[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
netfs part of a console server with server-client model
|
From: |
Marcus Brinkmann |
|
Subject: |
netfs part of a console server with server-client model |
|
Date: |
Sun, 2 Jun 2002 05:13:57 +0200 |
|
User-agent: |
Mutt/1.3.28i |
Hi,
as discussed before, it is desirable to have a server-client model for the
console:
VISUAL DISPLAY I \ /
VISUAL DISPLAY II | | TERM ON VIRTUAl CONSOLE I
... > CONSOLE SERVER < TERM ON VIRTUAL CONSOLE II
INPUT DEVICE I | | ...
INPUT DEVICE II | |
... / \
The console server provides an abstract hardware console in the local
encoding to the term server on the one hand, and on the other hand it
provides two interfaces for user interaction:
1. The display interface, which is basically a map'able file which contains
the text buffer and attribute (color, blinking, etc) matrix with the
possibility for the user to receive asynchronous update information on
the content. This also includes bells and gimmicks like this.
2. The input interface, which allows a kbd driver, mouse gestures, voice
recognition software etc to input keyboard and similar events into the
console server.
This side of the interface will be UCS-4/UTF-8 based only. It is expected
that usually, both parts are combined into a single program. This program
also handles virtual console switching etc. See past discussion for details
on what features this design makes possible.
Please see also past discussion about what difficulties are in the design of
the asynchronous update interface :) I gave a huge list, and did not make
progress on that part yet. I guess experimentation will help a lot.
(BTW, the Linux braille terminal driver polls the screen every so and so
often, because there are no file change notifications, we should be able to
do better than that).
What I attach to this mail is a file that incorporates some the virtual
console abstraction in console/console.c in the Hurd source tree, and new
parts which provide a netfs based filesystem that looks like this:
console:
drwxrwx--- 2 root root 0 2. Jun 04:27 1/
drwxrwx--- 2 root root 0 2. Jun 04:27 3/
drwxrwx--- 2 root root 0 2. Jun 04:27 4/
1:
crw-rw---- 1 root root 0, 0 2. Jun 04:27 console
-rw-rw---- 1 root root 2000 2. Jun 04:27 display
3:
crw-rw---- 1 root root 0, 0 2. Jun 04:27 console
-rw-rw---- 1 root root 2000 2. Jun 04:27 display
4:
crw-rw---- 1 root root 0, 0 2. Jun 04:27 console
-rw-rw---- 1 root root 2000 2. Jun 04:27 display
The idea is that here are three virtual consoles provided by the server,
with the IDs 1, 3 and 4. The "console" node in each directory provided by
the server is for term, and the "display" node is for the display/input
driver (better names are welcome). 2000 is 80x25 and the default size of a
screen, this is just an example.
I am inclined to give all virtual consoles for one console the same size and
local encoding, but this is not terribly important. Nothing of this is
set in stone, in particular the interface between the console server and the
display driver.
However, I think it might be a good idea to start with a simplicistic
interface based on existing RPCs, and work from there. I think this is
better than putting a lot of work into getting the existing code base with
the old, monolithic design to make it functional.
I have found a good description of xkb, btw, and will go for that. It just
needs some time, and it might be better to write a clean console server, and
a not-so-clean display and input driver so oskit-mach becomes usable, and
then go from there.
Anyway, the point of this mail is to put the netfs code into public. This
is the first netfs based filesystem I wrote, and I took some of the code
from hostmux (found a bug in there, too). The internals are like this:
Virtual consoles are created on the fly when looking up nodes in root. This
is unproblematic, as only available consoles are listed in
netfs_get_dirents, and usually programs will work with that. Probably some
parts of the virtual console should be allocated lazily, so you don't hit
the full cost just to look up a node you don't want.
A virtual console structure is also the place where created nodes are looked
up. This safes the inode hash table. For this to work, each node has to
keep a reference to the virtual console, and the virtual console have to
back reference to the node (but without acquiring a reference to it).
netfs_node_norefs removes the back reference before releasing the ref to the
virtual console. Locking is a bit peculiar, but not really difficult, as
the virtual console structure itself is fine-graded. All this is exactly
how hostmux stores and references the node and hostname structures.
