Released 2.7

Restore FAQ file

.gitignore

@@ -20,6 +20,7 @@ Makefile
 *.m4
 stamp-h*
 config.*
+sshfs.1
 /sshfs
 /ltmain.sh
 /configure

AUTHORS

@@ -15,8 +15,15 @@ Contributors (autogenerated list)
 
 Alan Jenkins <alan.christopher.jenkins@gmail.com>
 Alexander Neumann <alexander@bumpern.de>
+Benjamin Fleischer <fleiben@gmail.com>
 Chris Wolfe <cwolfe@chromium.org>
+gala <gala132@users.noreply.github.com>
+George Vlahavas <vlahavas@gmail.com>
+Julio Merino <jmmv@google.com>
+Julio Merino <jmmv@meroh.net>
 Mike Kelly <mike@pair.com>
 Miklos Szeredi <miklos@szeredi.hu>
+Miklos Szeredi <mszeredi@suse.cz>
 Nikolaus Rath <Nikolaus@rath.org>
 Percy Jahn <email@percyjahn.de>
+Qais Patankar <qaisjp@gmail.com>

ChangeLog

@@ -1,10 +1,19 @@
-Release 2.7 (2015-01-28)
+Release 2.6 (2015-03-01)
+------------------------
+
+* Integrated osxfuse's copy of sshfs, which means that sshfs now works
+  on OS X out of the box.
+* Added -o cache_max_size=N option to let users tune the maximum size of
+  the cache in number of entries.
+* Added -o cache_clean_interval=N and -o cache_min_clean_interval=N
+  options to let users tune the cleaning behavior of the cache.
+
+Release 2.6 (2015-01-28)
 ------------------------
 
 * New maintainer (Nikolaus Rath <Nikolaus@rath.org>)
 
-
-Release 2.6 (2014-01-14)
+Release 2.5 (2014-01-14)
 ------------------------
 
 * Some performance improvements for large directories.

