Subversion Repositories HelenOS

Rev

Rev 4437 | Rev 4445 | 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. /** Pending mount structure. */
  59. typedef struct {
  60.     link_t link;
  61.     char *fs_name;            /**< File system name */
  62.     char *mp;                 /**< Mount point */
  63.     char *opts;       /**< Mount options. */
  64.     ipc_callid_t callid;      /**< Call ID waiting for the mount */
  65.     ipc_callid_t rid;         /**< Request ID */
  66.     dev_handle_t dev_handle;  /**< Device handle */
  67. } pending_req_t;
  68.  
  69. LIST_INITIALIZE(pending_req);
  70.  
  71. /**
  72.  * This rwlock prevents the race between a triplet-to-VFS-node resolution and a
  73.  * concurrent VFS operation which modifies the file system namespace.
  74.  */
  75. RWLOCK_INITIALIZE(namespace_rwlock);
  76.  
  77. vfs_pair_t rootfs = {
  78.     .fs_handle = 0,
  79.     .dev_handle = 0
  80. };
  81.  
  82. static void vfs_mount_internal(ipc_callid_t rid, dev_handle_t dev_handle,
  83.     fs_handle_t fs_handle, char *mp, char *opts)
  84. {
  85.     vfs_lookup_res_t mp_res;
  86.     vfs_lookup_res_t mr_res;
  87.     vfs_node_t *mp_node = NULL;
  88.     vfs_node_t *mr_node;
  89.     fs_index_t rindex;
  90.     size_t rsize;
  91.     unsigned rlnkcnt;
  92.     ipcarg_t rc;
  93.     int phone;
  94.     aid_t msg;
  95.     ipc_call_t answer;
  96.            
  97.     /* Resolve the path to the mountpoint. */
  98.     rwlock_write_lock(&namespace_rwlock);
  99.     if (rootfs.fs_handle) {
  100.         /* We already have the root FS. */
  101.         if (str_cmp(mp, "/") == 0) {
  102.             /* Trying to mount root FS over root FS */
  103.             ipc_answer_0(rid, EBUSY);
  104.             rwlock_write_unlock(&namespace_rwlock);
  105.             return;
  106.         }
  107.        
  108.         rc = vfs_lookup_internal(mp, L_DIRECTORY, &mp_res, NULL);
  109.         if (rc != EOK) {
  110.             /* The lookup failed for some reason. */
  111.             ipc_answer_0(rid, rc);
  112.             rwlock_write_unlock(&namespace_rwlock);
  113.             return;
  114.         }
  115.        
  116.         mp_node = vfs_node_get(&mp_res);
  117.         if (!mp_node) {
  118.             ipc_answer_0(rid, ENOMEM);
  119.             rwlock_write_unlock(&namespace_rwlock);
  120.             return;
  121.         }
  122.        
  123.         /*
  124.          * Now we hold a reference to mp_node.
  125.          * It will be dropped upon the corresponding VFS_UNMOUNT.
  126.          * This prevents the mount point from being deleted.
  127.          */
  128.     } else {
  129.         /* We still don't have the root file system mounted. */
  130.         if (str_cmp(mp, "/") == 0) {
  131.             /*
  132.              * For this simple, but important case,
  133.              * we are almost done.
  134.              */
  135.            
  136.             /* Tell the mountee that it is being mounted. */
  137.             phone = vfs_grab_phone(fs_handle);
  138.             msg = async_send_1(phone, VFS_MOUNTED,
  139.                 (ipcarg_t) dev_handle, &answer);
  140.             /* send the mount options */
  141.             rc = ipc_data_write_start(phone, (void *)opts,
  142.                 str_size(opts));
  143.             if (rc != EOK) {
  144.                 async_wait_for(msg, NULL);
  145.                 vfs_release_phone(phone);
  146.                 ipc_answer_0(rid, rc);
  147.                 rwlock_write_unlock(&namespace_rwlock);
  148.                 return;
  149.             }
  150.             async_wait_for(msg, &rc);
  151.             vfs_release_phone(phone);
  152.            
  153.             if (rc != EOK) {
  154.                 ipc_answer_0(rid, rc);
  155.                 rwlock_write_unlock(&namespace_rwlock);
  156.                 return;
  157.             }
  158.  
  159.             rindex = (fs_index_t) IPC_GET_ARG1(answer);
  160.             rsize = (size_t) IPC_GET_ARG2(answer);
  161.             rlnkcnt = (unsigned) IPC_GET_ARG3(answer);
  162.            
  163.             mr_res.triplet.fs_handle = fs_handle;
  164.             mr_res.triplet.dev_handle = dev_handle;
  165.             mr_res.triplet.index = rindex;
  166.             mr_res.size = rsize;
  167.             mr_res.lnkcnt = rlnkcnt;
  168.             mr_res.type = VFS_NODE_DIRECTORY;
  169.            
  170.             rootfs.fs_handle = fs_handle;
  171.             rootfs.dev_handle = dev_handle;
  172.            
  173.             /* Add reference to the mounted root. */
  174.             mr_node = vfs_node_get(&mr_res);
  175.             assert(mr_node);
  176.            
  177.             ipc_answer_0(rid, rc);
  178.             rwlock_write_unlock(&namespace_rwlock);
  179.             return;
  180.         } else {
  181.             /*
  182.              * We can't resolve this without the root filesystem
  183.              * being mounted first.
