Subversion Repositories HelenOS

Rev

Rev 3352 | Rev 4010 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

  1. /*
  2.  * Copyright (c) 2008 Jakub Jermar
  3.  * All rights reserved.
  4.  *
  5.  * Redistribution and use in source and binary forms, with or without
  6.  * modification, are permitted provided that the following conditions
  7.  * are met:
  8.  *
  9.  * - Redistributions of source code must retain the above copyright
  10.  *   notice, this list of conditions and the following disclaimer.
  11.  * - Redistributions in binary form must reproduce the above copyright
  12.  *   notice, this list of conditions and the following disclaimer in the
  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
  15.  *   derived from this software without specific prior written permission.
  16.  *
  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
  19.  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  20.  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  21.  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  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
  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
  26.  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27.  */
  28.  
  29. /** @addtogroup fs
  30.  * @{
  31.  */
  32.  
  33. /**
  34.  * @file    vfs_ops.c
  35.  * @brief   Operations that VFS offers to its clients.
  36.  */
  37.  
  38. #include "vfs.h"
  39. #include <ipc/ipc.h>
  40. #include <async.h>
  41. #include <errno.h>
  42. #include <stdio.h>
  43. #include <stdlib.h>
  44. #include <string.h>
  45. #include <bool.h>
  46. #include <futex.h>
  47. #include <rwlock.h>
  48. #include <libadt/list.h>
  49. #include <unistd.h>
  50. #include <ctype.h>
  51. #include <fcntl.h>
  52. #include <assert.h>
  53. #include <vfs/canonify.h>
  54.  
  55. /* Forward declarations of static functions. */
  56. static int vfs_truncate_internal(fs_handle_t, dev_handle_t, fs_index_t, size_t);
  57.  
  58. /**
  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.
  61.  */
  62. RWLOCK_INITIALIZE(namespace_rwlock);
  63.  
  64. futex_t rootfs_futex = FUTEX_INITIALIZER;
  65. vfs_pair_t rootfs = {
  66.     .fs_handle = 0,
  67.     .dev_handle = 0
  68. };
  69.  
  70. void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
  71. {
  72.     dev_handle_t dev_handle;
  73.     vfs_node_t *mp_node = NULL;
  74.     ipc_callid_t callid;
  75.     ipc_call_t data;
  76.     int rc;
  77.     int phone;
  78.     size_t size;
  79.  
  80.     /*
  81.      * We expect the library to do the device-name to device-handle
  82.      * translation for us, thus the device handle will arrive as ARG1
  83.      * in the request.
  84.      */
  85.     dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);
  86.  
  87.     /*
  88.      * For now, don't make use of ARG2 and ARG3, but they can be used to
  89.      * carry mount options in the future.
  90.      */
  91.  
  92.     /*
  93.      * Now, we expect the client to send us data with the name of the file
  94.      * system.
  95.      */
  96.     if (!ipc_data_write_receive(&callid, &size)) {
  97.         ipc_answer_0(callid, EINVAL);
  98.         ipc_answer_0(rid, EINVAL);
  99.         return;
  100.     }
  101.  
  102.     /*
  103.      * Don't receive more than is necessary for storing a full file system
  104.      * name.
  105.      */
  106.     if (size < 1 || size > FS_NAME_MAXLEN) {
  107.         ipc_answer_0(callid, EINVAL);
  108.         ipc_answer_0(rid, EINVAL);
  109.         return;
  110.     }
  111.  
  112.     /* Deliver the file system name. */
  113.     char fs_name[FS_NAME_MAXLEN + 1];
  114.     (void) ipc_data_write_finalize(callid, fs_name, size);
  115.     fs_name[size] = '\0';
  116.    
  117.     /*
  118.      * Wait for IPC_M_PING so that we can return an error if we don't know
  119.      * fs_name.
  120.      */
  121.     callid = async_get_call(&data);
  122.     if (IPC_GET_METHOD(data) != IPC_M_PING) {
  123.         ipc_answer_0(callid, ENOTSUP);
  124.         ipc_answer_0(rid, ENOTSUP);
  125.         return;
  126.     }
  127.  
