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 <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 <atomic.h>
#include "vfs.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);
 
atomic_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("/", strlen("/"), 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);
    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);
 
    /*
     * 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, size, 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;
    }
 
    /*
     * Now we are on the verge of accepting the path.
     *
     * There is one optimization we could do in the future: copy the path
     * directly into the PLB using some kind of a callback.
     */
    char *path = malloc(len);
   
    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;
    }
   
    /*
     * 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, len, 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;
    }
 
    /*
     * Now we are on the verge of accepting the path.
     *
     * There is one optimization we could do in the future: copy the path
     * directly into the PLB using some kind of a callback.
     */
    char *path = malloc(len);
   
    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;
    }
   
    rwlock_write_lock(&namespace_rwlock);
    int lflag = L_DIRECTORY | L_CREATE | L_EXCLUSIVE;
    rc = vfs_lookup_internal(path, len, 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;
    }
 
    /*
     * Now we are on the verge of accepting the path.
     *
     * There is one optimization we could do in the future: copy the path
     * directly into the PLB using some kind of a callback.
     */
    char *path = malloc(len);
   
    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;
    }
   
    rwlock_write_lock(&namespace_rwlock);
    lflag &= L_DIRECTORY;   /* sanitize lflag */
    vfs_lookup_res_t lr;
    rc = vfs_lookup_internal(path, len, lflag | L_DESTROY, &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);
    node->lnkcnt--;
    rwlock_write_unlock(&namespace_rwlock);
    vfs_node_put(node);
    ipc_answer_0(rid, EOK);
}
 