  184.              */
  185.             ipc_answer_0(rid, ENOENT);
  186.             rwlock_write_unlock(&namespace_rwlock);
  187.             return;
  188.         }
  189.     }
  190.    
  191.     /*
  192.      * At this point, we have all necessary pieces: file system and device
  193.      * handles, and we know the mount point VFS node.
  194.      */
  195.    
  196.     int mountee_phone = vfs_grab_phone(fs_handle);
  197.     assert(mountee_phone >= 0);
  198.     vfs_release_phone(mountee_phone);
  199.  
  200.     phone = vfs_grab_phone(mp_res.triplet.fs_handle);
  201.     msg = async_send_4(phone, VFS_MOUNT,
  202.         (ipcarg_t) mp_res.triplet.dev_handle,
  203.         (ipcarg_t) mp_res.triplet.index,
  204.         (ipcarg_t) fs_handle,
  205.         (ipcarg_t) dev_handle, &answer);
  206.    
  207.     /* send connection */
  208.     rc = async_req_1_0(phone, IPC_M_CONNECTION_CLONE, mountee_phone);
  209.     if (rc != EOK) {
  210.         async_wait_for(msg, NULL);
  211.         vfs_release_phone(phone);
  212.         /* Mount failed, drop reference to mp_node. */
  213.         if (mp_node)
  214.             vfs_node_put(mp_node);
  215.         ipc_answer_0(rid, rc);
  216.         rwlock_write_unlock(&namespace_rwlock);
  217.         return;
  218.     }
  219.    
  220.     /* send the mount options */
  221.     rc = ipc_data_write_start(phone, (void *)opts, str_size(opts));
  222.     if (rc != EOK) {
  223.         async_wait_for(msg, NULL);
  224.         vfs_release_phone(phone);
  225.         /* Mount failed, drop reference to mp_node. */
  226.         if (mp_node)
  227.             vfs_node_put(mp_node);
  228.         ipc_answer_0(rid, rc);
  229.         rwlock_write_unlock(&namespace_rwlock);
  230.         return;
  231.     }
  232.     async_wait_for(msg, &rc);
  233.     vfs_release_phone(phone);
  234.    
  235.     if (rc == EOK) {
  236.         rindex = (fs_index_t) IPC_GET_ARG1(answer);
  237.         rsize = (size_t) IPC_GET_ARG2(answer);
  238.         rlnkcnt = (unsigned) IPC_GET_ARG3(answer);
  239.    
  240.         mr_res.triplet.fs_handle = fs_handle;
  241.         mr_res.triplet.dev_handle = dev_handle;
  242.         mr_res.triplet.index = rindex;
  243.         mr_res.size = rsize;
  244.         mr_res.lnkcnt = rlnkcnt;
  245.         mr_res.type = VFS_NODE_DIRECTORY;
  246.    
  247.         /* Add reference to the mounted root. */
  248.         mr_node = vfs_node_get(&mr_res);
  249.         assert(mr_node);
  250.     } else {
  251.         /* Mount failed, drop reference to mp_node. */
  252.         if (mp_node)
  253.             vfs_node_put(mp_node);
  254.     }
  255.  
  256.     ipc_answer_0(rid, rc);
  257.     rwlock_write_unlock(&namespace_rwlock);
  258. }
  259.  
  260. /** Process pending mount requests */
  261. void vfs_process_pending_mount(void)
  262. {
  263.     link_t *cur;
  264.    
  265. loop:
  266.     for (cur = pending_req.next; cur != &pending_req; cur = cur->next) {
  267.         pending_req_t *pr = list_get_instance(cur, pending_req_t, link);
  268.        
  269.         fs_handle_t fs_handle = fs_name_to_handle(pr->fs_name, true);
  270.         if (!fs_handle)
  271.             continue;
  272.        
  273.         /* Acknowledge that we know fs_name. */
  274.         ipc_answer_0(pr->callid, EOK);
  275.        
  276.         /* Do the mount */
  277.         vfs_mount_internal(pr->rid, pr->dev_handle, fs_handle, pr->mp,
  278.             pr->opts);
  279.        
  280.         free(pr->fs_name);
  281.         free(pr->mp);
  282.         free(pr->opts);
  283.         list_remove(cur);
  284.         free(pr);
  285.         goto loop;
  286.     }
  287. }
  288.  
  289. void vfs_mount(ipc_callid_t rid, ipc_call_t *request)
  290. {
  291.     /*
  292.      * We expect the library to do the device-name to device-handle
  293.      * translation for us, thus the device handle will arrive as ARG1
  294.      * in the request.
  295.      */
  296.     dev_handle_t dev_handle = (dev_handle_t) IPC_GET_ARG1(*request);
  297.    
  298.     /*
  299.      * Mount flags are passed as ARG2.
  300.      */
  301.     unsigned int flags = (unsigned int) IPC_GET_ARG2(*request);
  302.    
  303.     /*
  304.      * For now, don't make use of ARG3, but it can be used to
  305.      * carry mount options in the future.
  306.      */
  307.    
  308.     /* We want the client to send us the mount point. */
  309.     ipc_callid_t callid;
  310.     size_t size;
  311.     if (!ipc_data_write_receive(&callid, &size)) {
  312.         ipc_answer_0(callid, EINVAL);
  313.         ipc_answer_0(rid, EINVAL);
  314.         return;
  315.     }
  316.    