  128.     /*
  129.      * Check if we know a file system with the same name as is in fs_name.
  130.      * This will also give us its file system handle.
  131.      */
  132.     fs_handle_t fs_handle = fs_name_to_handle(fs_name, true);
  133.     if (!fs_handle) {
  134.         ipc_answer_0(callid, ENOENT);
  135.         ipc_answer_0(rid, ENOENT);
  136.         return;
  137.     }
  138.  
  139.     /* Acknowledge that we know fs_name. */
  140.     ipc_answer_0(callid, EOK);
  141.  
  142.     /* Now, we want the client to send us the mount point. */
  143.     if (!ipc_data_write_receive(&callid, &size)) {
  144.         ipc_answer_0(callid, EINVAL);
  145.         ipc_answer_0(rid, EINVAL);
  146.         return;
  147.     }
  148.  
  149.     /* Check whether size is reasonable wrt. the mount point. */
  150.     if (size < 1 || size > MAX_PATH_LEN) {
  151.         ipc_answer_0(callid, EINVAL);
  152.         ipc_answer_0(rid, EINVAL);
  153.         return;
  154.     }
  155.     /* Allocate buffer for the mount point data being received. */
  156.     char *buf;
  157.     buf = malloc(size + 1);
  158.     if (!buf) {
  159.         ipc_answer_0(callid, ENOMEM);
  160.         ipc_answer_0(rid, ENOMEM);
  161.         return;
  162.     }
  163.  
  164.     /* Deliver the mount point. */
  165.     (void) ipc_data_write_finalize(callid, buf, size);
  166.     buf[size] = '\0';
  167.  
  168.     /* Resolve the path to the mountpoint. */
  169.     vfs_lookup_res_t mp_res;
  170.     futex_down(&rootfs_futex);
  171.     if (rootfs.fs_handle) {
  172.         /* We already have the root FS. */
  173.         rwlock_write_lock(&namespace_rwlock);
  174.         if ((size == 1) && (buf[0] == '/')) {
  175.             /* Trying to mount root FS over root FS */
  176.             rwlock_write_unlock(&namespace_rwlock);
  177.             futex_up(&rootfs_futex);
  178.             free(buf);
  179.             ipc_answer_0(rid, EBUSY);
  180.             return;
  181.         }
  182.         rc = vfs_lookup_internal(buf, L_DIRECTORY, &mp_res, NULL);
  183.         if (rc != EOK) {
  184.             /* The lookup failed for some reason. */
  185.             rwlock_write_unlock(&namespace_rwlock);
  186.             futex_up(&rootfs_futex);
  187.             free(buf);
  188.             ipc_answer_0(rid, rc);
  189.             return;
  190.         }
  191.         mp_node = vfs_node_get(&mp_res);
  192.         if (!mp_node) {
  193.             rwlock_write_unlock(&namespace_rwlock);
  194.             futex_up(&rootfs_futex);
  195.             free(buf);
  196.             ipc_answer_0(rid, ENOMEM);
  197.             return;
  198.         }
  199.         /*
  200.          * Now we hold a reference to mp_node.
  201.          * It will be dropped upon the corresponding VFS_UNMOUNT.
  202.          * This prevents the mount point from being deleted.
  203.          */
  204.         rwlock_write_unlock(&namespace_rwlock);
  205.     } else {
  206.         /* We still don't have the root file system mounted. */
  207.         if ((size == 1) && (buf[0] == '/')) {
  208.             vfs_lookup_res_t mr_res;
  209.             vfs_node_t *mr_node;
  210.             ipcarg_t rindex;
  211.             ipcarg_t rsize;
  212.             ipcarg_t rlnkcnt;
  213.        
  214.             /*
  215.              * For this simple, but important case,
  216.              * we are almost done.
  217.              */
  218.             free(buf);
  219.            
  220.             /* Tell the mountee that it is being mounted. */
  221.             phone = vfs_grab_phone(fs_handle);
  222.             rc = async_req_1_3(phone, VFS_MOUNTED,
  223.                 (ipcarg_t) dev_handle, &rindex, &rsize, &rlnkcnt);
  224.             vfs_release_phone(phone);
  225.            