The code is tested and works as far as it is there. It has one serious
limitation, and that is the lookup of .., which currently assumes a flat
hierarchy (only one level of subdirectories). It doesn't integrate all
too much with the virtual console code, either, some comments show where
this can happen. As I said, the point here is to get the filesystem part
going, rather than the text buffer and input queue details (which I have
already written in the console/* code and just need to reuse and put some
nice abstractions in there).
Any comments welcome, of course!
Thanks,
Marcus
/* console -- a console server
Copyright (C) 1997, 1999, 2002 Free Software Foundation, Inc.
Written by Miles Bader and Marcus Brinkmann.
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2, or (at
your option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
#include <hurd/netfs.h>
#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <unistd.h>
#include <argp.h>
#include <error.h>
#include <string.h>
#include <fcntl.h>
#include <dirent.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <cthreads.h>
#include <mcheck.h>
#include <version.h>
#include "console.h"
const char *argp_program_version = STANDARD_HURD_VERSION (console);
char *netfs_server_name = "console";
char *netfs_server_version = HURD_VERSION;
int netfs_maxsymlinks = 16; /* Arbitrary. */
struct vcons
{
/* Protected by cons->lock. */
vcons_t next;
vcons_t prev;
/* We acquire one reference per netnode. */
int refcnt;
/* The following members remain constant over the lifetime of the
object and don't need to be locked. */
int id;
char *name;
cons_t cons;
struct mutex lock;
/* Indicates if OWNER_ID is initialized. */
int has_owner;
/* Specifies the ID of the process that should receive the WINCH
signal for this virtual console. */
int owner_id;
/* Nodes in the filesystem referring to this virtual console. */
struct node *dir_node;
struct node *cons_node;
struct node *disp_node;
/* The output queue holds the characters that are to be outputted.
The display driver might refuse to handle some incomplete
multi-byte or composed character at the end of the buffer, so we
have to keep them around. */
struct mutex output_lock;
int output_stopped;
struct condition output_resumed;
char *output;
size_t output_allocated;
size_t output_size;
struct mutex input_lock;
/* XXX input queue. */
char *input;
size_t input_allocated;
size_t input_size;
};
struct cons
{
/* The lock protects the console, all virtual consoles contained in
it and the reference counters. It also locks the configuration. */
struct mutex lock;
vcons_t vcons_list;
/* The encoding. */
const char *encoding;
struct node *node;
mach_port_t underlying;
/* A template for the stat information of all nodes. */
struct stat stat_template;
} mycons;
cons_t cons = &mycons;
/* Lookup the virtual console with number ID in the console CONS,
acquire a reference for it, and return it in R_VCONS. If CREATE is
true, the virtual console will be created if it doesn't exist yet.
If CREATE is true, and ID 0, the first free virtual console id is
used. */
error_t
vcons_lookup (cons_t cons, int id, int create, vcons_t *r_vcons)
{
vcons_t previous_vcons = 0;
vcons_t vcons;
if (!id && !create)
return EINVAL;
mutex_lock (&cons->lock);
if (id)
{
if (cons->vcons_list && cons->vcons_list->id <= id)
{
previous_vcons = cons->vcons_list;
while (previous_vcons->next && previous_vcons->next->id <= id)
previous_vcons = previous_vcons->next;
if (previous_vcons->id == id)
{
previous_vcons->refcnt++;
mutex_unlock (&cons->lock);
*r_vcons = previous_vcons;
return 0;
}
}
else if (!create)
{
mutex_unlock (&cons->lock);
return ESRCH;
}
}
else
{
id = 1;
if (cons->vcons_list && cons->vcons_list->id == 1)
{
previous_vcons = cons->vcons_list;
while (previous_vcons && previous_vcons->id == id)
{
id++;
previous_vcons = previous_vcons->next;
}
}
}
