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"
 
/**
 * 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);
 
    /*
     * 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);
}
 
void vfs_truncate(ipc_callid_t rid, ipc_call_t *request)
{
    int fd = IPC_GET_ARG1(*request);
    size_t size = IPC_GET_ARG2(*request);
    ipcarg_t 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);
    int fs_phone = vfs_grab_phone(file->node->fs_handle);
    rc = async_req_3_0(fs_phone, VFS_TRUNCATE,
        (ipcarg_t)file->node->dev_handle, (ipcarg_t)file->node->index,
        (ipcarg_t)size);
    vfs_release_phone(fs_phone);
    if (rc == EOK)
        file->node->size = size;
    rwlock_write_unlock(&file->node->contents_rwlock);
 
    futex_up(&file->lock);
    ipc_answer_0(rid, 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 2735	Rev 2742
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	/**	55	/**
56	* This rwlock prevents the race between a triplet-to-VFS-node resolution and a	56	* This rwlock prevents the race between a triplet-to-VFS-node resolution and a
57	* concurrent VFS operation which modifies the file system namespace.	57	* concurrent VFS operation which modifies the file system namespace.
58	*/	58	*/
59	RWLOCK_INITIALIZE(namespace_rwlock);	59	RWLOCK_INITIALIZE(namespace_rwlock);
60		60
61	atomic_t rootfs_futex = FUTEX_INITIALIZER;	61	atomic_t rootfs_futex = FUTEX_INITIALIZER;
62	vfs_triplet_t rootfs = {	62	vfs_triplet_t rootfs = {
63	.fs_handle = 0,	63	.fs_handle = 0,
64	.dev_handle = 0,	64	.dev_handle = 0,
65	.index = 0,	65	.index = 0,
66	};	66	};
67		67
68	static int lookup_root(int fs_handle, int dev_handle, vfs_lookup_res_t *result)	68	static int lookup_root(int fs_handle, int dev_handle, vfs_lookup_res_t *result)
69	{	69	{
70	vfs_pair_t altroot = {	70	vfs_pair_t altroot = {
71	.fs_handle = fs_handle,	71	.fs_handle = fs_handle,
72	.dev_handle = dev_handle,	72	.dev_handle = dev_handle,
73	};	73	};
74		74
75	return vfs_lookup_internal("/", strlen("/"), L_DIRECTORY, result,	75	return vfs_lookup_internal("/", strlen("/"), L_DIRECTORY, result,
76	&altroot);	76	&altroot);
77	}	77	}
78		78
79	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)	79	void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
80	{	80	{
81	int dev_handle;	81	int dev_handle;
82	vfs_node_t *mp_node = NULL;	82	vfs_node_t *mp_node = NULL;
83		83
84	/*	84	/*
85	* We expect the library to do the device-name to device-handle	85	* We expect the library to do the device-name to device-handle
86	* translation for us, thus the device handle will arrive as ARG1	86	* translation for us, thus the device handle will arrive as ARG1
87	* in the request.	87	* in the request.
88	*/	88	*/
89	dev_handle = IPC_GET_ARG1(*request);	89	dev_handle = IPC_GET_ARG1(*request);
90		90
91	/*	91	/*
92	* For now, don't make use of ARG2 and ARG3, but they can be used to	92	* For now, don't make use of ARG2 and ARG3, but they can be used to
93	* carry mount options in the future.	93	* carry mount options in the future.
94	*/	94	*/
95		95
96	ipc_callid_t callid;	96	ipc_callid_t callid;
97	size_t size;	97	size_t size;
98		98
99	/*	99	/*
100	* Now, we expect the client to send us data with the name of the file	100	* Now, we expect the client to send us data with the name of the file
101	* system.	101	* system.
102	*/	102	*/
103	if (!ipc_data_write_receive(&callid, &size)) {	103	if (!ipc_data_write_receive(&callid, &size)) {
104	ipc_answer_0(callid, EINVAL);	104	ipc_answer_0(callid, EINVAL);
105	ipc_answer_0(rid, EINVAL);	105	ipc_answer_0(rid, EINVAL);
106	return;	106	return;
107	}	107	}
108		108
109	/*	109	/*
110	* Don't receive more than is necessary for storing a full file system	110	* Don't receive more than is necessary for storing a full file system
111	* name.	111	* name.