  317.     /* Check whether size is reasonable wrt. the mount point. */
  318.     if ((size < 1) || (size > MAX_PATH_LEN)) {
  319.         ipc_answer_0(callid, EINVAL);
  320.         ipc_answer_0(rid, EINVAL);
  321.         return;
  322.     }
  323.    
  324.     /* Allocate buffer for the mount point data being received. */
  325.     char *mp = malloc(size + 1);
  326.     if (!mp) {
  327.         ipc_answer_0(callid, ENOMEM);
  328.         ipc_answer_0(rid, ENOMEM);
  329.         return;
  330.     }
  331.    
  332.     /* Deliver the mount point. */
  333.     ipcarg_t retval = ipc_data_write_finalize(callid, mp, size);
  334.     if (retval != EOK) {
  335.         ipc_answer_0(rid, retval);
  336.         free(mp);
  337.         return;
  338.     }
  339.     mp[size] = '\0';
  340.    
  341.     /* Now we expect to receive the mount options. */
  342.     if (!ipc_data_write_receive(&callid, &size)) {
  343.         ipc_answer_0(callid, EINVAL);
  344.         ipc_answer_0(rid, EINVAL);
  345.         free(mp);
  346.         return;
  347.     }
  348.  
  349.     /* Check the offered options size. */
  350.     if (size < 0 || size > MAX_MNTOPTS_LEN) {
  351.         ipc_answer_0(callid, EINVAL);
  352.         ipc_answer_0(rid, EINVAL);
  353.         free(mp);
  354.         return;
  355.     }
  356.  
  357.     /* Allocate buffer for the mount options. */
  358.     char *opts = (char *) malloc(size + 1);
  359.     if (!opts) {
  360.         ipc_answer_0(callid, ENOMEM);
  361.         ipc_answer_0(rid, ENOMEM);
  362.         free(mp);
  363.         return;
  364.     }
  365.  
  366.     /* Deliver the mount options. */
  367.     retval = ipc_data_write_finalize(callid, opts, size);
  368.     if (retval != EOK) {
  369.         ipc_answer_0(rid, retval);
  370.         free(mp);
  371.         free(opts);
  372.         return;
  373.     }
  374.     opts[size] = '\0';
  375.    
  376.     /*
  377.      * Now, we expect the client to send us data with the name of the file
  378.      * system.
  379.      */
  380.     if (!ipc_data_write_receive(&callid, &size)) {
  381.         ipc_answer_0(callid, EINVAL);
  382.         ipc_answer_0(rid, EINVAL);
  383.         free(mp);
  384.         free(opts);
  385.         return;
  386.     }
  387.    
  388.     /*
  389.      * Don't receive more than is necessary for storing a full file system
  390.      * name.
  391.      */
  392.     if ((size < 1) || (size > FS_NAME_MAXLEN)) {
  393.         ipc_answer_0(callid, EINVAL);
  394.         ipc_answer_0(rid, EINVAL);
  395.         free(mp);
  396.         free(opts);
  397.         return;
  398.     }
  399.    
  400.     /*
  401.      * Allocate buffer for file system name.
  402.      */
  403.     char *fs_name = (char *) malloc(size + 1);
  404.     if (fs_name == NULL) {
  405.         ipc_answer_0(callid, ENOMEM);
  406.         ipc_answer_0(rid, ENOMEM);
  407.         free(mp);
  408.         free(opts);
  409.         return;
  410.     }
  411.    
  412.     /* Deliver the file system name. */
  413.     retval = ipc_data_write_finalize(callid, fs_name, size);
  414.     if (retval != EOK) {
  415.         ipc_answer_0(rid, retval);
  416.         free(mp);
  417.         free(opts);
  418.         free(fs_name);
  419.         return;
  420.     }
  421.     fs_name[size] = '\0';
  422.  
  423.     /*
  424.      * Wait for IPC_M_PING so that we can return an error if we don't know
  425.      * fs_name.
  426.      */
  427.     ipc_call_t data;
  428.     callid = async_get_call(&data);
  429.     if (IPC_GET_METHOD(data) != IPC_M_PING) {
  430.         ipc_answer_0(callid, ENOTSUP);
  431.         ipc_answer_0(rid, ENOTSUP);
  432.         free(mp);
  433.         free(opts);
  434.         free(fs_name);
  435.         return;
  436.     }
  437.  
  438.     /*
  439.      * Check if we know a file system with the same name as is in fs_name.
  440.      * This will also give us its file system handle.
  441.      */
  442.     fs_handle_t fs_handle = fs_name_to_handle(fs_name, true);
  443.     if (!fs_handle) {
  444.         if (flags & IPC_FLAG_BLOCKING) {
  445.             pending_req_t *pr;
  446.  
  447.             /* Blocking mount, add to pending list */
  448.             pr = (pending_req_t *) malloc(sizeof(pending_req_t));
  449.             if (!pr) {
  450.                 ipc_answer_0(callid, ENOMEM);
  451.                 ipc_answer_0(rid, ENOMEM);
  452.                 free(mp);
  453.                 free(fs_name);
  454.                 free(opts);
  455.                 return;
  456.             }
  457.            