vcons = calloc (1, sizeof (struct vcons));
if (!vcons)
{
mutex_unlock (&vcons->cons->lock);
return ENOMEM;
}
vcons->cons = cons;
vcons->refcnt = 1;
vcons->id = id;
asprintf (&vcons->name, "%i", id);
mutex_init (&vcons->lock);
mutex_init (&vcons->output_lock);
condition_init (&vcons->output_resumed);
/* Insert the virtual console into the doubly linked list. */
if (previous_vcons)
{
vcons->prev = previous_vcons;
if (previous_vcons->next)
{
previous_vcons->next->prev = vcons;
vcons->next = previous_vcons->next;
}
previous_vcons->next = vcons;
}
else {
if (cons->vcons_list)
{
cons->vcons_list->prev = vcons;
vcons->next = cons->vcons_list;
}
cons->vcons_list = vcons;
}
mutex_unlock (&cons->lock);
*r_vcons = vcons;
return 0;
}
/* Acquire an additional reference to the virtual console VCONS. */
void
vcons_ref (vcons_t vcons)
{
cons_t cons = vcons->cons;
mutex_lock (&cons->lock);
vcons->refcnt++;
mutex_unlock (&cons->lock);
}
/* Release a reference to the virtual console VCONS. If this was the
last reference the virtual console is destroyed. */
void
vcons_release (vcons_t vcons)
{
cons_t cons = vcons->cons;
mutex_lock (&cons->lock);
if (!--vcons->refcnt)
{
/* As we keep a reference for all input focus groups pointing to
the virtual console, and a reference for the active console,
we know that without references, this virtual console is
neither active nor used by any input group. */
/* XXX Destroy the state. */
free (vcons->name);
if (vcons->prev)
vcons->prev->next = vcons->next;
if (vcons->next)
vcons->next->prev = vcons->prev;
if (!vcons->prev && !vcons->next)
vcons->cons->vcons_list = NULL;
free (vcons);
}
mutex_unlock (&cons->lock);
}
typedef enum
{
VCONS_NODE_DIR = 0,
VCONS_NODE_CONSOLE,
VCONS_NODE_DISPLAY
} vcons_node_type;
/* Make a new virtual node. Always consumes the ports. */
static error_t
new_node (struct node **np, vcons_t vcons, vcons_node_type type)
{
struct netnode *nn = calloc (1, sizeof *nn);
if (nn == 0)
return ENOMEM;
nn->vcons = vcons;
*np = netfs_make_node (nn);
if (*np == 0)
{
free (nn);
return ENOMEM;
}
(*np)->nn_stat = cons->stat_template;
switch (type)
{
case VCONS_NODE_DIR:
(*np)->nn_stat.st_ino = vcons->id << 2;
(*np)->nn_stat.st_mode |= S_IFDIR;
(*np)->nn_stat.st_size = 0;
break;
case VCONS_NODE_CONSOLE:
(*np)->nn_stat.st_ino = (vcons->id << 2) + 1;
(*np)->nn_stat.st_mode |= S_IFCHR; /* Don't set nn_translated! */
(*np)->nn_stat.st_size = 0;
break;
case VCONS_NODE_DISPLAY:
(*np)->nn_stat.st_ino = (vcons->id << 2) + 2;
(*np)->nn_stat.st_mode |= S_IFREG;
(*np)->nn_stat.st_size = 2000; /* XXX */
break;
}
/* If the underlying node isn't a directory, propagate read permission to
execute permission since we need that for lookups. */
if (! S_ISDIR (cons->stat_template.st_mode)
&& S_ISDIR ((*np)->nn_stat.st_mode))
{
if (cons->stat_template.st_mode & S_IRUSR)
(*np)->nn_stat.st_mode |= S_IXUSR;
if (cons->stat_template.st_mode & S_IRGRP)
(*np)->nn_stat.st_mode |= S_IXGRP;
if (cons->stat_template.st_mode & S_IROTH)
(*np)->nn_stat.st_mode |= S_IXOTH;
}
fshelp_touch (&(*np)->nn_stat, TOUCH_ATIME|TOUCH_MTIME|TOUCH_CTIME,
console_maptime);
return 0;
}
/* Node management. */
/* Node NP has no more references; free all its associated
storage. */
void
netfs_node_norefs (struct node *np)
{
vcons_t vcons = np->nn->vcons;
/* The root node does never go away. */
assert (!np->nn->cons && np->nn->vcons);
/* Find the back reference to ourself in the virtual console
structure, and delete it. */
mutex_lock (&vcons->lock);
if (np == vcons->dir_node)
vcons->dir_node = 0;
else if (np == vcons->cons_node)
vcons->cons_node = 0;
else
{
assert (np == vcons->disp_node);
vcons->disp_node = 0;
}
mutex_unlock (&vcons->lock);
/* Release our reference. */
vcons_release (vcons);
free (np->nn);
free (np);
}
/* Attempt to create a file named NAME in DIR for USER with MODE. Set
*NODE to the new node upon return. On any error, clear *NODE.