FAQ

@@ -0,0 +1,263 @@
+SSHFS FAQ
+=========
+
+1.     I've found a bug and there's no solution in this FAQ, what
+       should I do?
+2.     Advantage of sshfs over NFS and Samba
+3.     Create the device node
+4.     mv fails with "Operation not permitted"
+5.     cvs fails with "cvs [status aborted]: cannot get working
+       directory: No such file or directory" in a sshfs mounted directory
+6.     Changes on the server are not immediately visible in the
+       mounted directory.
+7.     Configuring the ssh connection
+8.     What are the no_readahead and sshfs_sync options for?
+9.    Why does df return strange values on partitions mounted via
+       sshfs?
+10.    How do I specify the remote mount point (since the example
+       defaults to the home directory)
+11.    sshfs hangs after a while
+12.    Following symlinks on the server side
+13.    Making absolute symlinks work
+14.    Mounting as root
+15.    Exporting via NFS
+16.    Automatical mounting using /etc/fstab
+17.    Why does SVN (etc...) fail with permission denied?
+18.    Why does SVN (etc...) fail to rename files?
+19.    Is there some neat way to do it in reverse?
+20.    Might it be reasonable to disallow loops?
+21.    How to mount through an intermediary ssh server, eg:
+       localhost -> A -> B; mount B from localhost
+22.    Alternative Solution
+23.    I seem to have successfully mounted a remote directory, but
+       performing an `ls -l` on the directory above the mount point shows
+       the mount point's attributes as `? ? ? ? ? ?`. Nothing shows up in
+       the directory either. What am I doing wrong?
+
+--------------------------------------------------------------------------
+
+1. I've found a bug and there's no solution in this FAQ, what should I do?
+
+   Please report the bug in the Github issue tracker:
+   https://github.com/libfuse/sshfs/issues
+
+   Also logs with debugging output can be useful for diagnosing the
+   problem. Try running sshfs with the following options:
+     sshfs -odebug,sshfs_debug,loglevel=debug ...
+
+   Doing strace on the application which fails may also sometimes help:
+     strace -f -o /tmp/strace application args ...
+
+   Note that large messages (over 40k) will be rejected from the mailing
+   list. So try to keep the logs as short as possible.
+
+
+2. Advantage of sshfs over NFS and Samba
+
+   Users can mount remote resources they already have ssh access to,
+   without requiring the remote machine to export the resource.
+
+   The remote resource can be mounted when it is needed in a location
+   that is convenient for the user at that time, without needing to rely
+   on a central, root-controlled file system table.
+
+   Automatic mounting, if desired, can be added to a shell script such as
+   .bashrc (provided authentication is done using RSA/DSA keys).
+
+   Resources can be mounted over slow and unreliable (distant)
+   connections.
+
+3. Create the device node
+
+   If you don't use udev, you may get this error message:
+
+   fusermount: failed to open /dev/fuse: No such device or address
+
+   Before loading the fuse kernel module, create the device node
+   manually:
+
+     mknod -m 666 /dev/fuse c 10 229
+
+4. mv fails with "Operation not permitted".
+
+   Use -o workaround=rename (requires sshfs version >= 1.3).
+
+5. cvs fails with "cvs status aborted?: cannot get working directory: No such
+   file or directory" in a sshfs mounted directory
+
+   Use the -oreaddir_ino option. Example:
+     sshfs -oreaddir_ino hostname:remote_dir mount_point
+
+6. Changes on the server are not immediately visible in the mounted directory.
+
+   By default, sshfs caches things for 20 seconds, use -o cache_timeout=N
+   to change the default cache timeout (in seconds) or -o cache=no for
+   disabling the cache.
+
+   You can also control cache timeouts for directory listing etc with
+   -o cache_stat_timeout=N,
+   -o cache_dir_timout=N, and
+   -o cache_link_timout=N.
+
+7. Configuring the ssh connection
+
+   In addition to flags like -C, -p, and -o SSHOPT...=, you may find it
+   easier to edit your /.ssh/config file. You can add an entry with any
+   customization you want, test it with ssh, and finally use it with
+   sshfs. As a bonus, you get a short mnemonic for your configuration.
+
+8. What are the no_readahead and sshfs_sync options for?
+
+   These disable read and write optimizations respectively. They don't
+   really make sense unless you're doing something special.
+
+9. Why does df return strange values on partitions mounted via sshfs?
+
+   Because the SFTP protocol doesn't have a statfs operation this is
+   currently not possible to display proper usage on remote partition.
+
+10. How do I specfy the remote mount point (since the example defaults to the
+    home directory)
+
+    The example shows:
+      sshfs hostname: mountpoint
+
+    To specify a remote mount point use:
+      sshfs hostname:remotemountpoint mountpoint
+
+    This might be obvious to others, but I ended up looking up the
+    interface to sftp to see if I could learn how to specify the remote
+    mount point, then thought about the way that scp specifies the remote
+    directory, and it worked.
+
+11. sshfs hangs after a while
+
+    Mounting works fine, I can use the files in Mountpoint as good as any
+    other files on my system, but after bit of time, changing nothing on
+    the remote files sshfs crashes. This means, I can not cd into the
+    Mountpoint (xterm hangs, nautilus hangs... every program trying to
+    access the Mountpoint gets stuck, and won't return).
+
+    Solution: add
+      ServerAliveInterval 15
+
+    in your .ssh/config (or use -o ServerAliveInterval=15 on the sshfs
+    command line but I did not test that solution). This will force the
+    ssh connection to stay alive even if you have no activity.
+
+12. Following symlinks on the server side
+
+    The -o follow_symlinks option will enable this.
+
+13. Making absolute symlinks work
+
+    Use the -o transform_symlinks option, which will transform absolute
+    symlinks (ones which point somewhere inside the mount) into relative
+    ones.
+
+14. Mounting as root
+
+    Generally it's not possible to use an sshfs mount as a "real"
+    filesystem shared between multiple users. Some of this functionality
+    can be enabled with the -o allow_other and -o default_permissions
+    options, but files will not be created with the correct ownership,
+    etc...
+
+15. Exporting via NFS
+
+    Use the userspace NFS daemon http://sourceforge.net/projects/unfs
+
+16. Automatical mounting using /etc/fstab
+
+    A line in /etc/fstab has the following format:
+      sshfs#USERNAME@REMOTE_HOST:REMOTE_PATH MOUNT_POINT fuse SSHFS_OPTIONS 0 0
+
+    eg.
+      sshfs#guest@guest.login.com:data /mnt/guest fuse \
+        uid=1003,gid=100,umask=0,allow_other 0 0
+
+17. Why does SVN (etc...) fail with permission denied?
+
+    This is a bug that happens when an application creates a read-only
+    file opened for writing (e.g. open("foo", O_WRONLY|O_CREAT, 0444))
+
+    It has been fixed in sshfs version 1.3, but also requires FUSE version
+    >=2.5.X and Linux kernel version >=2.6.15.
+
+18. Why does SVN (etc...) fail to rename files?
+
+    $ svn co svn://anonsvn.kde.org/home/kde/trunk/KDE/kdelibs
+    svn: Can't move 'kdelibs/.svn/tmp/entries' to 'kdelibs/.svn/entries':
+    Operation not permitted
+
+    The reason is that SFTP protocol version 3 (which is implemented by
+    OpenSSH's sftp-server) defines the rename operation slightly
+    differently than POSIX. The difference is that renaming to an existing
+    file or directory will fail instead of atomically replacing the old
+    file.
+
+    The -o workaround=rename option will try to emulate POSIX rename
+    semantics, but it cannot guarantee atomicity. In most of the cases
+    this doesn't matter, and things will work fine with this option.
+
+19. Is there some neat way to do it in reverse?
+
+    You want to mount a USB thumb drive onto a file server that is rather
+    remote.
+
+    Assuming this is difficult because the laptop with the thumb drive is
+    sitting behind NAT, firewalls, etc. then you need to create a
+    port-forward:
+      client$ ssh -R 2222:localhost:22 server
+      server$ sshfs -p 2222 localhost:/media/usb1 myusb1
+
+    (Now, is there is there a smarter way that does not involve port
+    opening login permissions in an undesireable direction?)
+
+20. Might it be reasonable to disallow loops?
+
+    sshfs localhost:/mnt /mnt
+
+    This seems to produce undesirable results. --JoshuaRodman
+
+21. How to mount through an intermediary ssh server, eg: localhost -> A -> B;
+    mount B from localhost
+
+    Start by mounting the folder you need that is on "a" to a folder on
+    "b" then mount the new folder that is on "a" to a folder on localhost.
+
+    IE: (These are NOT real commands, but a sequence of steps.
+    A mounts B:/home/x on /mnt/Bx
+    localhost mounts A:/mnt/Bx on ~/mydir
+
+22. Alternative Solution:
+
+    1) Create a shell script to wrap the tunneling of one ssh command over
+    another,
+      $ cat >Atunnel <<EOF
+      #!/bin/bash
+      ssh -q A ssh -q "$@"
+      EOF
+      $ chmod u+x Atunnel
+
+    N.B. make sure to put this somewhere on your path. 2) sshfs mount as
+    normal but using this script as the ssh command.
+      $ sshfs -o ssh_command='Atunnel' B: ~/mydir
+
+23. I seem to have successfully mounted a remote directory, but performing an
+    `ls -l` on the directory above the mount point shows the mount point's
+    attributes as `? ? ? ? ? ?`. Nothing shows up in the directory either. What
+    am I doing wrong?
+
+    You probably specified a remote path with the tilde (~) in it. This
+    doesn't seem to work. Instead, specify an absolute remote path:
+      sshfs username@remote.host:/home/username/whatever my/mount
+
+    What options do i use to make playing media files (music) over sshfs more
+    efficient?
+
+    MacFUSE doesn't seem to let me move files from one directory to another. It
+    first asks for my local user password (i.e. the password on my Macbook Pro)
+    and then produces the error message "The operation cannot be completed
+    because one or more required items cannot be found. (Error code -120).
+

