libfuse

• Added support for FUSE_LSEEK operation which can be used to report holes in sparse files.

• Added UFSD to whitelist (so users can now mount FUSE filesystems on mountpoints within UFSD filesystems).
• Added custom log message handler function support so that libfuse applications can direct messages to syslog(3) or other logging systems. stderr remains the default. See fuse_log.h for the new API.

• The init script is now installed to /etc/ rather than /usr/local/etc by default.
• Minor bugfixes

• Fixed version number (release 3.6.0 was shipped with a declared version of 3.0.0).

• Added a new example (passthrough_hp). The functionality is similar to passthrough_ll, but the implementation focuses on performance and correctness rather than simplicity.
• Added support for fuse kernel feature max_pages which allows to increase the maximum number of pages that can be used per request. This feature was introduced in kernel 4.20. max_pages is set based on the value in max_write. By default max_write will be 1MiB now for kernels that support max_pages. If you want smaller buffers or writes you have to set max_write manually.

• Changed ioctl commands to "unsigned int" in order to support commands which do not fit into a signed int. Commands issued by applications are still truncated to 32 bits.
• Added SMB2 to whitelist (so users can now mount FUSE filesystems on mountpoints within SMB 2.0 filesystems).
• Added a new cache_readdir flag to fuse_file_info to enable caching of readdir results. Supported by kernels 4.20 and newer.
• Add support and documentation for FUSE_CAP_NO_OPENDIR_SUPPORT.

• Fixed a memory leak in examples/passthrough_ll.c.
• Added OpenAFS and HFS+ to whitelist (so users can now mount FUSE filesystems within these filesystems).
• Documentation improvements.

Note: This is a maintenance release from the 2.9 branch of libfuse. Users are encouraged to transition to the
actively developed libfuse 3.x.

• Added OpenAFS to whitelist (so users can now mount FUSE filesystems on mountpoints within OpenAFS filesystems).
• Added a test of seekdir to test_syscalls.
• Fixed readdir bug when non-zero offsets are given to filler and the filesystem client, after reading a whole directory, re-reads it from a non-zero offset e. g. by calling seekdir followed by readdir.

• The examples/passthrough_ll.c example filesystem has been significantly extended.
• Support for copy_file_range has been added.
• Build system updates for non-Linux systems.

• The auto_unmount mode now works correctly in combination with
  autofs.

• The FUSE_CAP_READDIRPLUS_AUTO capability is no longer enabled by
  default unless the file system defines both a readdir() and a
  readdirplus() handler.

• The description of the FUSE_CAP_READDIRPLUS_AUTO flag has been
  improved.

• Allow open /dev/fuse file descriptors to be passed via mountpoints of the
  special format /dev/fd/%u. This allows mounting to be handled by the parent
  so the FUSE filesystem process can run fully unprivileged.

• Add a drop_privileges option to mount.fuse3 which causes it to open
  /dev/fuse and mount the file system itself, then run the FUSE file
  filesystem fully unprivileged and unable to re-acquire privilege via setuid,
  fscaps, etc.

• Documented under which conditions the fuse_lowlevel_notify_*
  functions may block.

• The fuse_main() function now returns more fine-grained error codes.
• FUSE filesystems may now be mounted on mountpoint within bcachefs, aufs and FAT filesystems.
• libfuse may now be used as a Meson subproject.
• Fix a few low-impact memory leaks.
• The fuse.conf file is no longer looked for in /etc, but in the sysconfdir directory (which can be set with meson configure). By default, the location is thus /usr/local/etc/fuse.conf.

Changes:

• SECURITY UPDATE: In previous versions of libfuse it was possible to for unprivileged users to specify the allow_other option even when this was forbidden in /etc/fuse.conf. The vulnerability is present only on systems where SELinux is active (including in permissive mode).
• libfuse no longer segfaults when fuse_interrupted() is called outside the event loop.
• The fusermount binary has been hardened in several ways to reduce potential attack surface. Most importantly, mountpoints and mount options must now match a hard-coded whitelist. It is expected that this whitelist covers all regular use-cases.
• Fixed rename deadlock on FreeBSD.

Changes:

• SECURITY UPDATE: In previous versions of libfuse it was possible to for unprivileged users to specify the allow_other option even when this was forbidden in /etc/fuse.conf. The vulnerability is present only on systems where SELinux is active (including in permissive mode).
• The fusermount binary has been hardened in several ways to reduce potential attack surface. Most importantly, mountpoints and mount options must now match a hard-coded whitelist. It is expected that this whitelist covers all regular use-cases.
• Added a test of seekdir to test_syscalls.
• Fixed readdir bug when non-zero offsets are given to filler and the filesystem client, after reading a whole directory, re-reads it from a non-zero offset e. g. by calling seekdir followed by readdir.

• Fixed rename deadlock on FreeBSD.
• Small bugfixes.

• Fixed a number of compiler warnings.

• Added example fuse.conf file.
• Added "support" for -o nofail mount option (the option is accepted and ignored).
• Various small bugfixes.

• Various small bugfixes.

• Support for building with autotools has been dropped.

• Added new fuse_invalidate_path() routine for cache invalidation from the high-level FUSE API, along with an example and tests.

• There's a new printcap example that can be used to determine the capabilities of the running kernel.

• fuse_loop_mt() now returns the minus the actual errno if there was an error (instead of just -1).

• fuse_loop() no longer returns a positive value if the filesystem loop was terminated without errors or signals.

• Improved documentation of fuse_lowlevel_notify_* functions.

• fuse_lowlevel_notify_inval_inode() and fuse_lowlevel_notify_inval_entry() now return -ENOSYS instead of an undefined error if the function is not supported by the kernel.

• Documented the special meaning of the zero offset for the fuse_fill_dir_t function.

• The passthrough_fh example now works under FreeBSD.

• libfuse can now be build without libiconv.

• Fixed support for FUSE_CAP_POSIX_ACL: setting this capability flag had no effect in the previous versions of libfuse 3.x; now ACLs should actually work.

• Fixed a number of compilation problems under FreeBSD.

• Fixed installation directory for udev rules.

• Fixed compilation with LTO.

• Documentation: clarified how filesystems are supposed to process open() and create() flags (see include/fuse_lowlevel.h).

• Fixed a compilation problem of the passthrough_ll example on 32 bit systems (wrong check and wrong error message).

• pkg-config is now used to determine the proper directory for udev rules.

• Fixed a symbol versioning problem that resulted in very strange failures (segfaults, unexpected behavior) in different situations.

• Fixed a test failure when /tmp is on btrfs.

• Incorporated several patches from the FreeBSD port. libfuse should now compile under FreeBSD without the need for patches.

• The passthrough_ll example now supports writeback caching.

• Added new fuse_lib_help() function. File-systems that previously passed a --help option to fuse_new() must now process the --help option internally and call fuse_lib_help() to print the help for generic FUSE options.
• Fixed description of the fuse_conn_info->time_gran. The default value of zero actually corresponds to full nanosecond resolution, not one second resolution.
• The init script is now installed into the right location ($DESTDIR/etc/init.d rather than $prefix/$sysconfdir/init.d)
• The example/passthrough_ll filesystem now supports creating and writing to files.
• fuse_main() / fuse_remove_signal_handlers(): do not reset SIGPIPE handler to SIG_DFL if it was not set by us.
• Documented the RENAME_EXCHANGE and RENAME_NOREPLACE flags that may be passed to the rename handler of both the high- and low-level API. Filesystem authors are strongly encouraged to check that these flags are handled correctly.