112	*/	112	*/
113	if (size < 1 \|\| size > FS_NAME_MAXLEN) {	113	if (size < 1 \|\| size > FS_NAME_MAXLEN) {
114	ipc_answer_0(callid, EINVAL);	114	ipc_answer_0(callid, EINVAL);
115	ipc_answer_0(rid, EINVAL);	115	ipc_answer_0(rid, EINVAL);
116	return;	116	return;
117	}	117	}
118		118
119	/* Deliver the file system name. */	119	/* Deliver the file system name. */
120	char fs_name[FS_NAME_MAXLEN + 1];	120	char fs_name[FS_NAME_MAXLEN + 1];
121	(void) ipc_data_write_finalize(callid, fs_name, size);	121	(void) ipc_data_write_finalize(callid, fs_name, size);
122	fs_name[size] = '\0';	122	fs_name[size] = '\0';
123		123
124	/*	124	/*
125	* Check if we know a file system with the same name as is in fs_name.	125	* Check if we know a file system with the same name as is in fs_name.
126	* This will also give us its file system handle.	126	* This will also give us its file system handle.
127	*/	127	*/
128	int fs_handle = fs_name_to_handle(fs_name, true);	128	int fs_handle = fs_name_to_handle(fs_name, true);
129	if (!fs_handle) {	129	if (!fs_handle) {
130	ipc_answer_0(rid, ENOENT);	130	ipc_answer_0(rid, ENOENT);
131	return;	131	return;
132	}	132	}
133		133
134	/* Now, we want the client to send us the mount point. */	134	/* Now, we want the client to send us the mount point. */
135	if (!ipc_data_write_receive(&callid, &size)) {	135	if (!ipc_data_write_receive(&callid, &size)) {
136	ipc_answer_0(callid, EINVAL);	136	ipc_answer_0(callid, EINVAL);
137	ipc_answer_0(rid, EINVAL);	137	ipc_answer_0(rid, EINVAL);
138	return;	138	return;
139	}	139	}
140		140
141	/* Check whether size is reasonable wrt. the mount point. */	141	/* Check whether size is reasonable wrt. the mount point. */
142	if (size < 1 \|\| size > MAX_PATH_LEN) {	142	if (size < 1 \|\| size > MAX_PATH_LEN) {
143	ipc_answer_0(callid, EINVAL);	143	ipc_answer_0(callid, EINVAL);
144	ipc_answer_0(rid, EINVAL);	144	ipc_answer_0(rid, EINVAL);
145	return;	145	return;
146	}	146	}
147	/* Allocate buffer for the mount point data being received. */	147	/* Allocate buffer for the mount point data being received. */
148	uint8_t *buf;	148	uint8_t *buf;
149	buf = malloc(size);	149	buf = malloc(size);
150	if (!buf) {	150	if (!buf) {
151	ipc_answer_0(callid, ENOMEM);	151	ipc_answer_0(callid, ENOMEM);
152	ipc_answer_0(rid, ENOMEM);	152	ipc_answer_0(rid, ENOMEM);
153	return;	153	return;
154	}	154	}
155		155
156	/* Deliver the mount point. */	156	/* Deliver the mount point. */
157	(void) ipc_data_write_finalize(callid, buf, size);	157	(void) ipc_data_write_finalize(callid, buf, size);
158		158
159	/*	159	/*
160	* Lookup the root node of the filesystem being mounted.	160	* Lookup the root node of the filesystem being mounted.
161	* In this case, we don't need to take the namespace_futex as the root	161	* In this case, we don't need to take the namespace_futex as the root
162	* node cannot be removed. However, we do take a reference to it so	162	* node cannot be removed. However, we do take a reference to it so
163	* that we can track how many times it has been mounted.	163	* that we can track how many times it has been mounted.