/**
 * @}
 */
 

Rev 2748	Rev 2749
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 <ipc/ipc.h>	38	#include <ipc/ipc.h>
39	#include <async.h>	39	#include <async.h>
40	#include <errno.h>	40	#include <errno.h>
41	#include <stdio.h>	41	#include <stdio.h>
42	#include <stdlib.h>	42	#include <stdlib.h>
43	#include <string.h>	43	#include <string.h>
44	#include <bool.h>	44	#include <bool.h>
45	#include <futex.h>	45	#include <futex.h>
46	#include <rwlock.h>	46	#include <rwlock.h>
47	#include <libadt/list.h>	47	#include <libadt/list.h>
48	#include <unistd.h>	48	#include <unistd.h>
49	#include <ctype.h>	49	#include <ctype.h>
50	#include <fcntl.h>	50	#include <fcntl.h>
51	#include <assert.h>	51	#include <assert.h>
52	#include <atomic.h>	52	#include <atomic.h>
53	#include "vfs.h"	53	#include "vfs.h"
54		54
55	/* Forward declarations of static functions. */	55	/* Forward declarations of static functions. */
56	static int vfs_truncate_internal(int, int, unsigned long, size_t);	56	static int vfs_truncate_internal(int, int, unsigned long, 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	atomic_t rootfs_futex = FUTEX_INITIALIZER;	64	atomic_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 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)
72	{	72	{
73	vfs_pair_t altroot = {	73	vfs_pair_t altroot = {
74	.fs_handle = fs_handle,	74	.fs_handle = fs_handle,
75	.dev_handle = dev_handle,	75	.dev_handle = dev_handle,
76	};	76	};
77		77
78	return vfs_lookup_internal("/", strlen("/"), L_DIRECTORY, result,	78	return vfs_lookup_internal("/", strlen("/"), L_DIRECTORY, result,
79	&altroot);	79	&altroot);
80	}	80	}
81		81
82	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)	82	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
83	{	83	{
84	int dev_handle;	84	int dev_handle;
85	vfs_node_t *mp_node = NULL;	85	vfs_node_t *mp_node = NULL;
86		86
87	/*	87	/*
88	* We expect the library to do the device-name to device-handle	88	* We expect the library to do the device-name to device-handle
89	* translation for us, thus the device handle will arrive as ARG1	89	* translation for us, thus the device handle will arrive as ARG1
90	* in the request.	90	* in the request.
91	*/	91	*/
92	dev_handle = IPC_GET_ARG1(*request);	92	dev_handle = IPC_GET_ARG1(*request);
93		93
94	/*	94	/*
95	* For now, don't make use of ARG2 and ARG3, but they can be used to	95	* For now, don't make use of ARG2 and ARG3, but they can be used to
96	* carry mount options in the future.	96	* carry mount options in the future.
97	*/	97	*/
98		98
99	ipc_callid_t callid;	99	ipc_callid_t callid;
100	size_t size;	100	size_t size;
101		101
102	/*	102	/*
103	* Now, we expect the client to send us data with the name of the file	103	* Now, we expect the client to send us data with the name of the file
104	* system.	104	* system.
105	*/	105	*/
106	if (!ipc_data_write_receive(&callid, &size)) {	106	if (!ipc_data_write_receive(&callid, &size)) {
107	ipc_answer_0(callid, EINVAL);	107	ipc_answer_0(callid, EINVAL);
108	ipc_answer_0(rid, EINVAL);	108	ipc_answer_0(rid, EINVAL);
109	return;	109	return;
110	}	110	}
111		111
112	/*	112	/*
113	* Don't receive more than is necessary for storing a full file system	113	* Don't receive more than is necessary for storing a full file system
114	* name.	114	* name.
115	*/	115	*/
116	if (size < 1 \|\| size > FS_NAME_MAXLEN) {	116	if (size < 1 \|\| size > FS_NAME_MAXLEN) {
117	ipc_answer_0(callid, EINVAL);	117	ipc_answer_0(callid, EINVAL);
118	ipc_answer_0(rid, EINVAL);	118	ipc_answer_0(rid, EINVAL);