Makefile.am

@@ -6,16 +6,27 @@ sshfs_SOURCES = sshfs.c cache.c cache.h
 if FUSE_OPT_COMPAT
 sshfs_SOURCES += compat/fuse_opt.c compat/fuse_opt.h
 endif
+if DARWIN_COMPAT
+sshfs_SOURCES += compat/darwin_compat.c compat/darwin_compat.h
+endif
 
 sshfs_LDADD = $(SSHFS_LIBS)
 sshfs_CFLAGS = $(SSHFS_CFLAGS)
-sshfs_CPPFLAGS = -D_REENTRANT -DFUSE_USE_VERSION=26 -DLIBDIR=\"$(libdir)\"
+sshfs_CPPFLAGS = -D_REENTRANT -DFUSE_USE_VERSION=26 -DLIBDIR=\"$(libdir)\" \
+                 -DIDMAP_DEFAULT="\"$(IDMAP_DEFAULT)\""
 
-EXTRA_DIST = sshnodelay.c
-CLEANFILES = sshnodelay.so
+EXTRA_DIST = sshnodelay.c sshfs.1.in
+CLEANFILES = sshnodelay.so sshfs.1 sshfs.1.tmp
 
 dist_man_MANS = sshfs.1
 
+sshfs.1: sshfs.1.in
+	$(AM_V_GEN)sed \
+	    -e 's,__IDMAP_DEFAULT__,$(IDMAP_DEFAULT),g' \
+	    -e 's,__UNMOUNT_COMMAND__,$(UNMOUNT_COMMAND),g' \
+	    <sshfs.1.in >sshfs.1.tmp || exit 1; \
+	mv sshfs.1.tmp sshfs.1
+
 if SSH_NODELAY_SO
 all-local: sshnodelay.so
 

README.md

@@ -50,7 +50,7 @@ sftp(1) and ssh_config(5)), including the remote port number
 
 To unmount the filesystem:
 
-    fusermount -u moutpoint
+    fusermount -u mountpoint
 
 
 Installing

cache.c

@@ -14,16 +14,19 @@
 #include <glib.h>
 #include <pthread.h>
 
-#define DEFAULT_CACHE_TIMEOUT 20
-#define MAX_CACHE_SIZE 10000
-#define MIN_CACHE_CLEAN_INTERVAL 5
-#define CACHE_CLEAN_INTERVAL 60
+#define DEFAULT_CACHE_TIMEOUT_SECS 20
+#define DEFAULT_MAX_CACHE_SIZE 10000
+#define DEFAULT_CACHE_CLEAN_INTERVAL_SECS 60
+#define DEFAULT_MIN_CACHE_CLEAN_INTERVAL_SECS 5
 
 struct cache {
 	int on;
-	unsigned stat_timeout;
-	unsigned dir_timeout;
-	unsigned link_timeout;
+	unsigned int stat_timeout_secs;
+	unsigned int dir_timeout_secs;
+	unsigned int link_timeout_secs;
+	unsigned int max_size;
+	unsigned int clean_interval_secs;
+	unsigned int min_clean_interval_secs;
 	struct fuse_cache_operations *next_oper;
 	GHashTable *table;
 	pthread_mutex_t lock;
@@ -70,9 +73,9 @@ static int cache_clean_entry(void *key_, struct node *node, time_t *now)
 static void cache_clean(void)
 {
 	time_t now = time(NULL);
-	if (now > cache.last_cleaned + MIN_CACHE_CLEAN_INTERVAL &&
-	    (g_hash_table_size(cache.table) > MAX_CACHE_SIZE ||
-	     now > cache.last_cleaned + CACHE_CLEAN_INTERVAL)) {
+	if (now > cache.last_cleaned + cache.min_clean_interval_secs &&
+	    (g_hash_table_size(cache.table) > cache.max_size ||
+	     now > cache.last_cleaned + cache.clean_interval_secs)) {
 		g_hash_table_foreach_remove(cache.table,
 					    (GHRFunc) cache_clean_entry, &now);
 		cache.last_cleaned = now;
@@ -172,7 +175,7 @@ void cache_add_attr(const char *path, const struct stat *stbuf, uint64_t wrctr)
 	if (wrctr == cache.write_ctr) {
 		node = cache_get(path);
 		node->stat = *stbuf;
-		node->stat_valid = time(NULL) + cache.stat_timeout;
+		node->stat_valid = time(NULL) + cache.stat_timeout_secs;
 		if (node->stat_valid > node->valid)
 			node->valid = node->stat_valid;
 		cache_clean();
@@ -188,7 +191,7 @@ static void cache_add_dir(const char *path, char **dir)
 	node = cache_get(path);
 	g_strfreev(node->dir);
 	node->dir = dir;
-	node->dir_valid = time(NULL) + cache.dir_timeout;
+	node->dir_valid = time(NULL) + cache.dir_timeout_secs;
 	if (node->dir_valid > node->valid)
 		node->valid = node->dir_valid;
 	cache_clean();
@@ -210,7 +213,7 @@ static void cache_add_link(const char *path, const char *link, size_t size)
 	node = cache_get(path);
 	g_free(node->link);
 	node->link = g_strndup(link, my_strnlen(link, size-1));
-	node->link_valid = time(NULL) + cache.link_timeout;
+	node->link_valid = time(NULL) + cache.link_timeout_secs;
 	if (node->link_valid > node->valid)
 		node->valid = node->link_valid;
 	cache_clean();
@@ -570,20 +573,28 @@ struct fuse_operations *cache_init(struct fuse_cache_operations *oper)
 static const struct fuse_opt cache_opts[] = {
 	{ "cache=yes", offsetof(struct cache, on), 1 },
 	{ "cache=no", offsetof(struct cache, on), 0 },
-	{ "cache_timeout=%u", offsetof(struct cache, stat_timeout), 0 },
-	{ "cache_timeout=%u", offsetof(struct cache, dir_timeout), 0 },
-	{ "cache_timeout=%u", offsetof(struct cache, link_timeout), 0 },
-	{ "cache_stat_timeout=%u", offsetof(struct cache, stat_timeout), 0 },
-	{ "cache_dir_timeout=%u", offsetof(struct cache, dir_timeout), 0 },
-	{ "cache_link_timeout=%u", offsetof(struct cache, link_timeout), 0 },
+	{ "cache_timeout=%u", offsetof(struct cache, stat_timeout_secs), 0 },
+	{ "cache_timeout=%u", offsetof(struct cache, dir_timeout_secs), 0 },
+	{ "cache_timeout=%u", offsetof(struct cache, link_timeout_secs), 0 },
+	{ "cache_stat_timeout=%u", offsetof(struct cache, stat_timeout_secs), 0 },
+	{ "cache_dir_timeout=%u", offsetof(struct cache, dir_timeout_secs), 0 },
+	{ "cache_link_timeout=%u", offsetof(struct cache, link_timeout_secs), 0 },
+	{ "cache_max_size=%u", offsetof(struct cache, max_size), 0 },
+	{ "cache_clean_interval=%u", offsetof(struct cache,
+	                                      clean_interval_secs), 0 },
+	{ "cache_min_clean_interval=%u", offsetof(struct cache,
+	                                          min_clean_interval_secs), 0 },
 	FUSE_OPT_END
 };
 