*NODE should be locked on success; no matter what, unlock DIR
before returning. */
error_t
netfs_attempt_create_file (struct iouser *user, struct node *dir,
char *name, mode_t mode, struct node **np)
{
/* We create virtual consoles dynamically on the fly, so there is no
need for an explicit create operation. */
*np = 0;
mutex_unlock (&dir->lock);
return EOPNOTSUPP;
}
/* Node NODE is being opened by USER, with FLAGS. NEWNODE is nonzero
if we just created this node. Return an error if we should not
permit the open to complete because of a permission
restriction. */
error_t
netfs_check_open_permissions (struct iouser *user, struct node *node,
int flags, int newnode)
{
error_t err = 0;
if (flags & O_READ)
err = fshelp_access (&node->nn_stat, S_IREAD, user);
if (!err && (flags & O_WRITE))
err = fshelp_access (&node->nn_stat, S_IWRITE, user);
if (!err && (flags & O_EXEC))
err = fshelp_access (&node->nn_stat, S_IEXEC, user);
return err;
}
/* This should attempt a utimes call for the user specified by CRED on node
NODE, to change the atime to ATIME and the mtime to MTIME. */
error_t
netfs_attempt_utimes (struct iouser *cred, struct node *node,
struct timespec *atime, struct timespec *mtime)
{
error_t err = fshelp_isowner (&node->nn_stat, cred);
int flags = TOUCH_CTIME;
if (! err)
{
if (mtime)
{
node->nn_stat.st_mtime = mtime->tv_sec;
node->nn_stat.st_mtime_usec = mtime->tv_nsec / 1000;
}
else
flags |= TOUCH_MTIME;
if (atime)
{
node->nn_stat.st_atime = atime->tv_sec;
node->nn_stat.st_atime_usec = atime->tv_nsec / 1000;
}
else
flags |= TOUCH_ATIME;
fshelp_touch (&node->nn_stat, flags, console_maptime);
}
return err;
}
/* Return the valid access types (bitwise OR of O_READ, O_WRITE, and O_EXEC)
in *TYPES for file NODE and user CRED. */
error_t
netfs_report_access (struct iouser *cred, struct node *node, int *types)
{
*types = 0;
if (fshelp_access (&node->nn_stat, S_IREAD, cred) == 0)
*types |= O_READ;
if (fshelp_access (&node->nn_stat, S_IWRITE, cred) == 0)
*types |= O_WRITE;
if (fshelp_access (&node->nn_stat, S_IEXEC, cred) == 0)
*types |= O_EXEC;
return 0;
}
/* Make sure that NP->nn_stat is filled with the most current
information. CRED identifies the user responsible for the
operation. NP is locked. */
error_t
netfs_validate_stat (struct node *np, struct iouser *cred)
{
/* We are always uptodate. */
return 0;
}
/* This should sync the file NODE completely to disk, for the user
CRED. If WAIT is set, return only after sync is completely
finished. */
error_t
netfs_attempt_sync (struct iouser *cred, struct node *np, int wait)
{
return 0;
}
/* Directory management. */
/* Lookup NAME in DIR for USER; set *NODE to the found name upon
return. If the name was not found, then return ENOENT. On any
error, clear *NODE. (*NODE, if found, should be locked, this call
should unlock DIR no matter what.) */
error_t
netfs_attempt_lookup (struct iouser *user, struct node *dir,
char *name, struct node **node)
{
error_t err;
*node = 0;
err = fshelp_access (&dir->nn_stat, S_IEXEC, user);
if (err)
goto out;
if (strcmp (name, ".") == 0)
{
/* Current directory -- just add an additional reference to DIR
and return it. */
netfs_nref (dir);
*node = dir;
goto out;
}
if (strcmp (name, "..") == 0)
{
/* Parent directory -- if this is the root directory, return
EAGAIN. Otherwise return the root node, because we know
that our hierarchy is only one level deep. */
if (dir->nn->cons)
err = EAGAIN;
else
{
netfs_nref (netfs_root_node);