119	return;	119	return;
120	}	120	}
121		121
122	/* Deliver the file system name. */	122	/* Deliver the file system name. */
123	char fs_name[FS_NAME_MAXLEN + 1];	123	char fs_name[FS_NAME_MAXLEN + 1];
124	(void) ipc_data_write_finalize(callid, fs_name, size);	124	(void) ipc_data_write_finalize(callid, fs_name, size);
125	fs_name[size] = '\0';	125	fs_name[size] = '\0';
126		126
127	/*	127	/*
128	* Check if we know a file system with the same name as is in fs_name.	128	* 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.	129	* This will also give us its file system handle.
130	*/	130	*/
131	int fs_handle = fs_name_to_handle(fs_name, true);	131	int fs_handle = fs_name_to_handle(fs_name, true);
132	if (!fs_handle) {	132	if (!fs_handle) {
133	ipc_answer_0(rid, ENOENT);	133	ipc_answer_0(rid, ENOENT);
134	return;	134	return;
135	}	135	}
136		136
137	/* Now, we want the client to send us the mount point. */	137	/* Now, we want the client to send us the mount point. */
138	if (!ipc_data_write_receive(&callid, &size)) {	138	if (!ipc_data_write_receive(&callid, &size)) {
139	ipc_answer_0(callid, EINVAL);	139	ipc_answer_0(callid, EINVAL);
140	ipc_answer_0(rid, EINVAL);	140	ipc_answer_0(rid, EINVAL);
141	return;	141	return;
142	}	142	}
143		143
144	/* Check whether size is reasonable wrt. the mount point. */	144	/* Check whether size is reasonable wrt. the mount point. */
145	if (size < 1 \|\| size > MAX_PATH_LEN) {	145	if (size < 1 \|\| size > MAX_PATH_LEN) {
146	ipc_answer_0(callid, EINVAL);	146	ipc_answer_0(callid, EINVAL);
147	ipc_answer_0(rid, EINVAL);	147	ipc_answer_0(rid, EINVAL);
148	return;	148	return;
149	}	149	}
150	/* Allocate buffer for the mount point data being received. */	150	/* Allocate buffer for the mount point data being received. */
151	uint8_t *buf;	151	uint8_t *buf;
152	buf = malloc(size);	152	buf = malloc(size);
153	if (!buf) {	153	if (!buf) {
154	ipc_answer_0(callid, ENOMEM);	154	ipc_answer_0(callid, ENOMEM);
155	ipc_answer_0(rid, ENOMEM);	155	ipc_answer_0(rid, ENOMEM);
156	return;	156	return;
157	}	157	}
158		158
159	/* Deliver the mount point. */	159	/* Deliver the mount point. */
160	(void) ipc_data_write_finalize(callid, buf, size);	160	(void) ipc_data_write_finalize(callid, buf, size);
161		161
162	/*	162	/*
163	* Lookup the root node of the filesystem being mounted.	163	* Lookup the root node of the filesystem being mounted.
164	* 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
165	* 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
166	* that we can track how many times it has been mounted.	166	* that we can track how many times it has been mounted.
167	*/	167	*/
168	int rc;	168	int rc;
169	vfs_lookup_res_t mr_res;	169	vfs_lookup_res_t mr_res;
170	rc = lookup_root(fs_handle, dev_handle, &mr_res);	170	rc = lookup_root(fs_handle, dev_handle, &mr_res);
171	if (rc != EOK) {	171	if (rc != EOK) {
172	free(buf);	172	free(buf);
173	ipc_answer_0(rid, rc);	173	ipc_answer_0(rid, rc);
174	return;	174	return;
175	}	175	}
176	vfs_node_t *mr_node = vfs_node_get(&mr_res);	176	vfs_node_t *mr_node = vfs_node_get(&mr_res);
177	if (!mr_node) {	177	if (!mr_node) {
178	free(buf);	178	free(buf);
179	ipc_answer_0(rid, ENOMEM);	179	ipc_answer_0(rid, ENOMEM);
180	return;	180	return;
181	}	181	}
182		182
183	/* Finally, we need to resolve the path to the mountpoint. */	183	/* Finally, we need to resolve the path to the mountpoint. */
184	vfs_lookup_res_t mp_res;	184	vfs_lookup_res_t mp_res;
185	futex_down(&rootfs_futex);	185	futex_down(&rootfs_futex);