164	*/	164	*/
165	int rc;	165	int rc;
166	vfs_lookup_res_t mr_res;	166	vfs_lookup_res_t mr_res;
167	rc = lookup_root(fs_handle, dev_handle, &mr_res);	167	rc = lookup_root(fs_handle, dev_handle, &mr_res);
168	if (rc != EOK) {	168	if (rc != EOK) {
169	free(buf);	169	free(buf);
170	ipc_answer_0(rid, rc);	170	ipc_answer_0(rid, rc);
171	return;	171	return;
172	}	172	}
173	vfs_node_t *mr_node = vfs_node_get(&mr_res);	173	vfs_node_t *mr_node = vfs_node_get(&mr_res);
174	if (!mr_node) {	174	if (!mr_node) {
175	free(buf);	175	free(buf);
176	ipc_answer_0(rid, ENOMEM);	176	ipc_answer_0(rid, ENOMEM);
177	return;	177	return;
178	}	178	}
179		179
180	/* Finally, we need to resolve the path to the mountpoint. */	180	/* Finally, we need to resolve the path to the mountpoint. */
181	vfs_lookup_res_t mp_res;	181	vfs_lookup_res_t mp_res;
182	futex_down(&rootfs_futex);	182	futex_down(&rootfs_futex);
183	if (rootfs.fs_handle) {	183	if (rootfs.fs_handle) {
184	/* We already have the root FS. */	184	/* We already have the root FS. */
185	rwlock_write_lock(&namespace_rwlock);	185	rwlock_write_lock(&namespace_rwlock);
186	rc = vfs_lookup_internal(buf, size, L_DIRECTORY, &mp_res,	186	rc = vfs_lookup_internal(buf, size, L_DIRECTORY, &mp_res,
187	NULL);	187	NULL);
188	if (rc != EOK) {	188	if (rc != EOK) {
189	/* The lookup failed for some reason. */	189	/* The lookup failed for some reason. */
190	rwlock_write_unlock(&namespace_rwlock);	190	rwlock_write_unlock(&namespace_rwlock);
191	futex_up(&rootfs_futex);	191	futex_up(&rootfs_futex);
192	vfs_node_put(mr_node); /* failed -> drop reference */	192	vfs_node_put(mr_node); /* failed -> drop reference */
193	free(buf);	193	free(buf);
194	ipc_answer_0(rid, rc);	194	ipc_answer_0(rid, rc);
195	return;	195	return;
196	}	196	}
197	mp_node = vfs_node_get(&mp_res);	197	mp_node = vfs_node_get(&mp_res);
198	if (!mp_node) {	198	if (!mp_node) {
199	rwlock_write_unlock(&namespace_rwlock);	199	rwlock_write_unlock(&namespace_rwlock);
200	futex_up(&rootfs_futex);	200	futex_up(&rootfs_futex);
201	vfs_node_put(mr_node); /* failed -> drop reference */	201	vfs_node_put(mr_node); /* failed -> drop reference */
202	free(buf);	202	free(buf);
203	ipc_answer_0(rid, ENOMEM);	203	ipc_answer_0(rid, ENOMEM);
204	return;	204	return;
205	}	205	}
206	/*	206	/*
207	* Now we hold a reference to mp_node.	207	* Now we hold a reference to mp_node.
208	* It will be dropped upon the corresponding VFS_UNMOUNT.	208	* It will be dropped upon the corresponding VFS_UNMOUNT.
209	* This prevents the mount point from being deleted.	209	* This prevents the mount point from being deleted.
210	*/	210	*/
211	rwlock_write_unlock(&namespace_rwlock);	211	rwlock_write_unlock(&namespace_rwlock);
212	} else {	212	} else {
213	/* We still don't have the root file system mounted. */	213	/* We still don't have the root file system mounted. */
214	if ((size == 1) && (buf[0] == '/')) {	214	if ((size == 1) && (buf[0] == '/')) {
215	/* For this simple, but important case, we are done. */	215	/* For this simple, but important case, we are done. */
216	rootfs = mr_res.triplet;	216	rootfs = mr_res.triplet;
217	futex_up(&rootfs_futex);	217	futex_up(&rootfs_futex);
218	free(buf);	218	free(buf);
219	ipc_answer_0(rid, EOK);	219	ipc_answer_0(rid, EOK);
220	return;	220	return;
221	} else {	221	} else {
222	/*	222	/*
223	* We can't resolve this without the root filesystem	223	* We can't resolve this without the root filesystem
224	* being mounted first.	224	* being mounted first.
225	*/	225	*/
226	futex_up(&rootfs_futex);	226	futex_up(&rootfs_futex);
227	free(buf);	227	free(buf);
228	vfs_node_put(mr_node); /* failed -> drop reference */	228	vfs_node_put(mr_node); /* failed -> drop reference */
229	ipc_answer_0(rid, ENOENT);	229	ipc_answer_0(rid, ENOENT);
230	return;	230	return;
231	}	231	}
232	}	232	}
233	futex_up(&rootfs_futex);	233	futex_up(&rootfs_futex);
234		234
235	free(buf); /* The buffer is not needed anymore. */	235	free(buf); /* The buffer is not needed anymore. */
236		236
237	/*	237	/*
238	* At this point, we have all necessary pieces: file system and device	238	* At this point, we have all necessary pieces: file system and device
239	* handles, and we know the mount point VFS node and also the root node	239	* handles, and we know the mount point VFS node and also the root node
240	* of the file system being mounted.	240	* of the file system being mounted.