  226.             if (rc != EOK) {
  227.                 futex_up(&rootfs_futex);
  228.                 ipc_answer_0(rid, rc);
  229.                 return;
  230.             }
  231.  
  232.             mr_res.triplet.fs_handle = fs_handle;
  233.             mr_res.triplet.dev_handle = dev_handle;
  234.             mr_res.triplet.index = (fs_index_t) rindex;
  235.             mr_res.size = (size_t) rsize;
  236.             mr_res.lnkcnt = (unsigned) rlnkcnt;
  237.  
  238.             rootfs.fs_handle = fs_handle;
  239.             rootfs.dev_handle = dev_handle;
  240.             futex_up(&rootfs_futex);
  241.  
  242.             /* Add reference to the mounted root. */
  243.             mr_node = vfs_node_get(&mr_res);
  244.             assert(mr_node);
  245.  
  246.             ipc_answer_0(rid, rc);
  247.             return;
  248.         } else {
  249.             /*
  250.              * We can't resolve this without the root filesystem
  251.              * being mounted first.
  252.              */
  253.             futex_up(&rootfs_futex);
  254.             free(buf);
  255.             ipc_answer_0(rid, ENOENT);
  256.             return;
  257.         }
  258.     }
  259.     futex_up(&rootfs_futex);
  260.    
  261.     free(buf);  /* The buffer is not needed anymore. */
  262.    
  263.     /*
  264.      * At this point, we have all necessary pieces: file system and device
  265.      * handles, and we know the mount point VFS node.
  266.      */
  267.  
  268.     phone = vfs_grab_phone(mp_res.triplet.fs_handle);
  269.     rc = async_req_4_0(phone, VFS_MOUNT,
  270.         (ipcarg_t) mp_res.triplet.dev_handle,
  271.         (ipcarg_t) mp_res.triplet.index,
  272.         (ipcarg_t) fs_handle,
  273.         (ipcarg_t) dev_handle);
  274.     vfs_release_phone(phone);
  275.  
  276.     if (rc != EOK) {
  277.         /* Mount failed, drop reference to mp_node. */
  278.         if (mp_node)
  279.             vfs_node_put(mp_node);
  280.     }
  281.    
  282.     ipc_answer_0(rid, rc);
  283. }
  284.  
  285. void vfs_open(ipc_callid_t rid, ipc_call_t *request)
  286. {
  287.     if (!vfs_files_init()) {
  288.         ipc_answer_0(rid, ENOMEM);
  289.         return;
  290.     }
  291.  
  292.     /*
  293.      * The POSIX interface is open(path, oflag, mode).
  294.      * We can receive oflags and mode along with the VFS_OPEN call; the path
  295.      * will need to arrive in another call.
  296.      *
  297.      * We also receive one private, non-POSIX set of flags called lflag
  298.      * used to pass information to vfs_lookup_internal().
  299.      */
  300.     int lflag = IPC_GET_ARG1(*request);
  301.     int oflag = IPC_GET_ARG2(*request);
  302.     int mode = IPC_GET_ARG3(*request);
  303.     size_t len;
  304.  
  305.     if (oflag & O_CREAT)
  306.         lflag |= L_CREATE;
  307.     if (oflag & O_EXCL)
  308.         lflag |= L_EXCLUSIVE;
  309.  
  310.     ipc_callid_t callid;
  311.  
  312.     if (!ipc_data_write_receive(&callid, &len)) {
  313.         ipc_answer_0(callid, EINVAL);
  314.         ipc_answer_0(rid, EINVAL);
  315.         return;
  316.     }
  317.     char *path = malloc(len + 1);
  318.     if (!path) {
  319.         ipc_answer_0(callid, ENOMEM);
  320.         ipc_answer_0(rid, ENOMEM);
  321.         return;
  322.     }
  323.     int rc;
  324.     if ((rc = ipc_data_write_finalize(callid, path, len))) {
  325.         ipc_answer_0(rid, rc);
  326.         free(path);
  327.         return;
  328.     }
  329.     path[len] = '\0';
  330.    
  331.     /*
  332.      * Avoid the race condition in which the file can be deleted before we
  333.      * find/create-and-lock the VFS node corresponding to the looked-up
  334.      * triplet.
