184	vfs_lookup_res_t mp_res;
186	futex_down(&rootfs_futex);	185	futex_down(&rootfs_futex);
187	if (rootfs.fs_handle) {	186	if (rootfs.fs_handle) {
188	/* We already have the root FS. */	187	/* We already have the root FS. */
189	rwlock_write_lock(&namespace_rwlock);	188	rwlock_write_lock(&namespace_rwlock);
190	rc = vfs_lookup_internal(buf, L_DIRECTORY, &mp_res, NULL);	189	rc = vfs_lookup_internal(buf, L_DIRECTORY, &mp_res, NULL);
191	if (rc != EOK) {	190	if (rc != EOK) {
192	/* The lookup failed for some reason. */	191	/* The lookup failed for some reason. */
193	rwlock_write_unlock(&namespace_rwlock);	192	rwlock_write_unlock(&namespace_rwlock);
194	futex_up(&rootfs_futex);	193	futex_up(&rootfs_futex);
195	vfs_node_put(mr_node); /* failed -> drop reference */	194	vfs_node_put(mr_node); /* failed -> drop reference */
196	free(buf);	195	free(buf);
197	ipc_answer_0(rid, rc);	196	ipc_answer_0(rid, rc);
198	return;	197	return;
199	}	198	}
200	mp_node = vfs_node_get(&mp_res);	199	mp_node = vfs_node_get(&mp_res);
201	if (!mp_node) {	200	if (!mp_node) {
202	rwlock_write_unlock(&namespace_rwlock);	201	rwlock_write_unlock(&namespace_rwlock);
203	futex_up(&rootfs_futex);	202	futex_up(&rootfs_futex);
204	vfs_node_put(mr_node); /* failed -> drop reference */	203	vfs_node_put(mr_node); /* failed -> drop reference */
205	free(buf);	204	free(buf);
206	ipc_answer_0(rid, ENOMEM);	205	ipc_answer_0(rid, ENOMEM);
207	return;	206	return;
208	}	207	}
209	/*	208	/*
210	* Now we hold a reference to mp_node.	209	* Now we hold a reference to mp_node.
211	* It will be dropped upon the corresponding VFS_UNMOUNT.	210	* It will be dropped upon the corresponding VFS_UNMOUNT.
212	* This prevents the mount point from being deleted.	211	* This prevents the mount point from being deleted.
213	*/	212	*/
214	rwlock_write_unlock(&namespace_rwlock);	213	rwlock_write_unlock(&namespace_rwlock);
215	} else {	214	} else {
216	/* We still don't have the root file system mounted. */	215	/* We still don't have the root file system mounted. */
217	if ((size == 1) && (buf[0] == '/')) {	216	if ((size == 1) && (buf[0] == '/')) {
218	/* For this simple, but important case, we are done. */	217	/* For this simple, but important case, we are done. */
219	rootfs = mr_res.triplet;	218	rootfs = mr_res.triplet;
220	futex_up(&rootfs_futex);	219	futex_up(&rootfs_futex);
221	free(buf);	220	free(buf);
222	ipc_answer_0(rid, EOK);	221	ipc_answer_0(rid, EOK);
223	return;	222	return;
224	} else {	223	} else {
225	/*	224	/*
226	* We can't resolve this without the root filesystem	225	* We can't resolve this without the root filesystem
227	* being mounted first.	226	* being mounted first.
228	*/	227	*/
229	futex_up(&rootfs_futex);	228	futex_up(&rootfs_futex);
230	free(buf);	229	free(buf);
231	vfs_node_put(mr_node); /* failed -> drop reference */	230	vfs_node_put(mr_node); /* failed -> drop reference */
232	ipc_answer_0(rid, ENOENT);	231	ipc_answer_0(rid, ENOENT);
233	return;	232	return;
234	}	233	}
235	}	234	}
236	futex_up(&rootfs_futex);	235	futex_up(&rootfs_futex);
237		236
238	free(buf); /* The buffer is not needed anymore. */	237	free(buf); /* The buffer is not needed anymore. */
239		238
240	/*	239	/*
241	* At this point, we have all necessary pieces: file system and device	240	* At this point, we have all necessary pieces: file system and device
242	* handles, and we know the mount point VFS node and also the root node	241	* handles, and we know the mount point VFS node and also the root node
243	* of the file system being mounted.	242	* of the file system being mounted.