241	*/	241	*/
242		242
243	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);	243	int phone = vfs_grab_phone(mp_res.triplet.fs_handle);
244	/* Later we can use ARG3 to pass mode/flags. */	244	/* Later we can use ARG3 to pass mode/flags. */
245	aid_t req1 = async_send_3(phone, VFS_MOUNT,	245	aid_t req1 = async_send_3(phone, VFS_MOUNT,
246	(ipcarg_t) mp_res.triplet.dev_handle,	246	(ipcarg_t) mp_res.triplet.dev_handle,
247	(ipcarg_t) mp_res.triplet.index, 0, NULL);	247	(ipcarg_t) mp_res.triplet.index, 0, NULL);
248	/* The second call uses the same method. */	248	/* The second call uses the same method. */
249	aid_t req2 = async_send_3(phone, VFS_MOUNT,	249	aid_t req2 = async_send_3(phone, VFS_MOUNT,
250	(ipcarg_t) mr_res.triplet.fs_handle,	250	(ipcarg_t) mr_res.triplet.fs_handle,
251	(ipcarg_t) mr_res.triplet.dev_handle,	251	(ipcarg_t) mr_res.triplet.dev_handle,
252	(ipcarg_t) mr_res.triplet.index, NULL);	252	(ipcarg_t) mr_res.triplet.index, NULL);
253	vfs_release_phone(phone);	253	vfs_release_phone(phone);
254		254
255	ipcarg_t rc1;	255	ipcarg_t rc1;
256	ipcarg_t rc2;	256	ipcarg_t rc2;
257	async_wait_for(req1, &rc1);	257	async_wait_for(req1, &rc1);
258	async_wait_for(req2, &rc2);	258	async_wait_for(req2, &rc2);
259		259
260	if ((rc1 != EOK) \|\| (rc2 != EOK)) {	260	if ((rc1 != EOK) \|\| (rc2 != EOK)) {
261	/* Mount failed, drop references to mr_node and mp_node. */	261	/* Mount failed, drop references to mr_node and mp_node. */
262	vfs_node_put(mr_node);	262	vfs_node_put(mr_node);
263	if (mp_node)	263	if (mp_node)
264	vfs_node_put(mp_node);	264	vfs_node_put(mp_node);
265	}	265	}
266		266
267	if (rc2 == EOK)	267	if (rc2 == EOK)
268	ipc_answer_0(rid, rc1);	268	ipc_answer_0(rid, rc1);
269	else if (rc1 == EOK)	269	else if (rc1 == EOK)
270	ipc_answer_0(rid, rc2);	270	ipc_answer_0(rid, rc2);
271	else	271	else
272	ipc_answer_0(rid, rc1);	272	ipc_answer_0(rid, rc1);
273	}	273	}
274		274
275	void vfs_open(ipc_callid_t rid, ipc_call_t *request)	275	void vfs_open(ipc_callid_t rid, ipc_call_t *request)
276	{	276	{
277	if (!vfs_files_init()) {	277	if (!vfs_files_init()) {
278	ipc_answer_0(rid, ENOMEM);	278	ipc_answer_0(rid, ENOMEM);
279	return;	279	return;
280	}	280	}
281		281
282	/*	282	/*
283	* The POSIX interface is open(path, oflag, mode).	283	* The POSIX interface is open(path, oflag, mode).
284	* We can receive oflags and mode along with the VFS_OPEN call; the path	284	* We can receive oflags and mode along with the VFS_OPEN call; the path
285	* will need to arrive in another call.	285	* will need to arrive in another call.
286	*	286	*
287	* We also receive one private, non-POSIX set of flags called lflag	287	* We also receive one private, non-POSIX set of flags called lflag
288	* used to pass information to vfs_lookup_internal().	288	* used to pass information to vfs_lookup_internal().