 int cache_parse_options(struct fuse_args *args)
 {
-	cache.stat_timeout = DEFAULT_CACHE_TIMEOUT;
-	cache.dir_timeout = DEFAULT_CACHE_TIMEOUT;
-	cache.link_timeout = DEFAULT_CACHE_TIMEOUT;
+	cache.stat_timeout_secs = DEFAULT_CACHE_TIMEOUT_SECS;
+	cache.dir_timeout_secs = DEFAULT_CACHE_TIMEOUT_SECS;
+	cache.link_timeout_secs = DEFAULT_CACHE_TIMEOUT_SECS;
+	cache.max_size = DEFAULT_MAX_CACHE_SIZE;
+	cache.clean_interval_secs = DEFAULT_CACHE_CLEAN_INTERVAL_SECS;
+	cache.min_clean_interval_secs = DEFAULT_MIN_CACHE_CLEAN_INTERVAL_SECS;
 	cache.on = 1;
 
 	return fuse_opt_parse(args, &cache, cache_opts, NULL);

compat/darwin_compat.c

@@ -0,0 +1,246 @@
+/*
+ * Copyright (c) 2006-2008 Amit Singh/Google Inc.
+ * Copyright (c) 2012 Anatol Pomozov
+ * Copyright (c) 2011-2013 Benjamin Fleischer
+ */
+
+#include "darwin_compat.h"
+
+#include <assert.h>
+#include <errno.h>
+#include <sys/types.h>
+
+/*
+ * Semaphore implementation based on:
+ *
+ * Copyright (C) 2000,02 Free Software Foundation, Inc.
+ * This file is part of the GNU C Library.
+ * Written by Ga<EB>l Le Mignot <address@hidden>
+ *
+ * The GNU C Library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Library General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * The GNU C Library 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
+ * Library General Public License for more details.
+ *
+ * You should have received a copy of the GNU Library General Public
+ * License along with the GNU C Library; see the file COPYING.LIB.  If not,
+ * write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 02111-1307, USA.
+ */
+
+/* Semaphores */
+
+#define __SEM_ID_NONE  ((int)0x0)
+#define __SEM_ID_LOCAL ((int)0xcafef00d)
+
+/* http://www.opengroup.org/onlinepubs/007908799/xsh/sem_init.html */
+int
+darwin_sem_init(darwin_sem_t *sem, int pshared, unsigned int value)
+{
+    if (pshared) {
+        errno = ENOSYS;
+        return -1;
+    }
+
+    sem->id = __SEM_ID_NONE;
+
+    if (pthread_cond_init(&sem->__data.local.count_cond, NULL)) {
+        goto cond_init_fail;
+    }
+
+    if (pthread_mutex_init(&sem->__data.local.count_lock, NULL)) {
+        goto mutex_init_fail;
+    }
+
+    sem->__data.local.count = value;
+    sem->id = __SEM_ID_LOCAL;
+
+    return 0;
+
+mutex_init_fail:
+
+    pthread_cond_destroy(&sem->__data.local.count_cond);
+
+cond_init_fail:
+
+    return -1;
+}
+
+/* http://www.opengroup.org/onlinepubs/007908799/xsh/sem_destroy.html */
+int
+darwin_sem_destroy(darwin_sem_t *sem)
+{
+    int res = 0;
+
+    pthread_mutex_lock(&sem->__data.local.count_lock);
+
+    sem->id = __SEM_ID_NONE;
+    pthread_cond_broadcast(&sem->__data.local.count_cond);
+
+    if (pthread_cond_destroy(&sem->__data.local.count_cond)) {
+        res = -1;
+    }
+
+    pthread_mutex_unlock(&sem->__data.local.count_lock);
+
+    if (pthread_mutex_destroy(&sem->__data.local.count_lock)) {
+        res = -1;
+    }
+
+    return res;
+}
+
+int
+darwin_sem_getvalue(darwin_sem_t *sem, unsigned int *sval)
+{
+    int res = 0;
+
+    pthread_mutex_lock(&sem->__data.local.count_lock);
+
+    if (sem->id != __SEM_ID_LOCAL) {
+        res = -1;
+        errno = EINVAL;
+    } else {
+        *sval = sem->__data.local.count;
+    }
+
+    pthread_mutex_unlock(&sem->__data.local.count_lock);
+
+    return res;
+}
+
+/* http://www.opengroup.org/onlinepubs/007908799/xsh/sem_post.html */
+int
+darwin_sem_post(darwin_sem_t *sem)
+{
+    int res = 0;
+
+    pthread_mutex_lock(&sem->__data.local.count_lock);
+
+    if (sem->id != __SEM_ID_LOCAL) {
+        res = -1;
+        errno = EINVAL;
+    } else if (sem->__data.local.count < DARWIN_SEM_VALUE_MAX) {
+        sem->__data.local.count++;
+        if (sem->__data.local.count == 1) {
+            pthread_cond_signal(&sem->__data.local.count_cond);
+        }
+    } else {
+        errno = ERANGE;
+        res = -1;
+    }
+
+    pthread_mutex_unlock(&sem->__data.local.count_lock);
+
+    return res;
+}
+
+/* http://www.opengroup.org/onlinepubs/009695399/functions/sem_timedwait.html */
+int
+darwin_sem_timedwait(darwin_sem_t *sem, const struct timespec *abs_timeout)
+{
+    int res = 0;
+