  335.      */
  336.     if (lflag & L_CREATE)
  337.         rwlock_write_lock(&namespace_rwlock);
  338.     else
  339.         rwlock_read_lock(&namespace_rwlock);
  340.  
  341.     /* The path is now populated and we can call vfs_lookup_internal(). */
  342.     vfs_lookup_res_t lr;
  343.     rc = vfs_lookup_internal(path, lflag, &lr, NULL);
  344.     if (rc) {
  345.         if (lflag & L_CREATE)
  346.             rwlock_write_unlock(&namespace_rwlock);
  347.         else
  348.             rwlock_read_unlock(&namespace_rwlock);
  349.         ipc_answer_0(rid, rc);
  350.         free(path);
  351.         return;
  352.     }
  353.  
  354.     /* Path is no longer needed. */
  355.     free(path);
  356.  
  357.     vfs_node_t *node = vfs_node_get(&lr);
  358.     if (lflag & L_CREATE)
  359.         rwlock_write_unlock(&namespace_rwlock);
  360.     else
  361.         rwlock_read_unlock(&namespace_rwlock);
  362.  
  363.     /* Truncate the file if requested and if necessary. */
  364.     if (oflag & O_TRUNC) {
  365.         rwlock_write_lock(&node->contents_rwlock);
  366.         if (node->size) {
  367.             rc = vfs_truncate_internal(node->fs_handle,
  368.                 node->dev_handle, node->index, 0);
  369.             if (rc) {
  370.                 rwlock_write_unlock(&node->contents_rwlock);
  371.                 vfs_node_put(node);
  372.                 ipc_answer_0(rid, rc);
  373.                 return;
  374.             }
  375.             node->size = 0;
  376.         }
  377.         rwlock_write_unlock(&node->contents_rwlock);
  378.     }
  379.  
  380.     /*
  381.      * Get ourselves a file descriptor and the corresponding vfs_file_t
  382.      * structure.
  383.      */
  384.     int fd = vfs_fd_alloc();
  385.     if (fd < 0) {
  386.         vfs_node_put(node);
  387.         ipc_answer_0(rid, fd);
  388.         return;
  389.     }
  390.     vfs_file_t *file = vfs_file_get(fd);
  391.     file->node = node;
  392.     if (oflag & O_APPEND)
  393.         file->append = true;
  394.  
  395.     /*
  396.      * The following increase in reference count is for the fact that the
  397.      * file is being opened and that a file structure is pointing to it.
  398.      * It is necessary so that the file will not disappear when
  399.      * vfs_node_put() is called. The reference will be dropped by the
  400.      * respective VFS_CLOSE.
  401.      */
  402.     vfs_node_addref(node);
  403.     vfs_node_put(node);
  404.  
  405.     /* Success! Return the new file descriptor to the client. */
  406.     ipc_answer_1(rid, EOK, fd);
  407. }
  408.  
  409. void vfs_close(ipc_callid_t rid, ipc_call_t *request)
  410. {
  411.     int fd = IPC_GET_ARG1(*request);
  412.     int rc = vfs_fd_free(fd);
  413.     ipc_answer_0(rid, rc);
  414. }
  415.  
  416. static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
  417. {
  418.  
  419.     /*
  420.      * The following code strongly depends on the fact that the files data
  421.      * structure can be only accessed by a single fibril and all file
  422.      * operations are serialized (i.e. the reads and writes cannot
  423.      * interleave and a file cannot be closed while it is being read).
  424.      *
  425.      * Additional synchronization needs to be added once the table of
  426.      * open files supports parallel access!
  427.      */
  428.  
  429.     int fd = IPC_GET_ARG1(*request);
  430.    
  431.     /* Lookup the file structure corresponding to the file descriptor. */
  432.     vfs_file_t *file = vfs_file_get(fd);
  433.     if (!file) {
  434.         ipc_answer_0(rid, ENOENT);
  435.         return;
  436.     }
  437.    