289	*/	289	*/
290	int lflag = IPC_GET_ARG1(*request);	290	int lflag = IPC_GET_ARG1(*request);
291	int oflag = IPC_GET_ARG2(*request);	291	int oflag = IPC_GET_ARG2(*request);
292	int mode = IPC_GET_ARG3(*request);	292	int mode = IPC_GET_ARG3(*request);
293	size_t len;	293	size_t len;
294		294
295	if (oflag & O_CREAT)	295	if (oflag & O_CREAT)
296	lflag \|= L_CREATE;	296	lflag \|= L_CREATE;
297	if (oflag & O_EXCL)	297	if (oflag & O_EXCL)
298	lflag \|= L_EXCLUSIVE;	298	lflag \|= L_EXCLUSIVE;
299		299
300	ipc_callid_t callid;	300	ipc_callid_t callid;
301		301
302	if (!ipc_data_write_receive(&callid, &len)) {	302	if (!ipc_data_write_receive(&callid, &len)) {
303	ipc_answer_0(callid, EINVAL);	303	ipc_answer_0(callid, EINVAL);
304	ipc_answer_0(rid, EINVAL);	304	ipc_answer_0(rid, EINVAL);
305	return;	305	return;
306	}	306	}
307		307
308	/*	308	/*
309	* Now we are on the verge of accepting the path.	309	* Now we are on the verge of accepting the path.
310	*	310	*
311	* There is one optimization we could do in the future: copy the path	311	* There is one optimization we could do in the future: copy the path
312	* directly into the PLB using some kind of a callback.	312	* directly into the PLB using some kind of a callback.
313	*/	313	*/
314	char *path = malloc(len);	314	char *path = malloc(len);
315		315
316	if (!path) {	316	if (!path) {
317	ipc_answer_0(callid, ENOMEM);	317	ipc_answer_0(callid, ENOMEM);
318	ipc_answer_0(rid, ENOMEM);	318	ipc_answer_0(rid, ENOMEM);
319	return;	319	return;
320	}	320	}
321		321
322	int rc;	322	int rc;
323	if ((rc = ipc_data_write_finalize(callid, path, len))) {	323	if ((rc = ipc_data_write_finalize(callid, path, len))) {
324	ipc_answer_0(rid, rc);	324	ipc_answer_0(rid, rc);
325	free(path);	325	free(path);
326	return;	326	return;
327	}	327	}
328		328
329	/*	329	/*
330	* Avoid the race condition in which the file can be deleted before we	330	* Avoid the race condition in which the file can be deleted before we
331	* find/create-and-lock the VFS node corresponding to the looked-up	331	* find/create-and-lock the VFS node corresponding to the looked-up
332	* triplet.	332	* triplet.
333	*/	333	*/
334	if (lflag & L_CREATE)	334	if (lflag & L_CREATE)
335	rwlock_write_lock(&namespace_rwlock);	335	rwlock_write_lock(&namespace_rwlock);
336	else	336	else
337	rwlock_read_lock(&namespace_rwlock);	337	rwlock_read_lock(&namespace_rwlock);
338		338
339	/* The path is now populated and we can call vfs_lookup_internal(). */	339	/* The path is now populated and we can call vfs_lookup_internal(). */
340	vfs_lookup_res_t lr;	340	vfs_lookup_res_t lr;
341	rc = vfs_lookup_internal(path, len, lflag, &lr, NULL);	341	rc = vfs_lookup_internal(path, len, lflag, &lr, NULL);
342	if (rc) {	342	if (rc) {
343	if (lflag & L_CREATE)	343	if (lflag & L_CREATE)
344	rwlock_write_unlock(&namespace_rwlock);	344	rwlock_write_unlock(&namespace_rwlock);
345	else	345	else
346	rwlock_read_unlock(&namespace_rwlock);	346	rwlock_read_unlock(&namespace_rwlock);
347	ipc_answer_0(rid, rc);	347	ipc_answer_0(rid, rc);
348	free(path);	348	free(path);
349	return;	349	return;
350	}	350	}
351		351
352	/** Path is no longer needed. */	352	/** Path is no longer needed. */
353	free(path);	353	free(path);
354		354
355	vfs_node_t *node = vfs_node_get(&lr);	355	vfs_node_t *node = vfs_node_get(&lr);
356	if (lflag & L_CREATE)	356	if (lflag & L_CREATE)
357	rwlock_write_unlock(&namespace_rwlock);	357	rwlock_write_unlock(&namespace_rwlock);
358	else	358	else
359	rwlock_read_unlock(&namespace_rwlock);	359	rwlock_read_unlock(&namespace_rwlock);
360		360
361	/*	361	/*
362	* Get ourselves a file descriptor and the corresponding vfs_file_t	362	* Get ourselves a file descriptor and the corresponding vfs_file_t
363	* structure.	363	* structure.