186	if (rootfs.fs_handle) {	186	if (rootfs.fs_handle) {
187	/* We already have the root FS. */	187	/* We already have the root FS. */
188	rwlock_write_lock(&namespace_rwlock);	188	rwlock_write_lock(&namespace_rwlock);
189	rc = vfs_lookup_internal(buf, size, L_DIRECTORY, &mp_res,	189	rc = vfs_lookup_internal(buf, size, L_DIRECTORY, &mp_res,
190	NULL);	190	NULL);
191	if (rc != EOK) {	191	if (rc != EOK) {
192	/* The lookup failed for some reason. */	192	/* The lookup failed for some reason. */
193	rwlock_write_unlock(&namespace_rwlock);	193	rwlock_write_unlock(&namespace_rwlock);
194	futex_up(&rootfs_futex);	194	futex_up(&rootfs_futex);
195	vfs_node_put(mr_node); /* failed -> drop reference */	195	vfs_node_put(mr_node); /* failed -> drop reference */
196	free(buf);	196	free(buf);
197	ipc_answer_0(rid, rc);	197	ipc_answer_0(rid, rc);
198	return;	198	return;
199	}	199	}
200	mp_node = vfs_node_get(&mp_res);	200	mp_node = vfs_node_get(&mp_res);
201	if (!mp_node) {	201	if (!mp_node) {
202	rwlock_write_unlock(&namespace_rwlock);	202	rwlock_write_unlock(&namespace_rwlock);
203	futex_up(&rootfs_futex);	203	futex_up(&rootfs_futex);
204	vfs_node_put(mr_node); /* failed -> drop reference */	204	vfs_node_put(mr_node); /* failed -> drop reference */
205	free(buf);	205	free(buf);
206	ipc_answer_0(rid, ENOMEM);	206	ipc_answer_0(rid, ENOMEM);
207	return;	207	return;
208	}	208	}
209	/*	209	/*
210	* Now we hold a reference to mp_node.	210	* Now we hold a reference to mp_node.
211	* It will be dropped upon the corresponding VFS_UNMOUNT.	211	* It will be dropped upon the corresponding VFS_UNMOUNT.
212	* This prevents the mount point from being deleted.	212	* This prevents the mount point from being deleted.
213	*/	213	*/
214	rwlock_write_unlock(&namespace_rwlock);	214	rwlock_write_unlock(&namespace_rwlock);
215	} else {	215	} else {
216	/* We still don't have the root file system mounted. */	216	/* We still don't have the root file system mounted. */
217	if ((size == 1) && (buf[0] == '/')) {	217	if ((size == 1) && (buf[0] == '/')) {
218	/* For this simple, but important case, we are done. */	218	/* For this simple, but important case, we are done. */
219	rootfs = mr_res.triplet;	219	rootfs = mr_res.triplet;
220	futex_up(&rootfs_futex);	220	futex_up(&rootfs_futex);
221	free(buf);	221	free(buf);
222	ipc_answer_0(rid, EOK);	222	ipc_answer_0(rid, EOK);
223	return;	223	return;
224	} else {	224	} else {
225	/*	225	/*
226	* We can't resolve this without the root filesystem	226	* We can't resolve this without the root filesystem
227	* being mounted first.	227	* being mounted first.
228	*/	228	*/
229	futex_up(&rootfs_futex);	229	futex_up(&rootfs_futex);
230	free(buf);	230	free(buf);
231	vfs_node_put(mr_node); /* failed -> drop reference */	231	vfs_node_put(mr_node); /* failed -> drop reference */
232	ipc_answer_0(rid, ENOENT);	232	ipc_answer_0(rid, ENOENT);
233	return;	233	return;
234	}	234	}
235	}	235	}
236	futex_up(&rootfs_futex);	236	futex_up(&rootfs_futex);
237		237
238	free(buf); /* The buffer is not needed anymore. */	238	free(buf); /* The buffer is not needed anymore. */
239		239
240	/*	240	/*
241	* At this point, we have all necessary pieces: file system and device	241	* 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	242	* handles, and we know the mount point VFS node and also the root node
243	* of the file system being mounted.	243	* of the file system being mounted.
244	*/	244	*/
245		245
246	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);	246	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);