244	*/	243	*/
245		244
246	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);	245	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);
247	/* Later we can use ARG3 to pass mode/flags. */	246	/* Later we can use ARG3 to pass mode/flags. */
248	aid_t req1 = async_send_3(phone, VFS_MOUNT,	247	aid_t req1 = async_send_3(phone, VFS_MOUNT,
249	(ipcarg_t) mp_res.triplet.dev_handle,	248	(ipcarg_t) mp_res.triplet.dev_handle,
250	(ipcarg_t) mp_res.triplet.index, 0, NULL);	249	(ipcarg_t) mp_res.triplet.index, 0, NULL);
251	/* The second call uses the same method. */	250	/* The second call uses the same method. */
252	aid_t req2 = async_send_3(phone, VFS_MOUNT,	251	aid_t req2 = async_send_3(phone, VFS_MOUNT,
253	(ipcarg_t) mr_res.triplet.fs_handle,	252	(ipcarg_t) mr_res.triplet.fs_handle,
254	(ipcarg_t) mr_res.triplet.dev_handle,	253	(ipcarg_t) mr_res.triplet.dev_handle,
255	(ipcarg_t) mr_res.triplet.index, NULL);	254	(ipcarg_t) mr_res.triplet.index, NULL);
256	vfs_release_phone(phone);	255	vfs_release_phone(phone);
257		256
258	ipcarg_t rc1;	257	ipcarg_t rc1;
259	ipcarg_t rc2;	258	ipcarg_t rc2;
260	async_wait_for(req1, &rc1);	259	async_wait_for(req1, &rc1);
261	async_wait_for(req2, &rc2);	260	async_wait_for(req2, &rc2);
262		261
263	if ((rc1 != EOK) \|\| (rc2 != EOK)) {	262	if ((rc1 != EOK) \|\| (rc2 != EOK)) {
264	/* Mount failed, drop references to mr_node and mp_node. */	263	/* Mount failed, drop references to mr_node and mp_node. */
265	vfs_node_put(mr_node);	264	vfs_node_put(mr_node);
266	if (mp_node)	265	if (mp_node)
267	vfs_node_put(mp_node);	266	vfs_node_put(mp_node);
268	}	267	}
269		268
270	if (rc2 == EOK)	269	if (rc2 == EOK)
271	ipc_answer_0(rid, rc1);	270	ipc_answer_0(rid, rc1);
272	else if (rc1 == EOK)	271	else if (rc1 == EOK)
273	ipc_answer_0(rid, rc2);	272	ipc_answer_0(rid, rc2);
274	else	273	else
275	ipc_answer_0(rid, rc1);	274	ipc_answer_0(rid, rc1);
276	}	275	}
277		276
278	void vfs_open(ipc_callid_t rid, ipc_call_t *request)	277	void vfs_open(ipc_callid_t rid, ipc_call_t *request)
279	{	278	{
280	if (!vfs_files_init()) {	279	if (!vfs_files_init()) {
281	ipc_answer_0(rid, ENOMEM);	280	ipc_answer_0(rid, ENOMEM);
282	return;	281	return;
283	}	282	}
284		283
285	/*	284	/*
286	* The POSIX interface is open(path, oflag, mode).	285	* The POSIX interface is open(path, oflag, mode).
287	* We can receive oflags and mode along with the VFS_OPEN call; the path	286	* We can receive oflags and mode along with the VFS_OPEN call; the path
288	* will need to arrive in another call.	287	* will need to arrive in another call.
289	*	288	*
290	* We also receive one private, non-POSIX set of flags called lflag	289	* We also receive one private, non-POSIX set of flags called lflag
291	* used to pass information to vfs_lookup_internal().	290	* used to pass information to vfs_lookup_internal().