364	*/	364	*/
365	int fd = vfs_fd_alloc();	365	int fd = vfs_fd_alloc();
366	if (fd < 0) {	366	if (fd < 0) {
367	vfs_node_put(node);	367	vfs_node_put(node);
368	ipc_answer_0(rid, fd);	368	ipc_answer_0(rid, fd);
369	return;	369	return;
370	}	370	}
371	vfs_file_t *file = vfs_file_get(fd);	371	vfs_file_t *file = vfs_file_get(fd);
372	file->node = node;	372	file->node = node;
373	if (oflag & O_APPEND)	373	if (oflag & O_APPEND)
374	file->append = true;	374	file->append = true;
375		375
376	/*	376	/*
377	* The following increase in reference count is for the fact that the	377	* The following increase in reference count is for the fact that the
378	* file is being opened and that a file structure is pointing to it.	378	* file is being opened and that a file structure is pointing to it.
379	* It is necessary so that the file will not disappear when	379	* It is necessary so that the file will not disappear when
380	* vfs_node_put() is called. The reference will be dropped by the	380	* vfs_node_put() is called. The reference will be dropped by the
381	* respective VFS_CLOSE.	381	* respective VFS_CLOSE.
382	*/	382	*/
383	vfs_node_addref(node);	383	vfs_node_addref(node);
384	vfs_node_put(node);	384	vfs_node_put(node);
385		385
386	/* Success! Return the new file descriptor to the client. */	386	/* Success! Return the new file descriptor to the client. */
387	ipc_answer_1(rid, EOK, fd);	387	ipc_answer_1(rid, EOK, fd);
388	}	388	}
389		389
390	void vfs_close(ipc_callid_t rid, ipc_call_t *request)	390	void vfs_close(ipc_callid_t rid, ipc_call_t *request)
391	{	391	{
392	int fd = IPC_GET_ARG1(*request);	392	int fd = IPC_GET_ARG1(*request);
393	if (fd >= MAX_OPEN_FILES) {	393	if (fd >= MAX_OPEN_FILES) {
394	ipc_answer_0(rid, EBADF);	394	ipc_answer_0(rid, EBADF);
395	return;	395	return;
396	}	396	}
397	vfs_fd_free(fd);	397	vfs_fd_free(fd);
398	ipc_answer_0(rid, EOK);	398	ipc_answer_0(rid, EOK);
399	}	399	}
400		400
401	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)	401	static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
402	{	402	{
403		403
404	/*	404	/*
405	* The following code strongly depends on the fact that the files data	405	* The following code strongly depends on the fact that the files data
406	* structure can be only accessed by a single fibril and all file	406	* structure can be only accessed by a single fibril and all file
407	* operations are serialized (i.e. the reads and writes cannot	407	* operations are serialized (i.e. the reads and writes cannot
408	* interleave and a file cannot be closed while it is being read).	408	* interleave and a file cannot be closed while it is being read).
409	*	409	*
410	* Additional synchronization needs to be added once the table of	410	* Additional synchronization needs to be added once the table of
411	* open files supports parallel access!	411	* open files supports parallel access!
412	*/	412	*/
413		413
414	int fd = IPC_GET_ARG1(*request);	414	int fd = IPC_GET_ARG1(*request);
415		415
416	/* Lookup the file structure corresponding to the file descriptor. */	416	/* Lookup the file structure corresponding to the file descriptor. */
417	vfs_file_t *file = vfs_file_get(fd);	417	vfs_file_t *file = vfs_file_get(fd);
418	if (!file) {	418	if (!file) {
419	ipc_answer_0(rid, ENOENT);	419	ipc_answer_0(rid, ENOENT);
420	return;	420	return;
421	}	421	}
422		422
423	/*	423	/*
424	* Now we need to receive a call with client's	424	* Now we need to receive a call with client's
425	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.	425	* IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
426	*/	426	*/
427	ipc_callid_t callid;	427	ipc_callid_t callid;
428	int res;	428	int res;
429	if (read)	429	if (read)
430	res = ipc_data_read_receive(&callid, NULL);	430	res = ipc_data_read_receive(&callid, NULL);
431	else	431	else
432	res = ipc_data_write_receive(&callid, NULL);	432	res = ipc_data_write_receive(&callid, NULL);
433	if (!res) {	433	if (!res) {
434	ipc_answer_0(callid, EINVAL);	434	ipc_answer_0(callid, EINVAL);
435	ipc_answer_0(rid, EINVAL);	435	ipc_answer_0(rid, EINVAL);
436	return;	436	return;
437	}	437	}
438		438
439	/*	439	/*
440	* Lock the open file structure so that no other thread can manipulate	440	* Lock the open file structure so that no other thread can manipulate
441	* the same open file at a time.	441	* the same open file at a time.