247	/* Later we can use ARG3 to pass mode/flags. */	247	/* Later we can use ARG3 to pass mode/flags. */
248	aid_t req1 = async_send_3(phone, VFS_MOUNT,	248	aid_t req1 = async_send_3(phone, VFS_MOUNT,
249	(ipcarg_t) mp_res.triplet.dev_handle,	249	(ipcarg_t) mp_res.triplet.dev_handle,
250	(ipcarg_t) mp_res.triplet.index, 0, NULL);	250	(ipcarg_t) mp_res.triplet.index, 0, NULL);
251	/* The second call uses the same method. */	251	/* The second call uses the same method. */
252	aid_t req2 = async_send_3(phone, VFS_MOUNT,	252	aid_t req2 = async_send_3(phone, VFS_MOUNT,
253	(ipcarg_t) mr_res.triplet.fs_handle,	253	(ipcarg_t) mr_res.triplet.fs_handle,
254	(ipcarg_t) mr_res.triplet.dev_handle,	254	(ipcarg_t) mr_res.triplet.dev_handle,
255	(ipcarg_t) mr_res.triplet.index, NULL);	255	(ipcarg_t) mr_res.triplet.index, NULL);
256	vfs_release_phone(phone);	256	vfs_release_phone(phone);
257		257
258	ipcarg_t rc1;	258	ipcarg_t rc1;
259	ipcarg_t rc2;	259	ipcarg_t rc2;
260	async_wait_for(req1, &rc1);	260	async_wait_for(req1, &rc1);
261	async_wait_for(req2, &rc2);	261	async_wait_for(req2, &rc2);
262		262
263	if ((rc1 != EOK) \|\| (rc2 != EOK)) {	263	if ((rc1 != EOK) \|\| (rc2 != EOK)) {
264	/* Mount failed, drop references to mr_node and mp_node. */	264	/* Mount failed, drop references to mr_node and mp_node. */
265	vfs_node_put(mr_node);	265	vfs_node_put(mr_node);
266	if (mp_node)	266	if (mp_node)
267	vfs_node_put(mp_node);	267	vfs_node_put(mp_node);
268	}	268	}
269		269
270	if (rc2 == EOK)	270	if (rc2 == EOK)
271	ipc_answer_0(rid, rc1);	271	ipc_answer_0(rid, rc1);
272	else if (rc1 == EOK)	272	else if (rc1 == EOK)
273	ipc_answer_0(rid, rc2);	273	ipc_answer_0(rid, rc2);
274	else	274	else
275	ipc_answer_0(rid, rc1);	275	ipc_answer_0(rid, rc1);
276	}	276	}
277		277
278	void vfs_open(ipc_callid_t rid, ipc_call_t *request)	278	void vfs_open(ipc_callid_t rid, ipc_call_t *request)
279	{	279	{
280	if (!vfs_files_init()) {	280	if (!vfs_files_init()) {
281	ipc_answer_0(rid, ENOMEM);	281	ipc_answer_0(rid, ENOMEM);
282	return;	282	return;
283	}	283	}
284		284
285	/*	285	/*
286	* The POSIX interface is open(path, oflag, mode).	286	* The POSIX interface is open(path, oflag, mode).
287	* We can receive oflags and mode along with the VFS_OPEN call; the path	287	* We can receive oflags and mode along with the VFS_OPEN call; the path
288	* will need to arrive in another call.	288	* will need to arrive in another call.
289	*	289	*
290	* We also receive one private, non-POSIX set of flags called lflag	290	* We also receive one private, non-POSIX set of flags called lflag
291	* used to pass information to vfs_lookup_internal().	291	* used to pass information to vfs_lookup_internal().
292	*/	292	*/
293	int lflag = IPC_GET_ARG1(*request);	293	int lflag = IPC_GET_ARG1(*request);
294	int oflag = IPC_GET_ARG2(*request);	294	int oflag = IPC_GET_ARG2(*request);
295	int mode = IPC_GET_ARG3(*request);	295	int mode = IPC_GET_ARG3(*request);
296	size_t len;	296	size_t len;
297		297
298	if (oflag & O_CREAT)	298	if (oflag & O_CREAT)
299	lflag \|= L_CREATE;	299	lflag \|= L_CREATE;
300	if (oflag & O_EXCL)	300	if (oflag & O_EXCL)
301	lflag \|= L_EXCLUSIVE;	301	lflag \|= L_EXCLUSIVE;
302		302
303	ipc_callid_t callid;	303	ipc_callid_t callid;
304		304
305	if (!ipc_data_write_receive(&callid, &len)) {	305	if (!ipc_data_write_receive(&callid, &len)) {
306	ipc_answer_0(callid, EINVAL);	306	ipc_answer_0(callid, EINVAL);
307	ipc_answer_0(rid, EINVAL);	307	ipc_answer_0(rid, EINVAL);
308	return;	308	return;
309	}	309	}