+    if (abs_timeout &&
+        (abs_timeout->tv_nsec < 0 || abs_timeout->tv_nsec >= 1000000000)) {
+        errno = EINVAL;
+        return -1;
+    }
+
+    pthread_cleanup_push((void(*)(void*))&pthread_mutex_unlock,
+                 &sem->__data.local.count_lock);
+
+    pthread_mutex_lock(&sem->__data.local.count_lock);
+
+    if (sem->id != __SEM_ID_LOCAL) {
+        errno = EINVAL;
+        res = -1;
+    } else {
+        if (!sem->__data.local.count) {
+            res = pthread_cond_timedwait(&sem->__data.local.count_cond,
+                             &sem->__data.local.count_lock,
+                             abs_timeout);
+        }
+        if (res) {
+            assert(res == ETIMEDOUT);
+            res = -1;
+            errno = ETIMEDOUT;
+        } else if (sem->id != __SEM_ID_LOCAL) {
+            res = -1;
+            errno = EINVAL;
+        } else {
+            sem->__data.local.count--;
+        }
+    }
+
+    pthread_cleanup_pop(1);
+
+    return res;
+}
+
+/* http://www.opengroup.org/onlinepubs/007908799/xsh/sem_trywait.html */
+int
+darwin_sem_trywait(darwin_sem_t *sem)
+{
+    int res = 0;
+
+    pthread_mutex_lock(&sem->__data.local.count_lock);
+
+    if (sem->id != __SEM_ID_LOCAL) {
+        res = -1;
+        errno = EINVAL;
+    } else if (sem->__data.local.count) {
+        sem->__data.local.count--;
+    } else {
+        res = -1;
+        errno = EAGAIN;
+    }
+
+    pthread_mutex_unlock (&sem->__data.local.count_lock);
+
+    return res;
+}
+
+/* http://www.opengroup.org/onlinepubs/007908799/xsh/sem_wait.html */
+int
+darwin_sem_wait(darwin_sem_t *sem)
+{
+    int res = 0;
+
+    pthread_cleanup_push((void(*)(void*))&pthread_mutex_unlock,
+                 &sem->__data.local.count_lock);
+
+    pthread_mutex_lock(&sem->__data.local.count_lock);
+
+    if (sem->id != __SEM_ID_LOCAL) {
+        errno = EINVAL;
+        res = -1;
+    } else {
+        if (!sem->__data.local.count) {
+            pthread_cond_wait(&sem->__data.local.count_cond,
+                      &sem->__data.local.count_lock);
+            if (!sem->__data.local.count) {
+                /* spurious wakeup, assume it is an interruption */
+                res = -1;
+                errno = EINTR;
+                goto out;
+            }
+        }
+        if (sem->id != __SEM_ID_LOCAL) {
+            res = -1;
+            errno = EINVAL;
+        } else {
+            sem->__data.local.count--;
+        }
+    }
+
+out:
+    pthread_cleanup_pop(1);
+
+    return res;
+}

compat/darwin_compat.h

@@ -0,0 +1,49 @@
+/*
+ * Copyright (c) 2006-2008 Amit Singh/Google Inc.
+ * Copyright (c) 2011-2013 Benjamin Fleischer
+ */
+
+#ifndef _DARWIN_COMPAT_
+#define _DARWIN_COMPAT_
+
+#include <pthread.h>
+
+/* Semaphores */
+
+typedef struct darwin_sem {
+    int id;
+    union {
+        struct
+        {
+            unsigned int    count;
+            pthread_mutex_t count_lock;
+            pthread_cond_t  count_cond;
+        } local;
+    } __data;
+} darwin_sem_t;
+
+#define DARWIN_SEM_VALUE_MAX ((int32_t)32767)
+
+int darwin_sem_init(darwin_sem_t *sem, int pshared, unsigned int value);
+int darwin_sem_destroy(darwin_sem_t *sem);
+int darwin_sem_getvalue(darwin_sem_t *sem, unsigned int *value);
+int darwin_sem_post(darwin_sem_t *sem);
+int darwin_sem_timedwait(darwin_sem_t *sem, const struct timespec *abs_timeout);
+int darwin_sem_trywait(darwin_sem_t *sem);
+int darwin_sem_wait(darwin_sem_t *sem);
+
+/* Caller must not include <semaphore.h> */
+
+typedef darwin_sem_t sem_t;
+
+#define sem_init(s, p, v)   darwin_sem_init(s, p, v)
+#define sem_destroy(s)      darwin_sem_destroy(s)
+#define sem_getvalue(s, v)  darwin_sem_getvalue(s, v)
+#define sem_post(s)         darwin_sem_post(s)
+#define sem_timedwait(s, t) darwin_sem_timedwait(s, t)
+#define sem_trywait(s)      darwin_sem_trywait(s)
+#define sem_wait(s)         darwin_sem_wait(s)
+
+#define SEM_VALUE_MAX       DARWIN_SEM_VALUE_MAX
+
+#endif /* _DARWIN_COMPAT_ */

configure.ac

@@ -1,4 +1,5 @@
-AC_INIT(sshfs, 2.6)
+AC_INIT(sshfs, 2.7)
+AC_CANONICAL_TARGET
 AM_INIT_AUTOMAKE([foreign subdir-objects])
 AM_CONFIG_HEADER(config.h)
 
