Subversion Repositories HelenOS

/*

 * 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 <ipc/services.h>

#include <async.h>

#include <fibril.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 <as.h>

#include <assert.h>

#include <atomic.h>

#include "vfs.h"

#define min(a, b)   ((a) < (b) ? (a) : (b))

/**

 * 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 plb_futex = FUTEX_INITIALIZER;

link_t plb_head;    /**< PLB entry ring buffer. */

uint8_t *plb = NULL;

/** Perform a path lookup.

 *

 * @param path      Path to be resolved; it needn't be an ASCIIZ string.

 * @param len       Number of path characters pointed by path.

 * @param result    Empty structure where the lookup result will be stored.

 * @param altroot   If non-empty, will be used instead of rootfs as the root

 *          of the whole VFS tree.

 *

 * @return      EOK on success or an error code from errno.h.

 */

int vfs_lookup_internal(char *path, size_t len, vfs_lookup_res_t *result,

    vfs_pair_t *altroot)

{

    vfs_pair_t *root;

    if (!len)

        return EINVAL;

    if (altroot)

        root = altroot;

    else

        root = (vfs_pair_t *) &rootfs;

    if (!root->fs_handle)

        return ENOENT;

    futex_down(&plb_futex);

    plb_entry_t entry;

    link_initialize(&entry.plb_link);

    entry.len = len;

    off_t first;    /* the first free index */

    off_t last; /* the last free index */

    if (list_empty(&plb_head)) {

        first = 0;

        last = PLB_SIZE - 1;

    } else {

        plb_entry_t *oldest = list_get_instance(plb_head.next,

            plb_entry_t, plb_link);

        plb_entry_t *newest = list_get_instance(plb_head.prev,

            plb_entry_t, plb_link);

        first = (newest->index + newest->len) % PLB_SIZE;

        last = (oldest->index - 1) % PLB_SIZE;

    }

    if (first <= last) {

        if ((last - first) + 1 < len) {

            /*

             * The buffer cannot absorb the path.

             */

            futex_up(&plb_futex);

            return ELIMIT;

        }

    } else {

        if (PLB_SIZE - ((first - last) + 1) < len) {

            /*

             * The buffer cannot absorb the path.

             */

            futex_up(&plb_futex);

            return ELIMIT;

        }

    }

    /*

     * We know the first free index in PLB and we also know that there is

     * enough space in the buffer to hold our path.

     */

    entry.index = first;

    entry.len = len;

    /*

     * Claim PLB space by inserting the entry into the PLB entry ring

     * buffer.

     */

    list_append(&entry.plb_link, &plb_head);

    futex_up(&plb_futex);

    /*

     * Copy the path into PLB.

     */

    size_t cnt1 = min(len, (PLB_SIZE - first) + 1);

    size_t cnt2 = len - cnt1;

    memcpy(&plb[first], path, cnt1);

    memcpy(plb, &path[cnt1], cnt2);

    ipc_call_t answer;

    int phone = vfs_grab_phone(root->fs_handle);

    aid_t req = async_send_3(phone, VFS_LOOKUP, (ipcarg_t) first,

        (ipcarg_t) (first + len - 1) % PLB_SIZE,

        (ipcarg_t) root->dev_handle, &answer);

    vfs_release_phone(phone);

    ipcarg_t rc;

    async_wait_for(req, &rc);

    futex_down(&plb_futex);

    list_remove(&entry.plb_link);

    /*

     * Erasing the path from PLB will come handy for debugging purposes.

     */

    memset(&plb[first], 0, cnt1);

    memset(plb, 0, cnt2);

    futex_up(&plb_futex);

    if (rc == EOK) {

        result->triplet.fs_handle = (int) IPC_GET_ARG1(answer);

        result->triplet.dev_handle = (int) IPC_GET_ARG2(answer);

        result->triplet.index = (int) IPC_GET_ARG3(answer);

        result->size = (size_t) IPC_GET_ARG4(answer);

    }

    return rc;

}

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("/"), 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, &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, flags, mode).

     * We can receive flags and mode along with the VFS_OPEN call; the path

     * will need to arrive in another call.

     */

    int flags = IPC_GET_ARG1(*request);

    int mode = IPC_GET_ARG2(*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;

    }

    /*

     * 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.

     */

    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, &lr, NULL);

    if (rc) {

        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);

    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;

    /*

     * 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);

}

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;

    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. */

        file->node->size = IPC_GET_ARG2(answer);

        rwlock_write_unlock(&file->node->contents_rwlock);

    }

    /*

     * Update the position pointer and unlock the open file.