442	*/	442	*/
443	futex_down(&file->lock);	443	futex_down(&file->lock);
444		444
445	/*	445	/*
446	* Lock the file's node so that no other client can read/write to it at	446	* Lock the file's node so that no other client can read/write to it at
447	* the same time.	447	* the same time.
448	*/	448	*/
449	if (read)	449	if (read)
450	rwlock_read_lock(&file->node->contents_rwlock);	450	rwlock_read_lock(&file->node->contents_rwlock);
451	else	451	else
452	rwlock_write_lock(&file->node->contents_rwlock);	452	rwlock_write_lock(&file->node->contents_rwlock);
453		453
454	int fs_phone = vfs_grab_phone(file->node->fs_handle);	454	int fs_phone = vfs_grab_phone(file->node->fs_handle);
455		455
456	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */	456	/* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
457	aid_t msg;	457	aid_t msg;
458	ipc_call_t answer;	458	ipc_call_t answer;
459	if (!read && file->append)	459	if (!read && file->append)
460	file->pos = file->node->size;	460	file->pos = file->node->size;
461	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),	461	msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
462	file->node->dev_handle, file->node->index, file->pos, &answer);	462	file->node->dev_handle, file->node->index, file->pos, &answer);
463		463
464	/*	464	/*
465	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the	465	* Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
466	* destination FS server. The call will be routed as if sent by	466	* destination FS server. The call will be routed as if sent by
467	* ourselves. Note that call arguments are immutable in this case so we	467	* ourselves. Note that call arguments are immutable in this case so we
468	* don't have to bother.	468	* don't have to bother.
469	*/	469	*/
470	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);	470	ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
471		471
472	vfs_release_phone(fs_phone);	472	vfs_release_phone(fs_phone);
473		473
474	/* Wait for reply from the FS server. */	474	/* Wait for reply from the FS server. */
475	ipcarg_t rc;	475	ipcarg_t rc;
476	async_wait_for(msg, &rc);	476	async_wait_for(msg, &rc);
477	size_t bytes = IPC_GET_ARG1(answer);	477	size_t bytes = IPC_GET_ARG1(answer);
478		478
479	/* Unlock the VFS node. */	479	/* Unlock the VFS node. */
480	if (read)	480	if (read)
481	rwlock_read_unlock(&file->node->contents_rwlock);	481	rwlock_read_unlock(&file->node->contents_rwlock);
482	else {	482	else {
483	/* Update the cached version of node's size. */	483	/* Update the cached version of node's size. */
484	if (rc == EOK)	484	if (rc == EOK)
485	file->node->size = IPC_GET_ARG2(answer);	485	file->node->size = IPC_GET_ARG2(answer);
486	rwlock_write_unlock(&file->node->contents_rwlock);	486	rwlock_write_unlock(&file->node->contents_rwlock);
487	}	487	}
488		488
489	/* Update the position pointer and unlock the open file. */	489	/* Update the position pointer and unlock the open file. */
490	if (rc == EOK)	490	if (rc == EOK)
491	file->pos += bytes;	491	file->pos += bytes;
492	futex_up(&file->lock);	492	futex_up(&file->lock);
493		493
494	/*	494	/*
495	* FS server's reply is the final result of the whole operation we	495	* FS server's reply is the final result of the whole operation we
496	* return to the client.	496	* return to the client.
497	*/	497	*/
498	ipc_answer_1(rid, rc, bytes);	498	ipc_answer_1(rid, rc, bytes);
499	}	499	}
500		500

Subversion Repositories HelenOS

(root)/trunk/uspace/srv/vfs/vfs_ops.c – Rev 2735 → 2742