  458.             pr->fs_name = fs_name;
  459.             pr->mp = mp;
  460.             pr->opts = opts;
  461.             pr->callid = callid;
  462.             pr->rid = rid;
  463.             pr->dev_handle = dev_handle;
  464.             link_initialize(&pr->link);
  465.             list_append(&pr->link, &pending_req);
  466.             return;
  467.         }
  468.        
  469.         ipc_answer_0(callid, ENOENT);
  470.         ipc_answer_0(rid, ENOENT);
  471.         free(mp);
  472.         free(fs_name);
  473.         free(opts);
  474.         return;
  475.     }
  476.    
  477.     /* Acknowledge that we know fs_name. */
  478.     ipc_answer_0(callid, EOK);
  479.    
  480.     /* Do the mount */
  481.     vfs_mount_internal(rid, dev_handle, fs_handle, mp, opts);
  482.     free(mp);
  483.     free(fs_name);
  484.     free(opts);
  485. }
  486.  
  487. void vfs_open(ipc_callid_t rid, ipc_call_t *request)
  488. {
  489.     async_serialize_start();
  490.    
  491.     if (!vfs_files_init()) {
  492.         ipc_answer_0(rid, ENOMEM);
  493.         async_serialize_end();
  494.         return;
  495.     }
  496.  
  497.     /*
  498.      * The POSIX interface is open(path, oflag, mode).
  499.      * We can receive oflags and mode along with the VFS_OPEN call; the path
  500.      * will need to arrive in another call.
  501.      *
  502.      * We also receive one private, non-POSIX set of flags called lflag
  503.      * used to pass information to vfs_lookup_internal().
  504.      */
  505.     int lflag = IPC_GET_ARG1(*request);
  506.     int oflag = IPC_GET_ARG2(*request);
  507.     int mode = IPC_GET_ARG3(*request);
  508.     size_t len;
  509.  
  510.     /*
  511.      * Make sure that we are called with exactly one of L_FILE and
  512.      * L_DIRECTORY.
  513.      */
  514.     if ((lflag & (L_FILE | L_DIRECTORY)) == 0 ||
  515.         (lflag & (L_FILE | L_DIRECTORY)) == (L_FILE | L_DIRECTORY)) {
  516.         ipc_answer_0(rid, EINVAL);
  517.         async_serialize_end();
  518.         return;
  519.     }
  520.  
  521.     if (oflag & O_CREAT)
  522.         lflag |= L_CREATE;
  523.     if (oflag & O_EXCL)
  524.         lflag |= L_EXCLUSIVE;
  525.  
  526.     ipc_callid_t callid;
  527.  
  528.     if (!ipc_data_write_receive(&callid, &len)) {
  529.         ipc_answer_0(callid, EINVAL);
  530.         ipc_answer_0(rid, EINVAL);
  531.         async_serialize_end();
  532.         return;
  533.     }
  534.     char *path = malloc(len + 1);
  535.     if (!path) {
  536.         ipc_answer_0(callid, ENOMEM);
  537.         ipc_answer_0(rid, ENOMEM);
  538.         async_serialize_end();
  539.         return;
  540.     }
  541.     int rc;
  542.     if ((rc = ipc_data_write_finalize(callid, path, len))) {
  543.         ipc_answer_0(rid, rc);
  544.         free(path);
  545.         async_serialize_end();
  546.         return;
  547.     }
  548.     path[len] = '\0';
  549.    
  550.     /*
  551.      * Avoid the race condition in which the file can be deleted before we
  552.      * find/create-and-lock the VFS node corresponding to the looked-up
  553.      * triplet.
  554.      */
  555.     if (lflag & L_CREATE)
  556.         rwlock_write_lock(&namespace_rwlock);
  557.     else
  558.         rwlock_read_lock(&namespace_rwlock);
  559.  
  560.     /* The path is now populated and we can call vfs_lookup_internal(). */
  561.     vfs_lookup_res_t lr;
  562.     rc = vfs_lookup_internal(path, lflag, &lr, NULL);
  563.     if (rc) {
  564.         if (lflag & L_CREATE)
  565.             rwlock_write_unlock(&namespace_rwlock);
  566.         else
  567.             rwlock_read_unlock(&namespace_rwlock);
  568.         ipc_answer_0(rid, rc);
  569.         free(path);
  570.         async_serialize_end();
  571.         return;
  572.     }
  573.  
  574.     /* Path is no longer needed. */
  575.     free(path);
  576.  
  577.     vfs_node_t *node = vfs_node_get(&lr);
  578.     if (lflag & L_CREATE)
  579.         rwlock_write_unlock(&namespace_rwlock);
  580.     else
  581.         rwlock_read_unlock(&namespace_rwlock);
  582.  
  583.     /* Truncate the file if requested and if necessary. */
  584.     if (oflag & O_TRUNC) {
  585.         rwlock_write_lock(&node->contents_rwlock);
  586.         if (node->size) {
  587.             rc = vfs_truncate_internal(node->fs_handle,
  588.                 node->dev_handle, node->index, 0);
  589.             if (rc) {
  590.                 rwlock_write_unlock(&node->contents_rwlock);
  591.                 vfs_node_put(node);
  592.                 ipc_answer_0(rid, rc);
  593.                 async_serialize_end();
  594.                 return;
  595.             }
  596.             node->size = 0;
  597.         }
  598.         rwlock_write_unlock(&node->contents_rwlock);
  599.     }
  600.  