  438.     /*
  439.      * Now we need to receive a call with client's
  440.      * IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
  441.      */
  442.     ipc_callid_t callid;
  443.     int res;
  444.     if (read)
  445.         res = ipc_data_read_receive(&callid, NULL);
  446.     else
  447.         res = ipc_data_write_receive(&callid, NULL);
  448.     if (!res) {
  449.         ipc_answer_0(callid, EINVAL);
  450.         ipc_answer_0(rid, EINVAL);
  451.         return;
  452.     }
  453.    
  454.     /*
  455.      * Lock the open file structure so that no other thread can manipulate
  456.      * the same open file at a time.
  457.      */
  458.     futex_down(&file->lock);
  459.    
  460.     /*
  461.      * Lock the file's node so that no other client can read/write to it at
  462.      * the same time.
  463.      */
  464.     if (read)
  465.         rwlock_read_lock(&file->node->contents_rwlock);
  466.     else
  467.         rwlock_write_lock(&file->node->contents_rwlock);
  468.    
  469.     int fs_phone = vfs_grab_phone(file->node->fs_handle);  
  470.    
  471.     /* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
  472.     aid_t msg;
  473.     ipc_call_t answer;
  474.     if (!read && file->append)
  475.         file->pos = file->node->size;
  476.     msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
  477.         file->node->dev_handle, file->node->index, file->pos, &answer);
  478.    
  479.     /*
  480.      * Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
  481.      * destination FS server. The call will be routed as if sent by
  482.      * ourselves. Note that call arguments are immutable in this case so we
  483.      * don't have to bother.
  484.      */
  485.     ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
  486.    
  487.     vfs_release_phone(fs_phone);
  488.    
  489.     /* Wait for reply from the FS server. */
  490.     ipcarg_t rc;
  491.     async_wait_for(msg, &rc);
  492.     size_t bytes = IPC_GET_ARG1(answer);
  493.    
  494.     /* Unlock the VFS node. */
  495.     if (read)
  496.         rwlock_read_unlock(&file->node->contents_rwlock);
  497.     else {
  498.         /* Update the cached version of node's size. */
  499.         if (rc == EOK)
  500.             file->node->size = IPC_GET_ARG2(answer);
  501.         rwlock_write_unlock(&file->node->contents_rwlock);
  502.     }
  503.    
  504.     /* Update the position pointer and unlock the open file. */
  505.     if (rc == EOK)
  506.         file->pos += bytes;
  507.     futex_up(&file->lock);
  508.    
  509.     /*
  510.      * FS server's reply is the final result of the whole operation we
  511.      * return to the client.
  512.      */
  513.     ipc_answer_1(rid, rc, bytes);
  514. }
  515.  
  516. void vfs_read(ipc_callid_t rid, ipc_call_t *request)
  517. {
  518.     vfs_rdwr(rid, request, true);
  519. }
  520.  
  521. void vfs_write(ipc_callid_t rid, ipc_call_t *request)
  522. {
  523.     vfs_rdwr(rid, request, false);
  524. }
  525.  
  526. void vfs_seek(ipc_callid_t rid, ipc_call_t *request)
  527. {
  528.     int fd = (int) IPC_GET_ARG1(*request);
  529.     off_t off = (off_t) IPC_GET_ARG2(*request);
  530.     int whence = (int) IPC_GET_ARG3(*request);
  531.  
  532.  
  533.     /* Lookup the file structure corresponding to the file descriptor. */
  534.     vfs_file_t *file = vfs_file_get(fd);
  535.     if (!file) {
  536.         ipc_answer_0(rid, ENOENT);
  537.         return;
  538.     }
  539.  