@@ -11,6 +12,12 @@ sshnodelay_libs=$LIBS
 AC_SUBST(sshnodelay_libs)
 LIBS=
 
+case "$target_os" in
+    *linux*) osname=linux;;
+    *darwin*) osname=darwin;;
+    *) osname=unknown;;
+esac
+
 AC_ARG_ENABLE(sshnodelay,
 	[  --disable-sshnodelay    Don't compile NODELAY workaround for ssh])
 
@@ -43,10 +50,22 @@ oldlibs="$LIBS"
 LIBS="$LIBS $SSHFS_LIBS"
 AC_CHECK_FUNC([fuse_opt_parse], [have_fuse_opt_parse=yes])
 LIBS="$oldlibs"
-if test "$have_fuse_opt_parse" = no; then
-	CFLAGS="$CFLAGS -Icompat"
+if test "$have_fuse_opt_parse" = no -o "$osname" = darwin; then
+	CFLAGS="$CFLAGS -I${srcdir}/compat"
 fi
 AM_CONDITIONAL(FUSE_OPT_COMPAT, test "$have_fuse_opt_parse" = no)
+AM_CONDITIONAL(DARWIN_COMPAT, test "$osname" = darwin)
+
+AC_CHECK_PROG(UNMOUNT_COMMAND, fusermount, fusermount -u, umount)
+
+# TODO: Figure out why we special-case this in Darwin.  Would be nice if
+# the default setting was consistent across platforms so we wouldn't need
+# to care about it here.
+case "$osname" in
+    darwin) IDMAP_DEFAULT=user ;;
+    *) IDMAP_DEFAULT=none ;;
+esac
+AC_SUBST(IDMAP_DEFAULT)
 
 AC_CONFIG_FILES([Makefile])
 AC_OUTPUT

sshfs.1.in

@@ -7,7 +7,7 @@ SSHFS \- filesystem client based on ssh
 \fBsshfs\fP [\fIuser\fP@]\fBhost\fP:[\fIdir\fP] \fBmountpoint\fP [\fIoptions\fP]
 .SS unmounting
 .TP
-\fBfusermount -u mountpoint\fP
+\fB__UNMOUNT_COMMAND__ mountpoint\fP
 .SH DESCRIPTION
 SSHFS (Secure SHell FileSystem) is a file system for Linux (and other
 operating systems with a FUSE implementation, such as Mac OS X or FreeBSD)
@@ -97,14 +97,14 @@ fix buffer fillup bug in server (default: on)
 .RE
 .TP
 \fB\-o\fR idmap=TYPE
-user/group ID mapping, possible types are:
+user/group ID mapping (default: __IDMAP_DEFAULT__)
 .RS 8
 .TP
 none
-no translation of the ID space (default)
+no translation of the ID space
 .TP
 user
-only translate UID of connecting user
+only translate UID/GID of connecting user
 .TP
 file
 translate UIDs/GIDs based upon the contents of \fBuidfile \fR and
@@ -141,6 +141,8 @@ path to sftp server or subsystem (default: sftp)
 .TP
 \fB\-o\fR directport=PORT
 directly connect to PORT bypassing ssh
+\fB\-o\fR slave
+communicate over stdin and stdout bypassing network
 .TP
 \fB\-o\fR slave
 communicate over stdin and stdout bypassing network
@@ -282,5 +284,3 @@ SSHFS has been written by Miklos Szeredi <miklos@szeredi.hu>.
 .LP
 This man page was written by Bartosz Fenski <fenio@debian.org> for the
 Debian GNU/Linux distribution (but it may be used by others).
-
-

sshfs.c

@@ -12,6 +12,9 @@
 #include <fuse.h>
 #include <fuse_opt.h>
 #include <fuse_lowlevel.h>
+#ifdef __APPLE__
+#  include <fuse_darwin.h>
+#endif
 #include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -20,7 +23,9 @@
 #include <string.h>
 #include <stdint.h>
 #include <errno.h>
-#include <semaphore.h>
+#ifndef __APPLE__
+#  include <semaphore.h>
+#endif
 #include <pthread.h>
 #include <netdb.h>
 #include <signal.h>
@@ -38,6 +43,11 @@
 #include <pwd.h>
 #include <grp.h>
 #include <limits.h>
+#ifdef __APPLE__
+#  include <strings.h>
+#  include <libgen.h>
+#  include <darwin_compat.h>
+#endif
 
 #include "cache.h"
 
@@ -127,6 +137,10 @@
 
 #define MAX_PASSWORD 1024
 
+#ifdef __APPLE__
+static char sshfs_program_path[PATH_MAX] = { 0 };
+#endif /* __APPLE__ */
+
 struct buffer {
 	uint8_t *p;
 	size_t len;
@@ -243,6 +257,10 @@ struct sshfs {
 	int server_version;
 	unsigned remote_uid;
 	unsigned local_uid;
+#ifdef __APPLE__
+	unsigned remote_gid;
+	unsigned local_gid;
+#endif
 	int remote_uid_detected;
 	unsigned blksize;
 	char *progname;
@@ -739,8 +757,17 @@ static int buf_get_attrs(struct buffer *buf, struct stat *stbuf, int *flagsp)
 		}
 	}
 