292	*/	291	*/
293	int lflag = IPC_GET_ARG1(*request);	292	int lflag = IPC_GET_ARG1(*request);
294	int oflag = IPC_GET_ARG2(*request);	293	int oflag = IPC_GET_ARG2(*request);
295	int mode = IPC_GET_ARG3(*request);	294	int mode = IPC_GET_ARG3(*request);
296	size_t len;	295	size_t len;
297		296
298	if (oflag & O_CREAT)	297	if (oflag & O_CREAT)
299	lflag \|= L_CREATE;	298	lflag \|= L_CREATE;
300	if (oflag & O_EXCL)	299	if (oflag & O_EXCL)
301	lflag \|= L_EXCLUSIVE;	300	lflag \|= L_EXCLUSIVE;
302		301
303	ipc_callid_t callid;	302	ipc_callid_t callid;
304		303
305	if (!ipc_data_write_receive(&callid, &len)) {	304	if (!ipc_data_write_receive(&callid, &len)) {
306	ipc_answer_0(callid, EINVAL);	305	ipc_answer_0(callid, EINVAL);
307	ipc_answer_0(rid, EINVAL);	306	ipc_answer_0(rid, EINVAL);
308	return;	307	return;
309	}	308	}
310	char *path = malloc(len + 1);	309	char *path = malloc(len + 1);
311	if (!path) {	310	if (!path) {
312	ipc_answer_0(callid, ENOMEM);	311	ipc_answer_0(callid, ENOMEM);
313	ipc_answer_0(rid, ENOMEM);	312	ipc_answer_0(rid, ENOMEM);
314	return;	313	return;
315	}	314	}
316	int rc;	315	int rc;
317	if ((rc = ipc_data_write_finalize(callid, path, len))) {	316	if ((rc = ipc_data_write_finalize(callid, path, len))) {
318	ipc_answer_0(rid, rc);	317	ipc_answer_0(rid, rc);
319	free(path);	318	free(path);
320	return;	319	return;
321	}	320	}
322	path[len] = '\0';	321	path[len] = '\0';
323		322
324	/*	323	/*
325	* Avoid the race condition in which the file can be deleted before we	324	* Avoid the race condition in which the file can be deleted before we
326	* find/create-and-lock the VFS node corresponding to the looked-up	325	* find/create-and-lock the VFS node corresponding to the looked-up
327	* triplet.	326	* triplet.
328	*/	327	*/
329	if (lflag & L_CREATE)	328	if (lflag & L_CREATE)
330	rwlock_write_lock(&namespace_rwlock);	329	rwlock_write_lock(&namespace_rwlock);
331	else	330	else
332	rwlock_read_lock(&namespace_rwlock);	331	rwlock_read_lock(&namespace_rwlock);
333		332
334	/* The path is now populated and we can call vfs_lookup_internal(). */	333	/* The path is now populated and we can call vfs_lookup_internal(). */
335	vfs_lookup_res_t lr;	334	vfs_lookup_res_t lr;
336	rc = vfs_lookup_internal(path, lflag, &lr, NULL);	335	rc = vfs_lookup_internal(path, lflag, &lr, NULL);
337	if (rc) {	336	if (rc) {
338	if (lflag & L_CREATE)	337	if (lflag & L_CREATE)
339	rwlock_write_unlock(&namespace_rwlock);	338	rwlock_write_unlock(&namespace_rwlock);
340	else	339	else
341	rwlock_read_unlock(&namespace_rwlock);	340	rwlock_read_unlock(&namespace_rwlock);
342	ipc_answer_0(rid, rc);	341	ipc_answer_0(rid, rc);
343	free(path);	342	free(path);
344	return;	343	return;
345	}	344	}
346		345
347	/* Path is no longer needed. */	346	/* Path is no longer needed. */
348	free(path);	347	free(path);
349		348
350	vfs_node_t *node = vfs_node_get(&lr);	349	vfs_node_t *node = vfs_node_get(&lr);
351	if (lflag & L_CREATE)	350	if (lflag & L_CREATE)
352	rwlock_write_unlock(&namespace_rwlock);	351	rwlock_write_unlock(&namespace_rwlock);
353	else	352	else
354	rwlock_read_unlock(&namespace_rwlock);	353	rwlock_read_unlock(&namespace_rwlock);
355		354
356	/* Truncate the file if requested and if necessary. */	355	/* Truncate the file if requested and if necessary. */
357	if (oflag & O_TRUNC) {	356	if (oflag & O_TRUNC) {
358	rwlock_write_lock(&node->contents_rwlock);	357	rwlock_write_lock(&node->contents_rwlock);
359	if (node->size) {	358	if (node->size) {
360	rc = vfs_truncate_internal(node->fs_handle,	359	rc = vfs_truncate_internal(node->fs_handle,
361	node->dev_handle, node->index, 0);	360	node->dev_handle, node->index, 0);
362	if (rc) {	361	if (rc) {
363	rwlock_write_unlock(&node->contents_rwlock);	362	rwlock_write_unlock(&node->contents_rwlock);
364	vfs_node_put(node);	363	vfs_node_put(node);
365	ipc_answer_0(rid, rc);	364	ipc_answer_0(rid, rc);
366	return;	365	return;
367	}	366	}
368	node->size = 0;	367	node->size = 0;
369	}	368	}
370	rwlock_write_unlock(&node->contents_rwlock);	369	rwlock_write_unlock(&node->contents_rwlock);
371	}	370	}
372		371
373	/*	372	/*
374	* Get ourselves a file descriptor and the corresponding vfs_file_t	373	* Get ourselves a file descriptor and the corresponding vfs_file_t
375	* structure.	374	* structure.