  540.     off_t newpos;
  541.     futex_down(&file->lock);
  542.     if (whence == SEEK_SET) {
  543.         file->pos = off;
  544.         futex_up(&file->lock);
  545.         ipc_answer_1(rid, EOK, off);
  546.         return;
  547.     }
  548.     if (whence == SEEK_CUR) {
  549.         if (file->pos + off < file->pos) {
  550.             futex_up(&file->lock);
  551.             ipc_answer_0(rid, EOVERFLOW);
  552.             return;
  553.         }
  554.         file->pos += off;
  555.         newpos = file->pos;
  556.         futex_up(&file->lock);
  557.         ipc_answer_1(rid, EOK, newpos);
  558.         return;
  559.     }
  560.     if (whence == SEEK_END) {
  561.         rwlock_read_lock(&file->node->contents_rwlock);
  562.         size_t size = file->node->size;
  563.         rwlock_read_unlock(&file->node->contents_rwlock);
  564.         if (size + off < size) {
  565.             futex_up(&file->lock);
  566.             ipc_answer_0(rid, EOVERFLOW);
  567.             return;
  568.         }
  569.         newpos = size + off;
  570.         futex_up(&file->lock);
  571.         ipc_answer_1(rid, EOK, newpos);
  572.         return;
  573.     }
  574.     futex_up(&file->lock);
  575.     ipc_answer_0(rid, EINVAL);
  576. }
  577.  
  578. int
  579. vfs_truncate_internal(fs_handle_t fs_handle, dev_handle_t dev_handle,
  580.     fs_index_t index, size_t size)
  581. {
  582.     ipcarg_t rc;
  583.     int fs_phone;
  584.    
  585.     fs_phone = vfs_grab_phone(fs_handle);
  586.     rc = async_req_3_0(fs_phone, VFS_TRUNCATE, (ipcarg_t)dev_handle,
  587.         (ipcarg_t)index, (ipcarg_t)size);
  588.     vfs_release_phone(fs_phone);
  589.     return (int)rc;
  590. }
  591.  
  592. void vfs_truncate(ipc_callid_t rid, ipc_call_t *request)
  593. {
  594.     int fd = IPC_GET_ARG1(*request);
  595.     size_t size = IPC_GET_ARG2(*request);
  596.     int rc;
  597.  
  598.     vfs_file_t *file = vfs_file_get(fd);
  599.     if (!file) {
  600.         ipc_answer_0(rid, ENOENT);
  601.         return;
  602.     }
  603.     futex_down(&file->lock);
  604.  
  605.     rwlock_write_lock(&file->node->contents_rwlock);
  606.     rc = vfs_truncate_internal(file->node->fs_handle,
  607.         file->node->dev_handle, file->node->index, size);
  608.     if (rc == EOK)
  609.         file->node->size = size;
  610.     rwlock_write_unlock(&file->node->contents_rwlock);
  611.  
  612.     futex_up(&file->lock);
  613.     ipc_answer_0(rid, (ipcarg_t)rc);
  614. }
  615.  
  616. void vfs_mkdir(ipc_callid_t rid, ipc_call_t *request)
  617. {
  618.     int mode = IPC_GET_ARG1(*request);
  619.  
  620.     size_t len;
  621.     ipc_callid_t callid;
  622.  
  623.     if (!ipc_data_write_receive(&callid, &len)) {
  624.         ipc_answer_0(callid, EINVAL);
  625.         ipc_answer_0(rid, EINVAL);
  626.         return;
  627.     }
  628.     char *path = malloc(len + 1);
  629.     if (!path) {
  630.         ipc_answer_0(callid, ENOMEM);
  631.         ipc_answer_0(rid, ENOMEM);
  632.         return;
  633.     }
  634.     int rc;
  635.     if ((rc = ipc_data_write_finalize(callid, path, len))) {
  636.         ipc_answer_0(rid, rc);
  637.         free(path);
  638.         return;
  639.     }
  640.     path[len] = '\0';
  641.    
  642.     rwlock_write_lock(&namespace_rwlock);
  643.     int lflag = L_DIRECTORY | L_CREATE | L_EXCLUSIVE;
  644.     rc = vfs_lookup_internal(path, lflag, NULL, NULL);
  645.     rwlock_write_unlock(&namespace_rwlock);
  646.     free(path);
  647.     ipc_answer_0(rid, rc);
  648. }
  649.  
  650. void vfs_unlink(ipc_callid_t rid, ipc_call_t *request)
  651. {
  652.     int lflag = IPC_GET_ARG1(*request);
  653.  
  654.     size_t len;
  655.     ipc_callid_t callid;
  656.  