  601.     /*
  602.      * Get ourselves a file descriptor and the corresponding vfs_file_t
  603.      * structure.
  604.      */
  605.     int fd = vfs_fd_alloc();
  606.     if (fd < 0) {
  607.         vfs_node_put(node);
  608.         ipc_answer_0(rid, fd);
  609.         async_serialize_end();
  610.         return;
  611.     }
  612.     vfs_file_t *file = vfs_file_get(fd);
  613.     file->node = node;
  614.     if (oflag & O_APPEND)
  615.         file->append = true;
  616.  
  617.     /*
  618.      * The following increase in reference count is for the fact that the
  619.      * file is being opened and that a file structure is pointing to it.
  620.      * It is necessary so that the file will not disappear when
  621.      * vfs_node_put() is called. The reference will be dropped by the
  622.      * respective VFS_CLOSE.
  623.      */
  624.     vfs_node_addref(node);
  625.     vfs_node_put(node);
  626.  
  627.     /* Success! Return the new file descriptor to the client. */
  628.     ipc_answer_1(rid, EOK, fd);
  629.     async_serialize_end();
  630. }
  631.  
  632. void vfs_close(ipc_callid_t rid, ipc_call_t *request)
  633. {
  634.     int fd = IPC_GET_ARG1(*request);
  635.     int rc = vfs_fd_free(fd);
  636.     ipc_answer_0(rid, rc);
  637. }
  638.  
  639. static void vfs_rdwr(ipc_callid_t rid, ipc_call_t *request, bool read)
  640. {
  641.  
  642.     /*
  643.      * The following code strongly depends on the fact that the files data
  644.      * structure can be only accessed by a single fibril and all file
  645.      * operations are serialized (i.e. the reads and writes cannot
  646.      * interleave and a file cannot be closed while it is being read).
  647.      *
  648.      * Additional synchronization needs to be added once the table of
  649.      * open files supports parallel access!
  650.      */
  651.  
  652.     int fd = IPC_GET_ARG1(*request);
  653.    
  654.     /* Lookup the file structure corresponding to the file descriptor. */
  655.     vfs_file_t *file = vfs_file_get(fd);
  656.     if (!file) {
  657.         ipc_answer_0(rid, ENOENT);
  658.         return;
  659.     }
  660.    
  661.     /*
  662.      * Now we need to receive a call with client's
  663.      * IPC_M_DATA_READ/IPC_M_DATA_WRITE request.
  664.      */
  665.     ipc_callid_t callid;
  666.     int res;
  667.     if (read)
  668.         res = ipc_data_read_receive(&callid, NULL);
  669.     else
  670.         res = ipc_data_write_receive(&callid, NULL);
  671.     if (!res) {
  672.         ipc_answer_0(callid, EINVAL);
  673.         ipc_answer_0(rid, EINVAL);
  674.         return;
  675.     }
  676.    
  677.     /*
  678.      * Lock the open file structure so that no other thread can manipulate
  679.      * the same open file at a time.
  680.      */
  681.     futex_down(&file->lock);
  682.  
  683.     /*
  684.      * Lock the file's node so that no other client can read/write to it at
  685.      * the same time.
  686.      */
  687.     if (read)
  688.         rwlock_read_lock(&file->node->contents_rwlock);
  689.     else
  690.         rwlock_write_lock(&file->node->contents_rwlock);
  691.  
  692.     if (file->node->type == VFS_NODE_DIRECTORY) {
  693.         /*
  694.          * Make sure that no one is modifying the namespace
  695.          * while we are in readdir().
  696.          */
  697.         assert(read);
  698.         rwlock_read_lock(&namespace_rwlock);
  699.     }
  700.    
  701.     int fs_phone = vfs_grab_phone(file->node->fs_handle);  
  702.    
  703.     /* Make a VFS_READ/VFS_WRITE request at the destination FS server. */
  704.     aid_t msg;
  705.     ipc_call_t answer;
  706.     if (!read && file->append)
  707.         file->pos = file->node->size;
  708.     msg = async_send_3(fs_phone, IPC_GET_METHOD(*request),
  709.         file->node->dev_handle, file->node->index, file->pos, &answer);
  710.    
  711.     /*
  712.      * Forward the IPC_M_DATA_READ/IPC_M_DATA_WRITE request to the
  713.      * destination FS server. The call will be routed as if sent by
  714.      * ourselves. Note that call arguments are immutable in this case so we
  715.      * don't have to bother.
  716.      */
  717.     ipc_forward_fast(callid, fs_phone, 0, 0, 0, IPC_FF_ROUTE_FROM_ME);
  718.    
  719.     vfs_release_phone(fs_phone);
  720.    
  721.     /* Wait for reply from the FS server. */
  722.     ipcarg_t rc;
  723.     async_wait_for(msg, &rc);
  724.     size_t bytes = IPC_GET_ARG1(answer);
  725.  
  726.     if (file->node->type == VFS_NODE_DIRECTORY)
  727.         rwlock_read_unlock(&namespace_rwlock);
  728.    
  729.     /* Unlock the VFS node. */
  730.     if (read)
  731.         rwlock_read_unlock(&file->node->contents_rwlock);
  732.     else {
  733.         /* Update the cached version of node's size. */
  734.         if (rc == EOK)
  735.             file->node->size = IPC_GET_ARG2(answer);
  736.         rwlock_write_unlock(&file->node->contents_rwlock);
  737.     }
  738.    