310		310
311	/*	311	/*
312	* Now we are on the verge of accepting the path.	312	* Now we are on the verge of accepting the path.
313	*	313	*
314	* There is one optimization we could do in the future: copy the path	314	* There is one optimization we could do in the future: copy the path
315	* directly into the PLB using some kind of a callback.	315	* directly into the PLB using some kind of a callback.
316	*/	316	*/
317	char *path = malloc(len);	317	char *path = malloc(len);
318		318
319	if (!path) {	319	if (!path) {
320	ipc_answer_0(callid, ENOMEM);	320	ipc_answer_0(callid, ENOMEM);
321	ipc_answer_0(rid, ENOMEM);	321	ipc_answer_0(rid, ENOMEM);
322	return;	322	return;
323	}	323	}
324		324
325	int rc;	325	int rc;
326	if ((rc = ipc_data_write_finalize(callid, path, len))) {	326	if ((rc = ipc_data_write_finalize(callid, path, len))) {
327	ipc_answer_0(rid, rc);	327	ipc_answer_0(rid, rc);
328	free(path);	328	free(path);
329	return;	329	return;
330	}	330	}
331		331
332	/*	332	/*
333	* Avoid the race condition in which the file can be deleted before we	333	* Avoid the race condition in which the file can be deleted before we
334	* find/create-and-lock the VFS node corresponding to the looked-up	334	* find/create-and-lock the VFS node corresponding to the looked-up
335	* triplet.	335	* triplet.
336	*/	336	*/
337	if (lflag & L_CREATE)	337	if (lflag & L_CREATE)
338	rwlock_write_lock(&namespace_rwlock);	338	rwlock_write_lock(&namespace_rwlock);
339	else	339	else
340	rwlock_read_lock(&namespace_rwlock);	340	rwlock_read_lock(&namespace_rwlock);
341		341
342	/* The path is now populated and we can call vfs_lookup_internal(). */	342	/* The path is now populated and we can call vfs_lookup_internal(). */
343	vfs_lookup_res_t lr;	343	vfs_lookup_res_t lr;
344	rc = vfs_lookup_internal(path, len, lflag, &lr, NULL);	344	rc = vfs_lookup_internal(path, len, lflag, &lr, NULL);
345	if (rc) {	345	if (rc) {
346	if (lflag & L_CREATE)	346	if (lflag & L_CREATE)
347	rwlock_write_unlock(&namespace_rwlock);	347	rwlock_write_unlock(&namespace_rwlock);
348	else	348	else
349	rwlock_read_unlock(&namespace_rwlock);	349	rwlock_read_unlock(&namespace_rwlock);
350	ipc_answer_0(rid, rc);	350	ipc_answer_0(rid, rc);
351	free(path);	351	free(path);
352	return;	352	return;
353	}	353	}
354		354
355	/* Path is no longer needed. */	355	/* Path is no longer needed. */
356	free(path);	356	free(path);
357		357
358	vfs_node_t *node = vfs_node_get(&lr);	358	vfs_node_t *node = vfs_node_get(&lr);
359	if (lflag & L_CREATE)	359	if (lflag & L_CREATE)
360	rwlock_write_unlock(&namespace_rwlock);	360	rwlock_write_unlock(&namespace_rwlock);
361	else	361	else
362	rwlock_read_unlock(&namespace_rwlock);	362	rwlock_read_unlock(&namespace_rwlock);
363		363
364	/* Truncate the file if requested and if necessary. */	364	/* Truncate the file if requested and if necessary. */
365	if (oflag & O_TRUNC) {	365	if (oflag & O_TRUNC) {
366	futex_down(&node->contents_rwlock);	366	rwlock_write_lock(&node->contents_rwlock);
367	if (node->size) {	367	if (node->size) {
368	rc = vfs_truncate_internal(node->fs_handle,	368	rc = vfs_truncate_internal(node->fs_handle,
369	node->dev_handle, node->index, 0);	369	node->dev_handle, node->index, 0);
370	if (rc) {	370	if (rc) {
371	futex_up(&node->contents_rwlock);	371	rwlock_write_unlock(&node->contents_rwlock);
372	vfs_node_put(node);	372	vfs_node_put(node);
373	ipc_answer_0(rid, rc);	373	ipc_answer_0(rid, rc);
374	return;	374	return;
375	}	375	}
376	node->size = 0;	376	node->size = 0;
377	}	377	}