  657.     if (!ipc_data_write_receive(&callid, &len)) {
  658.         ipc_answer_0(callid, EINVAL);
  659.         ipc_answer_0(rid, EINVAL);
  660.         return;
  661.     }
  662.     char *path = malloc(len + 1);
  663.     if (!path) {
  664.         ipc_answer_0(callid, ENOMEM);
  665.         ipc_answer_0(rid, ENOMEM);
  666.         return;
  667.     }
  668.     int rc;
  669.     if ((rc = ipc_data_write_finalize(callid, path, len))) {
  670.         ipc_answer_0(rid, rc);
  671.         free(path);
  672.         return;
  673.     }
  674.     path[len] = '\0';
  675.    
  676.     rwlock_write_lock(&namespace_rwlock);
  677.     lflag &= L_DIRECTORY;   /* sanitize lflag */
  678.     vfs_lookup_res_t lr;
  679.     rc = vfs_lookup_internal(path, lflag | L_UNLINK, &lr, NULL);
  680.     free(path);
  681.     if (rc != EOK) {
  682.         rwlock_write_unlock(&namespace_rwlock);
  683.         ipc_answer_0(rid, rc);
  684.         return;
  685.     }
  686.  
  687.     /*
  688.      * The name has already been unlinked by vfs_lookup_internal().
  689.      * We have to get and put the VFS node to ensure that it is
  690.      * VFS_DESTROY'ed after the last reference to it is dropped.
  691.      */
  692.     vfs_node_t *node = vfs_node_get(&lr);
  693.     futex_down(&nodes_futex);
  694.     node->lnkcnt--;
  695.     futex_up(&nodes_futex);
  696.     rwlock_write_unlock(&namespace_rwlock);
  697.     vfs_node_put(node);
  698.     ipc_answer_0(rid, EOK);
  699. }
  700.  
  701. void vfs_rename(ipc_callid_t rid, ipc_call_t *request)
  702. {
  703.     size_t len;
  704.     ipc_callid_t callid;
  705.     int rc;
  706.  
  707.     /* Retrieve the old path. */
  708.     if (!ipc_data_write_receive(&callid, &len)) {
  709.         ipc_answer_0(callid, EINVAL);
  710.         ipc_answer_0(rid, EINVAL);
  711.         return;
  712.     }
  713.     char *old = malloc(len + 1);
  714.     if (!old) {
  715.         ipc_answer_0(callid, ENOMEM);
  716.         ipc_answer_0(rid, ENOMEM);
  717.         return;
  718.     }
  719.     if ((rc = ipc_data_write_finalize(callid, old, len))) {
  720.         ipc_answer_0(rid, rc);
  721.         free(old);
  722.         return;
  723.     }
  724.     old[len] = '\0';
  725.    
  726.     /* Retrieve the new path. */
  727.     if (!ipc_data_write_receive(&callid, &len)) {
  728.         ipc_answer_0(callid, EINVAL);
  729.         ipc_answer_0(rid, EINVAL);
  730.         free(old);
  731.         return;
  732.     }
  733.     char *new = malloc(len + 1);
  734.     if (!new) {
  735.         ipc_answer_0(callid, ENOMEM);
  736.         ipc_answer_0(rid, ENOMEM);
  737.         free(old);
  738.         return;
  739.     }
  740.     if ((rc = ipc_data_write_finalize(callid, new, len))) {
  741.         ipc_answer_0(rid, rc);
  742.         free(old);
  743.         free(new);
  744.         return;
  745.     }
  746.     new[len] = '\0';
  747.  
  748.     char *oldc = canonify(old, &len);
  749.     char *newc = canonify(new, NULL);
  750.     if (!oldc || !newc) {
  751.         ipc_answer_0(rid, EINVAL);
  752.         free(old);
  753.         free(new);
  754.         return;
  755.     }
  756.     if (!strncmp(newc, oldc, len)) {
  757.         /* oldc is a prefix of newc */
  758.         ipc_answer_0(rid, EINVAL);
  759.         free(old);
  760.         free(new);
  761.         return;
  762.     }
  763.    