*node = netfs_root_node;
}
goto out;
}
if (dir->nn->cons)
{
/* This is the root directory. Look up the desired virtual
console, creating it on the fly if necessary. */
vcons_t vcons;
int release_vcons = 0;
char *tail = NULL;
int id;
errno = 0;
id = strtol (name, &tail, 0);
if ((tail && *tail) || errno)
{
err = ENOENT;
goto out;
}
err = vcons_lookup (dir->nn->cons, id, 1, &vcons);
if (err == ESRCH || err == EINVAL)
err = ENOENT;
if (err)
goto out;
mutex_lock (&vcons->lock);
if (vcons->dir_node)
{
/* We already have a directory node for this virtual
console. Use that, acquire a reference for it, and drop
our extra reference to the virtual console. */
*node = vcons->dir_node;
netfs_nref (*node);
release_vcons = 1;
}
else
{
/* Create a new directory node, connsuming the reference to
the virtual console. */
err = new_node (node, vcons, VCONS_NODE_DIR);
if (!err)
vcons->dir_node = *node;
else
release_vcons = 1;
}
mutex_unlock (&vcons->lock);
if (release_vcons)
vcons_release (vcons);
}
else
{
/* This is a virtual console directory node. */
vcons_t vcons = dir->nn->vcons;
int ref_vcons = 0;
assert (dir == vcons->dir_node);
if (!strcmp (name, "console"))
{
mutex_lock (&vcons->lock);
if (vcons->cons_node)
{
*node = vcons->cons_node;
netfs_nref (*node);
}
else
{
err = new_node (node, vcons, VCONS_NODE_CONSOLE);
if (!err)
{
vcons->cons_node = *node;
ref_vcons = 1;
}
}
mutex_unlock (&vcons->lock);
}
else if (!strcmp (name, "display"))
{
mutex_lock (&vcons->lock);
if (vcons->disp_node)
{
*node = vcons->disp_node;
netfs_nref (*node);
}
else
{
err = new_node (node, vcons, VCONS_NODE_DISPLAY);
if (!err)
{
vcons->disp_node = *node;
ref_vcons = 1;
}
}
mutex_unlock (&vcons->lock);
}
else
err = ENOENT;
if (ref_vcons)
vcons_ref (vcons);
}
if (!err)
fshelp_touch (&dir->nn_stat, TOUCH_ATIME, console_maptime);
out:
mutex_unlock (&dir->lock);
if (err)
*node = 0;
else
mutex_lock (&(*node)->lock);
return err;
}
/* Returned directory entries are aligned to blocks this many bytes long.
Must be a power of two. */
#define DIRENT_ALIGN 4
#define DIRENT_NAME_OFFS offsetof (struct dirent, d_name)
/* Length is structure before the name + the name + '\0', all
padded to a four-byte alignment. */
#define DIRENT_LEN(name_len) \
((DIRENT_NAME_OFFS + (name_len) + 1 + (DIRENT_ALIGN - 1)) \
& ~(DIRENT_ALIGN - 1))
/* Implement the netfs_get_dirents callback as described in
<hurd/netfs.h>. */
error_t
netfs_get_dirents (struct iouser *cred, struct node *dir,
int first_entry, int num_entries, char **data,
mach_msg_type_number_t *data_len,
vm_size_t max_data_len, int *data_entries)
{
error_t err;
int count = 0;
size_t size = 0; /* Total size of our return block. */
struct vcons *first_vcons, *vcons;
/* Add the length of a directory entry for NAME to SIZE and return true,
unless it would overflow MAX_DATA_LEN or NUM_ENTRIES, in which case
return false. */
int bump_size (const char *name)
{
if (num_entries == -1 || count < num_entries)
{
size_t new_size = size + DIRENT_LEN (strlen (name));
if (max_data_len > 0 && new_size > max_data_len)
return 0;
size = new_size;
count++;
return 1;
}
else
return 0;
}
if (!dir->nn->cons && !(dir == dir->nn->vcons->dir_node))
return ENOTDIR;
if (dir->nn->cons)
{
mutex_lock (&dir->nn->cons->lock);
/* Find the first entry. */
for (first_vcons = dir->nn->cons->vcons_list, count = 2;
first_vcons && first_entry > count;
first_vcons = first_vcons->next)
count++;
count = 0;
}
/* Make space for the `.' and `..' entries. */