378	futex_up(&node->contents_rwlock);	378	rwlock_write_unlock(&node->contents_rwlock);
379	}	379	}
380		380
381	/*	381	/*
382	* Get ourselves a file descriptor and the corresponding vfs_file_t	382	* Get ourselves a file descriptor and the corresponding vfs_file_t
383	* structure.	383	* structure.
384	*/	384	*/
385	int fd = vfs_fd_alloc();	385	int fd = vfs_fd_alloc();
386	if (fd < 0) {	386	if (fd < 0) {
387	vfs_node_put(node);	387	vfs_node_put(node);
388	ipc_answer_0(rid, fd);	388	ipc_answer_0(rid, fd);
389	return;	389	return;
390	}	390	}
391	vfs_file_t *file = vfs_file_get(fd);	391	vfs_file_t *file = vfs_file_get(fd);
392	file->node = node;	392	file->node = node;
393	if (oflag & O_APPEND)	393	if (oflag & O_APPEND)
394	file->append = true;	394	file->append = true;
395		395
396	/*	396	/*
397	* The following increase in reference count is for the fact that the	397	* The following increase in reference count is for the fact that the
398	* file is being opened and that a file structure is pointing to it.	398	* file is being opened and that a file structure is pointing to it.
399	* It is necessary so that the file will not disappear when	399	* It is necessary so that the file will not disappear when
400	* vfs_node_put() is called. The reference will be dropped by the	400	* vfs_node_put() is called. The reference will be dropped by the
401	* respective VFS_CLOSE.	401	* respective VFS_CLOSE.
402	*/	402	*/
403	vfs_node_addref(node);	403	vfs_node_addref(node);
404	vfs_node_put(node);	404	vfs_node_put(node);
405		405
406	/* Success! Return the new file descriptor to the client. */	406	/* Success! Return the new file descriptor to the client. */
407	ipc_answer_1(rid, EOK, fd);	407	ipc_answer_1(rid, EOK, fd);
408	}	408	}
409		409
410	void vfs_close(ipc_callid_t rid, ipc_call_t *request)	410	void vfs_close(ipc_callid_t rid, ipc_call_t *request)
411	{	411	{
412	int fd = IPC_GET_ARG1(*request);	412	int fd = IPC_GET_ARG1(*request);
413	if (fd >= MAX_OPEN_FILES) {	413	if (fd >= MAX_OPEN_FILES) {
414	ipc_answer_0(rid, EBADF);	414	ipc_answer_0(rid, EBADF);
415	return;	415	return;
416	}	416	}
417	vfs_fd_free(fd);	417	vfs_fd_free(fd);
418	ipc_answer_0(rid, EOK);	418	ipc_answer_0(rid, EOK);
419	}	419	}
420		420
421	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)	421	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
422	{	422	{
423		423
424	/*	424	/*
425	* The following code strongly depends on the fact that the files data	425	* The following code strongly depends on the fact that the files data
426	* structure can be only accessed by a single fibril and all file	426	* structure can be only accessed by a single fibril and all file
427	* operations are serialized (i.e. the reads and writes cannot	427	* operations are serialized (i.e. the reads and writes cannot
428	* interleave and a file cannot be closed while it is being read).	428	* interleave and a file cannot be closed while it is being read).
429	*	429	*
430	* Additional synchronization needs to be added once the table of	430	* Additional synchronization needs to be added once the table of
431	* open files supports parallel access!	431	* open files supports parallel access!
432	*/	432	*/
433		433
434	int fd = IPC_GET_ARG1(*request);	434	int fd = IPC_GET_ARG1(*request);
435		435
436	/* Lookup the file structure corresponding to the file descriptor. */	436	/* Lookup the file structure corresponding to the file descriptor. */
437	vfs_file_t *file = vfs_file_get(fd);	437	vfs_file_t *file = vfs_file_get(fd);
438	if (!file) {	438	if (!file) {