+#ifdef __APPLE__
+	if (sshfs.remote_uid_detected) {
+		if (uid == sshfs.remote_uid)
+			uid = sshfs.local_uid;
+		if (gid == sshfs.remote_gid)
+			gid = sshfs.local_gid;
+	}
+#else /* !__APPLE__ */
 	if (sshfs.remote_uid_detected && uid == sshfs.remote_uid)
 		uid = sshfs.local_uid;
+#endif /* __APPLE__ */
 	if (sshfs.idmap == IDMAP_FILE && sshfs.uid_map)
 		if (translate_id(&uid, sshfs.uid_map) == -1)
 			return -EPERM;
@@ -849,11 +876,35 @@ static void ssh_add_arg(const char *arg)
 #ifdef SSH_NODELAY_WORKAROUND
 static int do_ssh_nodelay_workaround(void)
 {
+#ifdef __APPLE__
+	char *oldpreload = getenv("DYLD_INSERT_LIBRARIES");
+#else
 	char *oldpreload = getenv("LD_PRELOAD");
+#endif
 	char *newpreload;
 	char sopath[PATH_MAX];
 	int res;
 
+#ifdef __APPLE__
+	char *sshfs_program_path_base = NULL;
+	if (!sshfs_program_path[0]) {
+		goto nobundle;
+	}
+	sshfs_program_path_base = dirname(sshfs_program_path);
+	if (!sshfs_program_path_base) {
+		goto nobundle;
+	}
+	snprintf(sopath, sizeof(sopath), "%s/%s", sshfs_program_path_base,
+             SSHNODELAY_SO);
+	res = access(sopath, R_OK);
+	if (res == -1) {
+		goto nobundle;
+	}
+	goto pathok;
+
+nobundle:
+#endif /* __APPLE__ */
+
 	snprintf(sopath, sizeof(sopath), "%s/%s", LIBDIR, SSHNODELAY_SO);
 	res = access(sopath, R_OK);
 	if (res == -1) {
@@ -879,15 +930,24 @@ static int do_ssh_nodelay_workaround(void)
 		}
 	}
 
+#ifdef __APPLE__
+pathok:
+#endif
+
 	newpreload = g_strdup_printf("%s%s%s",
 				     oldpreload ? oldpreload : "",
 				     oldpreload ? " " : "",
 				     sopath);
 
+#ifdef __APPLE__
+	if (!newpreload || setenv("DYLD_INSERT_LIBRARIES", newpreload, 1) == -1)
+		fprintf(stderr, "warning: failed set DYLD_INSERT_LIBRARIES for ssh nodelay workaround\n");
+#else /* !__APPLE__ */
 	if (!newpreload || setenv("LD_PRELOAD", newpreload, 1) == -1) {
 		fprintf(stderr, "warning: failed set LD_PRELOAD "
 			"for ssh nodelay workaround\n");
 	}
+#endif /* __APPLE__ */
 	g_free(newpreload);
 	return 0;
 }
@@ -1607,6 +1667,10 @@ static void sftp_detect_uid()
 
 	sshfs.remote_uid = stbuf.st_uid;
 	sshfs.local_uid = getuid();
+#ifdef __APPLE__
+	sshfs.remote_gid = stbuf.st_gid;
+	sshfs.local_gid = getgid();
+#endif
 	sshfs.remote_uid_detected = 1;
 	DEBUG("remote_uid = %i\n", sshfs.remote_uid);
 
@@ -2364,8 +2428,17 @@ static int sshfs_chown(const char *path, uid_t uid, gid_t gid)
 	int err;
 	struct buffer buf;
 
+#ifdef __APPLE__
+	if (sshfs.remote_uid_detected) {
+		if (uid == sshfs.local_uid)
+			uid = sshfs.remote_uid;
+		if (gid == sshfs.local_gid)
+			gid = sshfs.remote_gid;
+	}
+#else /* !__APPLE__ */
 	if (sshfs.remote_uid_detected && uid == sshfs.local_uid)
 		uid = sshfs.remote_uid;
+#endif /* __APPLE__ */
 	if (sshfs.idmap == IDMAP_FILE && sshfs.r_uid_map)
 		if(translate_id(&uid, sshfs.r_uid_map) == -1)
 			return -EPERM;
@@ -3304,8 +3377,15 @@ static void usage(const char *progname)
 "    -o sync_readdir        synchronous readdir\n"
 "    -o sshfs_debug         print some debugging information\n"
 "    -o cache=BOOL          enable caching {yes,no} (default: yes)\n"
+"    -o cache_max_size=N    sets the maximum size of the cache (default: 10000)\n"
 "    -o cache_timeout=N     sets timeout for caches in seconds (default: 20)\n"
 "    -o cache_X_timeout=N   sets timeout for {stat,dir,link} cache\n"
+"    -o cache_clean_interval=N\n"
+"                           sets the interval for automatic cleaning of the\n"
+"                           cache (default: 60)\n"
+"    -o cache_min_clean_interval=N\n"
+"                           sets the interval for forced cleaning of the\n"
+"                           cache if full (default: 5)\n"
 "    -o workaround=LIST     colon separated list of workarounds\n"
 "             none             no workarounds enabled\n"
 "             all              all workarounds enabled\n"
@@ -3316,9 +3396,9 @@ static void usage(const char *progname)
 "             [no]nodelaysrv   set nodelay tcp flag in sshd (default: off)\n"
 "             [no]truncate     fix truncate for old servers (default: off)\n"
 "             [no]buflimit     fix buffer fillup bug in server (default: on)\n"
-"    -o idmap=TYPE          user/group ID mapping, possible types are:\n"
-"             none             no translation of the ID space (default)\n"
-"             user             only translate UID of connecting user\n"
+"    -o idmap=TYPE          user/group ID mapping (default: " IDMAP_DEFAULT ")\n"
+"             none             no translation of the ID space\n"
+"             user             only translate UID/GID of connecting user\n"
 "             file             translate UIDs/GIDs contained in uidfile/gidfile\n"
 "    -o uidfile=FILE        file containing username:remote_uid mappings\n"
 "    -o gidfile=FILE        file containing groupname:remote_gid mappings\n"
@@ -3494,6 +3574,13 @@ static int read_password(void)
 		perror("Failed to allocate locked page for password");
 		return -1;
 	}
+	if (mlock(sshfs.password, size) != 0) {
+		memset(sshfs.password, 0, size);
+		munmap(sshfs.password, size);
+		sshfs.password = NULL;
+		perror("Failed to allocate locked page for password");
+		return -1;
+	}
 