if (first_entry == 0)
bump_size (".");
if (first_entry <= 1)
bump_size ("..");
if (dir->nn->cons)
{
/* See how much space we need for the result. */
for (vcons = first_vcons; vcons; vcons = vcons->next)
#if 0
if (vcons->dir_node /* XXX see above */
&& !bump_size (vcons->name))
#else
if (!bump_size (vcons->name))
#endif
break;
}
else
{
if (first_entry <= 2)
bump_size ("console");
if (first_entry <= 3)
bump_size ("display");
}
/* Allocate it. */
*data = mmap (0, size, PROT_READ|PROT_WRITE, MAP_ANON, 0, 0);
err = ((void *) *data == (void *) -1) ? errno : 0;
if (! err)
/* Copy out the result. */
{
char *p = *data;
int add_dir_entry (const char *name, ino_t fileno, int type)
{
if (num_entries == -1 || count < num_entries)
{
struct dirent hdr;
size_t name_len = strlen (name);
size_t sz = DIRENT_LEN (name_len);
if (sz > size)
return 0;
else
size -= sz;
hdr.d_fileno = fileno;
hdr.d_reclen = sz;
hdr.d_type = type;
hdr.d_namlen = name_len;
memcpy (p, &hdr, DIRENT_NAME_OFFS);
strcpy (p + DIRENT_NAME_OFFS, name);
p += sz;
count++;
return 1;
}
else
return 0;
}
*data_len = size;
*data_entries = count;
count = 0;
if (dir->nn->cons)
{
/* Add `.' and `..' entries. */
if (first_entry == 0)
add_dir_entry (".", 2, DT_DIR);
if (first_entry <= 1)
add_dir_entry ("..", 2, DT_DIR);
/* Fill in the real directory entries. */
for (vcons = first_vcons; vcons; vcons = vcons->next)
if (
#if 0
/* XXX see above */ vcons->dir_node &&
#endif
!add_dir_entry (vcons->name,
vcons->id << 2, DT_DIR))
break;
mutex_unlock (&dir->nn->cons->lock);
}
else
{
/* Add `.' and `..' entries. */
if (first_entry == 0)
add_dir_entry (".", dir->nn_stat.st_ino, DT_DIR);
if (first_entry <= 1)
add_dir_entry ("..", 2, DT_DIR);
if (first_entry <= 2)
add_dir_entry ("console", (vcons->id << 2) + 1, DT_REG);
if (first_entry <= 3)
add_dir_entry ("display", (vcons->id << 2) + 2, DT_REG);
}
}
fshelp_touch (&dir->nn_stat, TOUCH_ATIME, console_maptime);
return err;
}
/* This should sync the entire remote filesystem. If WAIT is set, return
only after sync is completely finished. */
error_t
netfs_attempt_syncfs (struct iouser *cred, int wait)
{
return 0;
}
/* This should attempt a chmod call for the user specified by CRED on node
NODE, to change the owner to UID and the group to GID. */
error_t
netfs_attempt_chown (struct iouser *cred, struct node *node,
uid_t uid, uid_t gid)
{
cons_t cons = node->nn->cons;
vcons_t vcons;
error_t err;
if (!cons)
return EOPNOTSUPP;
err = file_chown (cons->underlying, uid, gid);
if (err)
return err;
/* Change NODE's owner. */
node->nn_stat.st_uid = uid;
node->nn_stat.st_gid = gid;
mutex_lock (&cons->lock);
cons->stat_template.st_uid = uid;
cons->stat_template.st_gid = gid;
/* Change the owner of each leaf node. */
for (vcons = cons->vcons_list; vcons; vcons = vcons->next)
{
if (vcons->dir_node)
{
vcons->dir_node->nn_stat.st_uid = uid;
vcons->dir_node->nn_stat.st_gid = gid;
}
if (vcons->cons_node)
{
vcons->cons_node->nn_stat.st_uid = uid;
vcons->cons_node->nn_stat.st_gid = gid;
}
if (vcons->disp_node)
{
vcons->disp_node->nn_stat.st_uid = uid;
vcons->disp_node->nn_stat.st_gid = gid;
}
}
mutex_unlock (&cons->lock);
fshelp_touch (&node->nn_stat, TOUCH_CTIME, console_maptime);
return err;
}
/* This should attempt a chauthor call for the user specified by CRED on node
NODE, to change the author to AUTHOR. */
error_t
netfs_attempt_chauthor (struct iouser *cred, struct node *node, uid_t author)
{
cons_t cons = node->nn->cons;
vcons_t vcons;
error_t err;