  739.     /* Update the position pointer and unlock the open file. */
  740.     if (rc == EOK)
  741.         file->pos += bytes;
  742.     futex_up(&file->lock);
  743.    
  744.     /*
  745.      * FS server's reply is the final result of the whole operation we
  746.      * return to the client.
  747.      */
  748.     ipc_answer_1(rid, rc, bytes);
  749. }
  750.  
  751. void vfs_read(ipc_callid_t rid, ipc_call_t *request)
  752. {
  753.     vfs_rdwr(rid, request, true);
  754. }
  755.  
  756. void vfs_write(ipc_callid_t rid, ipc_call_t *request)
  757. {
  758.     vfs_rdwr(rid, request, false);
  759. }
  760.  
  761. void vfs_seek(ipc_callid_t rid, ipc_call_t *request)
  762. {
  763.     int fd = (int) IPC_GET_ARG1(*request);
  764.     off_t off = (off_t) IPC_GET_ARG2(*request);
  765.     int whence = (int) IPC_GET_ARG3(*request);
  766.  
  767.  
  768.     /* Lookup the file structure corresponding to the file descriptor. */
  769.     vfs_file_t *file = vfs_file_get(fd);
  770.     if (!file) {
  771.         ipc_answer_0(rid, ENOENT);
  772.         return;
  773.     }
  774.  
  775.     off_t newpos;
  776.     futex_down(&file->lock);
  777.     if (whence == SEEK_SET) {
  778.         file->pos = off;
  779.         futex_up(&file->lock);
  780.         ipc_answer_1(rid, EOK, off);
  781.         return;
  782.     }
  783.     if (whence == SEEK_CUR) {
  784.         if (file->pos + off < file->pos) {
  785.             futex_up(&file->lock);
  786.             ipc_answer_0(rid, EOVERFLOW);
  787.             return;
  788.         }
  789.         file->pos += off;
  790.         newpos = file->pos;
  791.         futex_up(&file->lock);
  792.         ipc_answer_1(rid, EOK, newpos);
  793.         return;
  794.     }
  795.     if (whence == SEEK_END) {
  796.         rwlock_read_lock(&file->node->contents_rwlock);
  797.         size_t size = file->node->size;
  798.         rwlock_read_unlock(&file->node->contents_rwlock);
  799.         if (size + off < size) {
  800.             futex_up(&file->lock);
  801.             ipc_answer_0(rid, EOVERFLOW);
  802.             return;
  803.         }
  804.         newpos = size + off;
  805.         futex_up(&file->lock);
  806.         ipc_answer_1(rid, EOK, newpos);
  807.         return;
  808.     }
  809.     futex_up(&file->lock);
  810.     ipc_answer_0(rid, EINVAL);
  811. }
  812.  
  813. int
  814. vfs_truncate_internal(fs_handle_t fs_handle, dev_handle_t dev_handle,
  815.     fs_index_t index, size_t size)
  816. {
  817.     ipcarg_t rc;
  818.     int fs_phone;
  819.    
  820.     fs_phone = vfs_grab_phone(fs_handle);
  821.     rc = async_req_3_0(fs_phone, VFS_TRUNCATE, (ipcarg_t)dev_handle,
  822.         (ipcarg_t)index, (ipcarg_t)size);
  823.     vfs_release_phone(fs_phone);
  824.     return (int)rc;
  825. }
  826.  
  827. void vfs_truncate(ipc_callid_t rid, ipc_call_t *request)
  828. {
  829.     int fd = IPC_GET_ARG1(*request);
  830.     size_t size = IPC_GET_ARG2(*request);
  831.     int rc;
  832.  
  833.     vfs_file_t *file = vfs_file_get(fd);
  834.     if (!file) {
  835.         ipc_answer_0(rid, ENOENT);
  836.         return;
  837.     }
  838.     futex_down(&file->lock);
  839.  
  840.     rwlock_write_lock(&file->node->contents_rwlock);
  841.     rc = vfs_truncate_internal(file->node->fs_handle,
  842.         file->node->dev_handle, file->node->index, size);
  843.     if (rc == EOK)
  844.         file->node->size = size;
  845.     rwlock_write_unlock(&file->node->contents_rwlock);
  846.  
  847.     futex_up(&file->lock);
  848.     ipc_answer_0(rid, (ipcarg_t)rc);
  849. }
  850.  
  851. void vfs_mkdir(ipc_callid_t rid, ipc_call_t *request)
  852. {
  853.     int mode = IPC_GET_ARG1(*request);
  854.  
  855.     size_t len;
  856.     ipc_callid_t callid;
  857.  
  858.     if (!ipc_data_write_receive(&callid, &len)) {
  859.         ipc_answer_0(callid, EINVAL);
  860.         ipc_answer_0(rid, EINVAL);
  861.         return;
  862.     }
  863.     char *path = malloc(len + 1);
  864.     if (!path) {
  865.         ipc_answer_0(callid, ENOMEM);
  866.         ipc_answer_0(rid, ENOMEM);
  867.         return;
  868.     }
  869.     int rc;
  870.     if ((rc = ipc_data_write_finalize(callid, path, len))) {
  871.         ipc_answer_0(rid, rc);
  872.         free(path);
  873.         return;
  874.     }
  875.     path[len] = '\0';
  876.    