 	/* Don't use fgets() because password might stay in memory */
 	for (n = 0; n < max_password; n++) {
@@ -3819,7 +3906,11 @@ static inline void load_gid_map(void)
 	read_id_map(sshfs.gid_file, &groupname_to_gid, "gid", &sshfs.gid_map, &sshfs.r_gid_map);
 }
 
+#ifdef __APPLE__
+int main(int argc, char *argv[], __unused char *envp[], char **exec_path)
+#else
 int main(int argc, char *argv[])
+#endif
 {
 	int res;
 	struct fuse_args args = FUSE_ARGS_INIT(argc, argv);
@@ -3828,6 +3919,14 @@ int main(int argc, char *argv[])
 	const char *sftp_server;
 	int libver;
 
+#ifdef __APPLE__
+	if (!realpath(*exec_path, sshfs_program_path)) {
+		memset(sshfs_program_path, 0, PATH_MAX);
+	}
+
+	/* Until this gets fixed somewhere else. */
+	g_slice_set_config(G_SLICE_CONFIG_ALWAYS_MALLOC, TRUE);
+#endif /* __APPLE__ */
 	g_thread_init(NULL);
 
 	sshfs.blksize = 4096;
@@ -3836,7 +3935,11 @@ int main(int argc, char *argv[])
 	sshfs.max_write = 32768;
 	sshfs.nodelay_workaround = 1;
 	sshfs.nodelaysrv_workaround = 0;
+#ifdef __APPLE__
+	sshfs.rename_workaround = 1;
+#else
 	sshfs.rename_workaround = 0;
+#endif
 	sshfs.truncate_workaround = 0;
 	sshfs.buflimit_workaround = 1;
 	sshfs.ssh_ver = 2;
@@ -3848,7 +3951,15 @@ int main(int argc, char *argv[])
 	sshfs.delay_connect = 0;
 	sshfs.slave = 0;
 	sshfs.detect_uid = 0;
-	sshfs.idmap = IDMAP_NONE;
+	if (strcmp(IDMAP_DEFAULT, "none") == 0) {
+		sshfs.idmap = IDMAP_NONE;
+	} else if (strcmp(IDMAP_DEFAULT, "user") == 0) {
+		sshfs.idmap = IDMAP_USER;
+	} else {
+		fprintf(stderr, "bad idmap default value built into sshfs; "
+		    "assuming none (bad logic in configure script?)\n");
+		sshfs.idmap = IDMAP_NONE;
+	}
 	sshfs.nomap = NOMAP_ERROR;
 	ssh_add_arg("ssh");
 	ssh_add_arg("-x");

sshnodelay.c

@@ -5,14 +5,55 @@
 #include <netinet/in.h>
 #include <netinet/tcp.h>
 
-int connect(int sock, const struct sockaddr *addr, socklen_t addrlen)
+/* Wrapper around connect(2) to explicitly set TCP_NODELAY. */
+static int nodelay_connect(
+    int (*real_connect)(int, const struct sockaddr *, socklen_t),
+    int sock, const struct sockaddr *addr, socklen_t addrlen)
 {
-	int (*next_connect)(int, const struct sockaddr *, socklen_t) =
-		dlsym(RTLD_NEXT, "connect");
-	int res = next_connect(sock, addr, addrlen);
+	int res = real_connect(sock, addr, addrlen);
 	if (!res && addr->sa_family == AF_INET) {
 		int opt = 1;
 		setsockopt(sock, IPPROTO_TCP, TCP_NODELAY, &opt, sizeof(opt));
 	}
 	return res;
 }
+
+#if __APPLE__
+
+/* OS X does not have LD_PRELOAD but has DYLD_INSERT_LIBRARIES.  The right
+ * environment variable is set by sshfs.c when attempting to load the
+ * sshnodelay workaround.
+ *
+ * However, things are not that simple: DYLD_INSERT_LIBRARIES does not
+ * behave exactly like LD_PRELOAD.  Instead, the dyld dynamic linker will
+ * look for __DATA __interpose sections on the libraries given via the
+ * DYLD_INSERT_LIBRARIES variable.  The contents of this section are pairs
+ * of replacement functions and functions to be replaced, respectively.
+ * Prepare such section here. */
+
+int custom_connect(int sock, const struct sockaddr *addr, socklen_t addrlen);
+
+typedef struct interpose_s {
+	void *new_func;
+	void *orig_func;
+} interpose_t;
+
+static const interpose_t interposers[] \
+	__attribute__ ((section("__DATA, __interpose"))) = {
+	{ (void *)custom_connect,  (void *)connect  },
+};
+
+int custom_connect(int sock, const struct sockaddr *addr, socklen_t addrlen)
+{
+	return nodelay_connect(connect, sock, addr, addrlen);
+}
+
+#else /* !__APPLE__ */
+
+int connect(int sock, const struct sockaddr *addr, socklen_t addrlen)
+{
+	return nodelay_connect(dlsym(RTLD_NEXT, "connect"),
+	                       sock, addr, addrlen);
+}
+
+#endif /* !__APPLE__ */