if (!cons)
return EOPNOTSUPP;
err = file_chauthor (cons->underlying, author);
if (err)
return err;
/* Change NODE's owner. */
node->nn_stat.st_author = author;
mutex_lock (&cons->lock);
cons->stat_template.st_author = author;
/* Change the owner of each leaf node. */
for (vcons = cons->vcons_list; vcons; vcons = vcons->next)
{
if (vcons->dir_node)
vcons->dir_node->nn_stat.st_author = author;
if (vcons->cons_node)
vcons->cons_node->nn_stat.st_author = author;
if (vcons->disp_node)
vcons->disp_node->nn_stat.st_author = author;
}
mutex_unlock (&cons->lock);
fshelp_touch (&node->nn_stat, TOUCH_CTIME, console_maptime);
return err;
}
/* This should attempt a chmod call for the user specified by CRED on node
NODE, to change the mode to MODE. Unlike the normal Unix and Hurd meaning
of chmod, this function is also used to attempt to change files into other
types. If such a transition is attempted which is impossible, then return
EOPNOTSUPP. */
error_t
netfs_attempt_chmod (struct iouser *cred, struct node *node, mode_t mode)
{
error_t err;
mode &= ~S_ITRANS;
if ((mode & S_IFMT) == 0)
mode |= (node->nn_stat.st_mode & S_IFMT);
if (!node->nn->cons || ((mode & S_IFMT) != (node->nn_stat.st_mode & S_IFMT)))
return EOPNOTSUPP;
err = file_chmod (node->nn->cons->underlying, mode & ~S_IFMT);
if (err)
return err;
node->nn_stat.st_mode = mode;
fshelp_touch (&node->nn_stat, TOUCH_CTIME, console_maptime);
return err;
}
/* The user must define this function. Attempt to turn locked node NP
(user CRED) into a symlink with target NAME. */
error_t
netfs_attempt_mksymlink (struct iouser *cred, struct node *np, char *name)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_mkdev (struct iouser *cred, struct node *np,
mode_t type, dev_t indexes)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_chflags (struct iouser *cred, struct node *np, int flags)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_set_size (struct iouser *cred, struct node *np, off_t size)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_statfs (struct iouser *cred, struct node *np, struct statfs *st)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_mkdir (struct iouser *user, struct node *dir,
char *name, mode_t mode)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_unlink (struct iouser *user, struct node *dir, char *name)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_rename (struct iouser *user, struct node *fromdir,
char *fromname, struct node *todir,
char *toname, int excl)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_rmdir (struct iouser *user,
struct node *dir, char *name)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_link (struct iouser *user, struct node *dir,
struct node *file, char *name, int excl)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_mkfile (struct iouser *user, struct node *dir,
mode_t mode, struct node **np)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_readlink (struct iouser *user, struct node *np, char *buf)
{
return EOPNOTSUPP;
}
error_t
netfs_attempt_read (struct iouser *cred, struct node *np,
off_t offset, size_t *len, void *data)
{
vcons_t vcons = np->nn->vcons;
if (!vcons || np == vcons->dir_node)
return EOPNOTSUPP;
if (np == vcons->cons_node)
*len = 0; /* Pass input queue content to caller. */
else
{
assert (np == vcons->disp_node);
*len = 0; /* Pass display content to caller. */
}
return 0;
}
error_t
netfs_attempt_write (struct iouser *cred, struct node *np,