  877.     rwlock_write_lock(&namespace_rwlock);
  878.     int lflag = L_DIRECTORY | L_CREATE | L_EXCLUSIVE;
  879.     rc = vfs_lookup_internal(path, lflag, NULL, NULL);
  880.     rwlock_write_unlock(&namespace_rwlock);
  881.     free(path);
  882.     ipc_answer_0(rid, rc);
  883. }
  884.  
  885. void vfs_unlink(ipc_callid_t rid, ipc_call_t *request)
  886. {
  887.     int lflag = IPC_GET_ARG1(*request);
  888.  
  889.     size_t len;
  890.     ipc_callid_t callid;
  891.  
  892.     if (!ipc_data_write_receive(&callid, &len)) {
  893.         ipc_answer_0(callid, EINVAL);
  894.         ipc_answer_0(rid, EINVAL);
  895.         return;
  896.     }
  897.     char *path = malloc(len + 1);
  898.     if (!path) {
  899.         ipc_answer_0(callid, ENOMEM);
  900.         ipc_answer_0(rid, ENOMEM);
  901.         return;
  902.     }
  903.     int rc;
  904.     if ((rc = ipc_data_write_finalize(callid, path, len))) {
  905.         ipc_answer_0(rid, rc);
  906.         free(path);
  907.         return;
  908.     }
  909.     path[len] = '\0';
  910.    
  911.     rwlock_write_lock(&namespace_rwlock);
  912.     lflag &= L_DIRECTORY;   /* sanitize lflag */
  913.     vfs_lookup_res_t lr;
  914.     rc = vfs_lookup_internal(path, lflag | L_UNLINK, &lr, NULL);
  915.     free(path);
  916.     if (rc != EOK) {
  917.         rwlock_write_unlock(&namespace_rwlock);
  918.         ipc_answer_0(rid, rc);
  919.         return;
  920.     }
  921.  
  922.     /*
  923.      * The name has already been unlinked by vfs_lookup_internal().
  924.      * We have to get and put the VFS node to ensure that it is
  925.      * VFS_DESTROY'ed after the last reference to it is dropped.
  926.      */
  927.     vfs_node_t *node = vfs_node_get(&lr);
  928.     futex_down(&nodes_futex);
  929.     node->lnkcnt--;
  930.     futex_up(&nodes_futex);
  931.     rwlock_write_unlock(&namespace_rwlock);
  932.     vfs_node_put(node);
  933.     ipc_answer_0(rid, EOK);
  934. }
  935.  
  936. void vfs_rename(ipc_callid_t rid, ipc_call_t *request)
  937. {
  938.     size_t olen, nlen;
  939.     ipc_callid_t callid;
  940.     int rc;
  941.  
  942.     /* Retrieve the old path. */
  943.     if (!ipc_data_write_receive(&callid, &olen)) {
  944.         ipc_answer_0(callid, EINVAL);
  945.         ipc_answer_0(rid, EINVAL);
  946.         return;
  947.     }
  948.     char *old = malloc(olen + 1);
  949.     if (!old) {
  950.         ipc_answer_0(callid, ENOMEM);
  951.         ipc_answer_0(rid, ENOMEM);
  952.         return;
  953.     }
  954.     if ((rc = ipc_data_write_finalize(callid, old, olen))) {
  955.         ipc_answer_0(rid, rc);
  956.         free(old);
  957.         return;
  958.     }
  959.     old[olen] = '\0';
  960.    
  961.     /* Retrieve the new path. */
  962.     if (!ipc_data_write_receive(&callid, &nlen)) {
  963.         ipc_answer_0(callid, EINVAL);
  964.         ipc_answer_0(rid, EINVAL);
  965.         free(old);
  966.         return;
  967.     }
  968.     char *new = malloc(nlen + 1);
  969.     if (!new) {
  970.         ipc_answer_0(callid, ENOMEM);
  971.         ipc_answer_0(rid, ENOMEM);
  972.         free(old);
  973.         return;
  974.     }
  975.     if ((rc = ipc_data_write_finalize(callid, new, nlen))) {
  976.         ipc_answer_0(rid, rc);
  977.         free(old);
  978.         free(new);
  979.         return;
  980.     }
  981.     new[nlen] = '\0';
  982.  
  983.     char *oldc = canonify(old, &olen);
  984.     char *newc = canonify(new, &nlen);
  985.     if (!oldc || !newc) {
  986.         ipc_answer_0(rid, EINVAL);
  987.         free(old);
  988.         free(new);
  989.         return;
  990.     }
  991.     oldc[olen] = '\0';
  992.     newc[nlen] = '\0';
  993.     if ((!str_lcmp(newc, oldc, str_length(oldc))) &&
  994.         ((newc[str_length(oldc)] == '/') ||
  995.         (str_length(oldc) == 1) ||
  996.         (str_length(oldc) == str_length(newc)))) {
  997.             /*
  998.          * oldc is a prefix of newc and either
  999.          * - newc continues with a / where oldc ends, or
  1000.          * - oldc was / itself, or
  1001.          * - oldc and newc are equal.
  1002.          */
  1003.         ipc_answer_0(rid, EINVAL);
  1004.         free(old);
  1005.         free(new);
  1006.         return;
  1007.     }
  1008.    