  764.     vfs_lookup_res_t old_lr;
  765.     vfs_lookup_res_t new_lr;
  766.     vfs_lookup_res_t new_par_lr;
  767.     rwlock_write_lock(&namespace_rwlock);
  768.     /* Lookup the node belonging to the old file name. */
  769.     rc = vfs_lookup_internal(oldc, L_NONE, &old_lr, NULL);
  770.     if (rc != EOK) {
  771.         rwlock_write_unlock(&namespace_rwlock);
  772.         ipc_answer_0(rid, rc);
  773.         free(old);
  774.         free(new);
  775.         return;
  776.     }
  777.     vfs_node_t *old_node = vfs_node_get(&old_lr);
  778.     if (!old_node) {
  779.         rwlock_write_unlock(&namespace_rwlock);
  780.         ipc_answer_0(rid, ENOMEM);
  781.         free(old);
  782.         free(new);
  783.         return;
  784.     }
  785.     /* Lookup parent of the new file name. */
  786.     rc = vfs_lookup_internal(newc, L_PARENT, &new_par_lr, NULL);
  787.     if (rc != EOK) {
  788.         rwlock_write_unlock(&namespace_rwlock);
  789.         ipc_answer_0(rid, rc);
  790.         free(old);
  791.         free(new);
  792.         return;
  793.     }
  794.     /* Check whether linking to the same file system instance. */
  795.     if ((old_node->fs_handle != new_par_lr.triplet.fs_handle) ||
  796.         (old_node->dev_handle != new_par_lr.triplet.dev_handle)) {
  797.         rwlock_write_unlock(&namespace_rwlock);
  798.         ipc_answer_0(rid, EXDEV);   /* different file systems */
  799.         free(old);
  800.         free(new);
  801.         return;
  802.     }
  803.     /* Destroy the old link for the new name. */
  804.     vfs_node_t *new_node = NULL;
  805.     rc = vfs_lookup_internal(newc, L_UNLINK, &new_lr, NULL);
  806.     switch (rc) {
  807.     case ENOENT:
  808.         /* simply not in our way */
  809.         break;
  810.     case EOK:
  811.         new_node = vfs_node_get(&new_lr);
  812.         if (!new_node) {
  813.             rwlock_write_unlock(&namespace_rwlock);
  814.             ipc_answer_0(rid, ENOMEM);
  815.             free(old);
  816.             free(new);
  817.             return;
  818.         }
  819.         futex_down(&nodes_futex);
  820.         new_node->lnkcnt--;
  821.         futex_up(&nodes_futex);
  822.         break;
  823.     default:
  824.         rwlock_write_unlock(&namespace_rwlock);
  825.         ipc_answer_0(rid, ENOTEMPTY);
  826.         free(old);
  827.         free(new);
  828.         return;
  829.     }
  830.     /* Create the new link for the new name. */
  831.     rc = vfs_lookup_internal(newc, L_LINK, NULL, NULL, old_node->index);
  832.     if (rc != EOK) {
  833.         rwlock_write_unlock(&namespace_rwlock);
  834.         if (new_node)
  835.             vfs_node_put(new_node);
  836.         ipc_answer_0(rid, rc);
  837.         free(old);
  838.         free(new);
  839.         return;
  840.     }
  841.     futex_down(&nodes_futex);
  842.     old_node->lnkcnt++;
  843.     futex_up(&nodes_futex);
  844.     /* Destroy the link for the old name. */
  845.     rc = vfs_lookup_internal(oldc, L_UNLINK, NULL, NULL);
  846.     if (rc != EOK) {
  847.         rwlock_write_unlock(&namespace_rwlock);
  848.         vfs_node_put(old_node);
  849.         if (new_node)
  850.             vfs_node_put(new_node);
  851.         ipc_answer_0(rid, rc);
  852.         free(old);
  853.         free(new);
  854.         return;
  855.     }
  856.     futex_down(&nodes_futex);
  857.     old_node->lnkcnt--;
  858.     futex_up(&nodes_futex);
  859.     rwlock_write_unlock(&namespace_rwlock);
  860.     vfs_node_put(old_node);
  861.     if (new_node)
  862.         vfs_node_put(new_node);
  863.     free(old);
  864.     free(new);
  865.     ipc_answer_0(rid, EOK);
  866. }
  867.  
  868. /**
  869.  * @}
  870.  */
  871.