off_t offset, size_t *len, void *data)
{
vcons_t vcons = np->nn->vcons;
if (!vcons || np == vcons->dir_node)
return EOPNOTSUPP;
if (np == vcons->cons_node)
*len = 0; /* Write input to text matrix. */
else
{
assert (np == vcons->disp_node);
*len = 0; /* Put input into input queue. */
}
return 0;
}
int
main (int argc, char **argv)
{
error_t err;
mach_port_t bootstrap;
struct stat ul_stat;
struct netnode root_nn = { cons: cons, vcons: 0 };
mtrace();
struct argp argp
= { NULL, NULL, NULL,
"A translator that provides virtual console displays." };
/* Parse our command line arguments (all none of them). */
argp_parse (&argp, argc, argv, 0, 0, 0);
task_get_bootstrap_port (mach_task_self (), &bootstrap);
netfs_init ();
/* Create the root node (some attributes initialized below). */
netfs_root_node = netfs_make_node (&root_nn);
if (! netfs_root_node)
error (5, ENOMEM, "Cannot create root node");
err = maptime_map (0, 0, &console_maptime);
if (err)
error (6, err, "Cannot map time");
mutex_init (&cons->lock);
cons->node = netfs_root_node;
cons->underlying = netfs_startup (bootstrap, O_READ);
if (cons->underlying == MACH_PORT_NULL)
error (5, err, "Cannot get underlying node");
err = io_stat (cons->underlying, &ul_stat);
if (err)
error (6, err, "Cannot stat underlying node");
/* CONS.stat_template contains some fields that are inherited by all
nodes we create. */
cons->stat_template.st_uid = ul_stat.st_uid;
cons->stat_template.st_gid = ul_stat.st_gid;
cons->stat_template.st_author = ul_stat.st_author;
cons->stat_template.st_mode = (ul_stat.st_mode & ~S_IFMT & ~S_ITRANS);
cons->stat_template.st_fsid = getpid ();
cons->stat_template.st_nlink = 1;
cons->stat_template.st_fstype = FSTYPE_MISC;
/* Initialize the root node's stat information. */
netfs_root_node->nn_stat = cons->stat_template;
netfs_root_node->nn_stat.st_ino = 2;
netfs_root_node->nn_stat.st_mode |= S_IFDIR;
netfs_root_node->nn_translated = 0;
/* If the underlying node isn't a directory, propagate read permission to
execute permission since we need that for lookups. */
if (! S_ISDIR (ul_stat.st_mode))
{
if (ul_stat.st_mode & S_IRUSR)
netfs_root_node->nn_stat.st_mode |= S_IXUSR;
if (ul_stat.st_mode & S_IRGRP)
netfs_root_node->nn_stat.st_mode |= S_IXGRP;
if (ul_stat.st_mode & S_IROTH)
netfs_root_node->nn_stat.st_mode |= S_IXOTH;
}
fshelp_touch (&netfs_root_node->nn_stat, TOUCH_ATIME|TOUCH_MTIME|TOUCH_CTIME,
console_maptime);
netfs_server_loop (); /* Never returns. */
/*NOTREACHED*/
return 0;
}
--
`Rhubarb is no Egyptian god.' Debian http://www.debian.org brinkmd@debian.org
Marcus Brinkmann GNU http://www.gnu.org marcus@gnu.org
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de
- netfs part of a console server with server-client model,
Marcus Brinkmann <=
- Re: netfs part of a console server with server-client model, Niels Möller, 2002/06/02
- Re: netfs part of a console server with server-client model, Marcus Brinkmann, 2002/06/02
- Re: netfs part of a console server with server-client model, Niels Möller, 2002/06/02
- Re: netfs part of a console server with server-client model, Marcus Brinkmann, 2002/06/02
- Re: netfs part of a console server with server-client model, Niels Möller, 2002/06/02
- Re: netfs part of a console server with server-client model, Marcus Brinkmann, 2002/06/02
- Re: netfs part of a console server with server-client model, Niels Möller, 2002/06/03
Re: netfs part of a console server with server-client model, Marcus Brinkmann, 2002/06/02