376	*/	375	*/
377	int fd = vfs_fd_alloc();	376	int fd = vfs_fd_alloc();
378	if (fd < 0) {	377	if (fd < 0) {
379	vfs_node_put(node);	378	vfs_node_put(node);
380	ipc_answer_0(rid, fd);	379	ipc_answer_0(rid, fd);
381	return;	380	return;
382	}	381	}
383	vfs_file_t *file = vfs_file_get(fd);	382	vfs_file_t *file = vfs_file_get(fd);
384	file->node = node;	383	file->node = node;
385	if (oflag & O_APPEND)	384	if (oflag & O_APPEND)
386	file->append = true;	385	file->append = true;
387		386
388	/*	387	/*
389	* The following increase in reference count is for the fact that the	388	* The following increase in reference count is for the fact that the
390	* file is being opened and that a file structure is pointing to it.	389	* file is being opened and that a file structure is pointing to it.
391	* It is necessary so that the file will not disappear when	390	* It is necessary so that the file will not disappear when
392	* vfs_node_put() is called. The reference will be dropped by the	391	* vfs_node_put() is called. The reference will be dropped by the
393	* respective VFS_CLOSE.	392	* respective VFS_CLOSE.
394	*/	393	*/
395	vfs_node_addref(node);	394	vfs_node_addref(node);
396	vfs_node_put(node);	395	vfs_node_put(node);
397		396
398	/* Success! Return the new file descriptor to the client. */	397	/* Success! Return the new file descriptor to the client. */
399	ipc_answer_1(rid, EOK, fd);	398	ipc_answer_1(rid, EOK, fd);
400	}	399	}
401		400
402	void vfs_close(ipc_callid_t rid, ipc_call_t *request)	401	void vfs_close(ipc_callid_t rid, ipc_call_t *request)
403	{	402	{
404	int fd = IPC_GET_ARG1(*request);	403	int fd = IPC_GET_ARG1(*request);
405	if (fd >= MAX_OPEN_FILES) {	404	if (fd >= MAX_OPEN_FILES) {
406	ipc_answer_0(rid, EBADF);	405	ipc_answer_0(rid, EBADF);
407	return;	406	return;
408	}	407	}
409	vfs_fd_free(fd);	408	vfs_fd_free(fd);
410	ipc_answer_0(rid, EOK);	409	ipc_answer_0(rid, EOK);
411	}	410	}
412		411
413	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)	412	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
414	{	413	{
415		414
416	/*	415	/*
417	* The following code strongly depends on the fact that the files data	416	* The following code strongly depends on the fact that the files data
418	* structure can be only accessed by a single fibril and all file	417	* structure can be only accessed by a single fibril and all file
419	* operations are serialized (i.e. the reads and writes cannot	418	* operations are serialized (i.e. the reads and writes cannot
420	* interleave and a file cannot be closed while it is being read).	419	* interleave and a file cannot be closed while it is being read).
421	*	420	*
422	* Additional synchronization needs to be added once the table of	421	* Additional synchronization needs to be added once the table of
423	* open files supports parallel access!	422	* open files supports parallel access!