  1009.     vfs_lookup_res_t old_lr;
  1010.     vfs_lookup_res_t new_lr;
  1011.     vfs_lookup_res_t new_par_lr;
  1012.     rwlock_write_lock(&namespace_rwlock);
  1013.     /* Lookup the node belonging to the old file name. */
  1014.     rc = vfs_lookup_internal(oldc, L_NONE, &old_lr, NULL);
  1015.     if (rc != EOK) {
  1016.         rwlock_write_unlock(&namespace_rwlock);
  1017.         ipc_answer_0(rid, rc);
  1018.         free(old);
  1019.         free(new);
  1020.         return;
  1021.     }
  1022.     vfs_node_t *old_node = vfs_node_get(&old_lr);
  1023.     if (!old_node) {
  1024.         rwlock_write_unlock(&namespace_rwlock);
  1025.         ipc_answer_0(rid, ENOMEM);
  1026.         free(old);
  1027.         free(new);
  1028.         return;
  1029.     }
  1030.     /* Determine the path to the parent of the node with the new name. */
  1031.     char *parentc = str_dup(newc);
  1032.     if (!parentc) {
  1033.         rwlock_write_unlock(&namespace_rwlock);
  1034.         ipc_answer_0(rid, rc);
  1035.         free(old);
  1036.         free(new);
  1037.         return;
  1038.     }
  1039.     char *lastsl = str_rchr(parentc + 1, '/');
  1040.     if (lastsl)
  1041.         *lastsl = '\0';
  1042.     else
  1043.         parentc[1] = '\0';
  1044.     /* Lookup parent of the new file name. */
  1045.     rc = vfs_lookup_internal(parentc, L_NONE, &new_par_lr, NULL);
  1046.     free(parentc);  /* not needed anymore */
  1047.     if (rc != EOK) {
  1048.         rwlock_write_unlock(&namespace_rwlock);
  1049.         ipc_answer_0(rid, rc);
  1050.         free(old);
  1051.         free(new);
  1052.         return;
  1053.     }
  1054.     /* Check whether linking to the same file system instance. */
  1055.     if ((old_node->fs_handle != new_par_lr.triplet.fs_handle) ||
  1056.         (old_node->dev_handle != new_par_lr.triplet.dev_handle)) {
  1057.         rwlock_write_unlock(&namespace_rwlock);
  1058.         ipc_answer_0(rid, EXDEV);   /* different file systems */
  1059.         free(old);
  1060.         free(new);
  1061.         return;
  1062.     }
  1063.     /* Destroy the old link for the new name. */
  1064.     vfs_node_t *new_node = NULL;
  1065.     rc = vfs_lookup_internal(newc, L_UNLINK, &new_lr, NULL);
  1066.     switch (rc) {
  1067.     case ENOENT:
  1068.         /* simply not in our way */
  1069.         break;
  1070.     case EOK:
  1071.         new_node = vfs_node_get(&new_lr);
  1072.         if (!new_node) {
  1073.             rwlock_write_unlock(&namespace_rwlock);
  1074.             ipc_answer_0(rid, ENOMEM);
  1075.             free(old);
  1076.             free(new);
  1077.             return;
  1078.         }
  1079.         futex_down(&nodes_futex);
  1080.         new_node->lnkcnt--;
  1081.         futex_up(&nodes_futex);
  1082.         break;
  1083.     default:
  1084.         rwlock_write_unlock(&namespace_rwlock);
  1085.         ipc_answer_0(rid, ENOTEMPTY);
  1086.         free(old);
  1087.         free(new);
  1088.         return;
  1089.     }
  1090.     /* Create the new link for the new name. */
  1091.     rc = vfs_lookup_internal(newc, L_LINK, NULL, NULL, old_node->index);
  1092.     if (rc != EOK) {
  1093.         rwlock_write_unlock(&namespace_rwlock);
  1094.         if (new_node)
  1095.             vfs_node_put(new_node);
  1096.         ipc_answer_0(rid, rc);
  1097.         free(old);
  1098.         free(new);
  1099.         return;
  1100.     }
  1101.     futex_down(&nodes_futex);
  1102.     old_node->lnkcnt++;
  1103.     futex_up(&nodes_futex);
  1104.     /* Destroy the link for the old name. */
  1105.     rc = vfs_lookup_internal(oldc, L_UNLINK, NULL, NULL);
  1106.     if (rc != EOK) {
  1107.         rwlock_write_unlock(&namespace_rwlock);
  1108.         vfs_node_put(old_node);
  1109.         if (new_node)
  1110.             vfs_node_put(new_node);
  1111.         ipc_answer_0(rid, rc);
  1112.         free(old);
  1113.         free(new);
  1114.         return;
  1115.     }
  1116.     futex_down(&nodes_futex);
  1117.     old_node->lnkcnt--;
  1118.     futex_up(&nodes_futex);
  1119.     rwlock_write_unlock(&namespace_rwlock);
  1120.     vfs_node_put(old_node);
  1121.     if (new_node)
  1122.         vfs_node_put(new_node);
  1123.     free(old);
  1124.     free(new);
  1125.     ipc_answer_0(rid, EOK);
  1126. }
  1127.  
  1128. /**
  1129.  * @}
  1130.  */
  1131.