439	ipc_answer_0(rid, ENOENT);	439	ipc_answer_0(rid, ENOENT);
440	return;	440	return;
441	}	441	}
442		442
443	/*	443	/*
444	* Now we need to receive a call with client's	444	* Now we need to receive a call with client's
445	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.	445	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
446	*/	446	*/
447	ipc_callid_t callid;	447	ipc_callid_t callid;
448	int res;	448	int res;
449	if (read)	449	if (read)
450	res = ipc_data_read_receive(&callid, NULL);	450	res = ipc_data_read_receive(&callid, NULL);
451	else	451	else
452	res = ipc_data_write_receive(&callid, NULL);	452	res = ipc_data_write_receive(&callid, NULL);
453	if (!res) {	453	if (!res) {
454	ipc_answer_0(callid, EINVAL);	454	ipc_answer_0(callid, EINVAL);
455	ipc_answer_0(rid, EINVAL);	455	ipc_answer_0(rid, EINVAL);
456	return;	456	return;
457	}	457	}
458		458
459	/*	459	/*
460	* Lock the open file structure so that no other thread can manipulate	460	* Lock the open file structure so that no other thread can manipulate
461	* the same open file at a time.	461	* the same open file at a time.
462	*/	462	*/
463	futex_down(&file->lock);	463	futex_down(&file->lock);
464		464
465	/*	465	/*
466	* Lock the file's node so that no other client can read/write to it at	466	* Lock the file's node so that no other client can read/write to it at
467	* the same time.	467	* the same time.
468	*/	468	*/
469	if (read)	469	if (read)
470	rwlock_read_lock(&file->node->contents_rwlock);	470	rwlock_read_lock(&file->node->contents_rwlock);
471	else	471	else
472	rwlock_write_lock(&file->node->contents_rwlock);	472	rwlock_write_lock(&file->node->contents_rwlock);
473		473
474	int fs_phone = vfs_grab_phone(file->node->fs_handle);	474	int fs_phone = vfs_grab_phone(file->node->fs_handle);
475		475
476	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */	476	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
477	aid_t msg;	477	aid_t msg;
478	ipc_call_t answer;	478	ipc_call_t answer;
479	if (!read && file->append)	479	if (!read && file->append)
480	file->pos = file->node->size;	480	file->pos = file->node->size;
481	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),	481	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
482	file->node->dev_handle, file->node->index, file->pos, &answer);	482	file->node->dev_handle, file->node->index, file->pos, &answer);
483		483
484	/*	484	/*
485	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the	485	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
486	* destination FS server. The call will be routed as if sent by	486	* destination FS server. The call will be routed as if sent by
487	* ourselves. Note that call arguments are immutable in this case so we	487	* ourselves. Note that call arguments are immutable in this case so we
488	* don't have to bother.	488	* don't have to bother.
489	*/	489	*/
490	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);	490	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
491		491
492	vfs_release_phone(fs_phone);	492	vfs_release_phone(fs_phone);
493		493
494	/* Wait for reply from the FS server. */	494	/* Wait for reply from the FS server. */
495	ipcarg_t rc;	495	ipcarg_t rc;
496	async_wait_for(msg, &rc);	496	async_wait_for(msg, &rc);
497	size_t bytes = IPC_GET_ARG1(answer);	497	size_t bytes = IPC_GET_ARG1(answer);
498		498
499	/* Unlock the VFS node. */	499	/* Unlock the VFS node. */
500	if (read)	500	if (read)

Subversion Repositories HelenOS

(root)/trunk/uspace/srv/vfs/vfs_ops.c – Rev 2748 → 2749