424	*/	423	*/
425		424
426	int fd = IPC_GET_ARG1(*request);	425	int fd = IPC_GET_ARG1(*request);
427		426
428	/* Lookup the file structure corresponding to the file descriptor. */	427	/* Lookup the file structure corresponding to the file descriptor. */
429	vfs_file_t *file = vfs_file_get(fd);	428	vfs_file_t *file = vfs_file_get(fd);
430	if (!file) {	429	if (!file) {
431	ipc_answer_0(rid, ENOENT);	430	ipc_answer_0(rid, ENOENT);
432	return;	431	return;
433	}	432	}
434		433
435	/*	434	/*
436	* Now we need to receive a call with client's	435	* Now we need to receive a call with client's
437	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.	436	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
438	*/	437	*/
439	ipc_callid_t callid;	438	ipc_callid_t callid;
440	int res;	439	int res;
441	if (read)	440	if (read)
442	res = ipc_data_read_receive(&callid, NULL);	441	res = ipc_data_read_receive(&callid, NULL);
443	else	442	else
444	res = ipc_data_write_receive(&callid, NULL);	443	res = ipc_data_write_receive(&callid, NULL);
445	if (!res) {	444	if (!res) {
446	ipc_answer_0(callid, EINVAL);	445	ipc_answer_0(callid, EINVAL);
447	ipc_answer_0(rid, EINVAL);	446	ipc_answer_0(rid, EINVAL);
448	return;	447	return;
449	}	448	}
450		449
451	/*	450	/*
452	* Lock the open file structure so that no other thread can manipulate	451	* Lock the open file structure so that no other thread can manipulate
453	* the same open file at a time.	452	* the same open file at a time.
454	*/	453	*/
455	futex_down(&file->lock);	454	futex_down(&file->lock);
456		455
457	/*	456	/*
458	* Lock the file's node so that no other client can read/write to it at	457	* Lock the file's node so that no other client can read/write to it at
459	* the same time.	458	* the same time.
460	*/	459	*/
461	if (read)	460	if (read)
462	rwlock_read_lock(&file->node->contents_rwlock);	461	rwlock_read_lock(&file->node->contents_rwlock);
463	else	462	else
464	rwlock_write_lock(&file->node->contents_rwlock);	463	rwlock_write_lock(&file->node->contents_rwlock);
465		464
466	int fs_phone = vfs_grab_phone(file->node->fs_handle);	465	int fs_phone = vfs_grab_phone(file->node->fs_handle);
467		466
468	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */	467	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
469	aid_t msg;	468	aid_t msg;
470	ipc_call_t answer;	469	ipc_call_t answer;
471	if (!read && file->append)	470	if (!read && file->append)
472	file->pos = file->node->size;	471	file->pos = file->node->size;
473	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),	472	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
474	file->node->dev_handle, file->node->index, file->pos, &answer);	473	file->node->dev_handle, file->node->index, file->pos, &answer);
475		474
476	/*	475	/*
477	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the	476	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
478	* destination FS server. The call will be routed as if sent by	477	* destination FS server. The call will be routed as if sent by
479	* ourselves. Note that call arguments are immutable in this case so we	478	* ourselves. Note that call arguments are immutable in this case so we
480	* don't have to bother.	479	* don't have to bother.
481	*/	480	*/
482	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);	481	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
483		482
484	vfs_release_phone(fs_phone);	483	vfs_release_phone(fs_phone);
485		484
486	/* Wait for reply from the FS server. */	485	/* Wait for reply from the FS server. */
487	ipcarg_t rc;	486	ipcarg_t rc;
488	async_wait_for(msg, &rc);	487	async_wait_for(msg, &rc);
489	size_t bytes = IPC_GET_ARG1(answer);	488	size_t bytes = IPC_GET_ARG1(answer);
490		489
491	/* Unlock the VFS node. */	490	/* Unlock the VFS node. */
492	if (read)	491	if (read)
493	rwlock_read_unlock(&file->node->contents_rwlock);	492	rwlock_read_unlock(&file->node->contents_rwlock);
494	else {	493	else {
495	/* Update the cached version of node's size. */	494	/* Update the cached version of node's size. */
496	if (rc == EOK)	495	if (rc == EOK)
497	file->node->size = IPC_GET_ARG2(answer);	496	file->node->size = IPC_GET_ARG2(answer);
498	rwlock_write_unlock(&file->node->contents_rwlock);	497	rwlock_write_unlock(&file->node->contents_rwlock);
499	}	498	}
500		499

Subversion Repositories HelenOS

(root)/trunk/uspace/srv/vfs/vfs_ops.c – Rev 2763 → 2766