     */

    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);

}

atomic_t fs_head_futex = FUTEX_INITIALIZER;

link_t fs_head;

atomic_t fs_handle_next = {

    .count = 1

};

/** Verify the VFS info structure.

 *

 * @param info      Info structure to be verified.

 *

 * @return      Non-zero if the info structure is sane, zero otherwise.

 */

static bool vfs_info_sane(vfs_info_t *info)

{

    int i;

    /*

     * Check if the name is non-empty and is composed solely of ASCII

     * characters [a-z]+[a-z0-9_-]*.

     */

    if (!islower(info->name[0])) {

        dprintf("The name doesn't start with a lowercase character.\n");

        return false;

    }

    for (i = 1; i < FS_NAME_MAXLEN; i++) {

        if (!(islower(info->name[i]) || isdigit(info->name[i])) &&

            (info->name[i] != '-') && (info->name[i] != '_')) {

            if (info->name[i] == '\0') {

                break;

            } else {

                dprintf("The name contains illegal "

                    "characters.\n");

                return false;

            }

        }

    }

    /*

     * This check is not redundant. It ensures that the name is

     * NULL-terminated, even if FS_NAME_MAXLEN characters are used.

     */

    if (info->name[i] != '\0') {

        dprintf("The name is not properly NULL-terminated.\n");

        return false;

    }

    /*

     * Check if the FS implements mandatory VFS operations.

     */

    if (info->ops[IPC_METHOD_TO_VFS_OP(VFS_LOOKUP)] != VFS_OP_DEFINED) {

        dprintf("Operation VFS_LOOKUP not defined by the client.\n");

        return false;

    }

    if (info->ops[IPC_METHOD_TO_VFS_OP(VFS_OPEN)] != VFS_OP_DEFINED) {

        dprintf("Operation VFS_OPEN not defined by the client.\n");

        return false;

    }

    if (info->ops[IPC_METHOD_TO_VFS_OP(VFS_CLOSE)] != VFS_OP_DEFINED) {

        dprintf("Operation VFS_CLOSE not defined by the client.\n");

        return false;

    }

    if (info->ops[IPC_METHOD_TO_VFS_OP(VFS_READ)] != VFS_OP_DEFINED) {

        dprintf("Operation VFS_READ not defined by the client.\n");

        return false;

    }

    /*

     * Check if each operation is either not defined, defined or default.

     */

    for (i = VFS_FIRST; i < VFS_LAST_CLNT; i++) {

        if ((info->ops[IPC_METHOD_TO_VFS_OP(i)] != VFS_OP_NULL) &&

            (info->ops[IPC_METHOD_TO_VFS_OP(i)] != VFS_OP_DEFAULT) &&

            (info->ops[IPC_METHOD_TO_VFS_OP(i)] != VFS_OP_DEFINED)) {

            dprintf("Operation info not understood.\n");

            return false;

        }

    }

    return true;

}

/** VFS_REGISTER protocol function.

 *

 * @param rid       Hash of the call with the request.

 * @param request   Call structure with the request.

 */

void vfs_register(ipc_callid_t rid, ipc_call_t *request)

{

    ipc_callid_t callid;

    ipc_call_t call;

    int rc;

    size_t size;

    dprintf("Processing VFS_REGISTER request received from %p.\n",

        request->in_phone_hash);

    /*

     * The first call has to be IPC_M_DATA_SEND in which we receive the

     * VFS info structure from the client FS.

     */

    if (!ipc_data_write_receive(&callid, &size)) {

        /*

         * The client doesn't obey the same protocol as we do.

         */

        dprintf("Receiving of VFS info failed.\n");

        ipc_answer_0(callid, EINVAL);

        ipc_answer_0(rid, EINVAL);

        return;

    }

    dprintf("VFS info received, size = %d\n", size);

    /*

     * We know the size of the VFS info structure. See if the client

     * understands this easy concept too.

     */

    if (size != sizeof(vfs_info_t)) {

        /*

         * The client is sending us something, which cannot be

         * the info structure.

         */

        dprintf("Received VFS info has bad size.\n");

        ipc_answer_0(callid, EINVAL);

        ipc_answer_0(rid, EINVAL);

        return;

    }

    /*

     * Allocate and initialize a buffer for the fs_info structure.

     */

    fs_info_t *fs_info;

    fs_info = (fs_info_t *) malloc(sizeof(fs_info_t));

    if (!fs_info) {