Source Tree:

qfuse
   |---trunk
       |---setenv.sh 
       |---lib
             |---fuse, ublio, uuid
       |---utils 
             |---n 
                 |---ntfs-3g, ntfsprogs
             |---e
                  |---ext2fuse, e2fsprogs
             |---z
                  |---zfsfuse
       |---qa
            |---pjd-fstest
   |---vendor
       |---trunk
           |---lib
                 |---fuse-2.7.4, libublio-20070103
           |---utils
               |---n
                    |---ntfs-3g-2009.4.4, ntfsprogs-2.0.0
               |---e
                    |---ext2fuse-src-0.8.1, e2fsprogs-1.41.5
               |---z
                    |---zfsfuse
            |---qa
                 |---pjd-fstest-20090130-RC

How to build libfuse and other fs daemon?

1. Enter qfuse/trunk and run ". ./setenv.sh"
   
2. cd lib/fuse and run "make install" or "CPULIST=x86 make install" if you only want to build libfuse for x86 target.
   This will install libfuse to qfuse/trunk/install/cpudir/lib and fuse headers to qfuse/trunk/install/include.
   
3. cd lib/ublio and make install./* default ntfs-3g needs ublio but you can change*/
4. cd utils/n/ntfs-3g and run "make" or "make install". This will install ntfs-3g and libntfs to qfuse/trunk/install/cpudir too.
   

How to run filesystem daemon?

1. copy libfuse to /usr/local/lib
   
2. copy filesystem daemon's own library to /usr/local/lib, for example: ntfs-3g needs libntfs-3g.
   
3. copy filesystem daemon executable file to /usr/local/bin
   
4. run the fs daemon with arguments. For example: ntfs-3g -o debug,no_detach /dev/hd1t7 /ntfs
   

How to port other filesystem to QNX?

1. Get filesystem source code
2. Write a run.sh to try on QNX host. If configure complains something like you have to rerun automake or autoconf, then you have to install these on QNX(pkgsrc QNX port project you could try) or you can try on Linux host with QNX development tools installed. There is run.sh example in ntfs-3g.
3. If previous step succeeded then you can take a look at ext2fuse or ntfs-3g to see what you need to modify. Most likely you have to modify getopt because QNX doesn't support "--" long option yet.
4. If the filesystem registers FUSE lowlevel API then you have to modify its *_read function to use qfuse's buffer. There is example in libfuse or ext2fuse's op_read function. Because ntfs-3g registers highlevel API it doesn't care, necessary modification is in libfuse's default lowlevel API.
5. Try to port it to QNX' recursive compiling environment with multi-platform support.

How to test?

1. There is a posix filesystem test suite I ported from FreeBSD in trunk/qa/pjd-fstest.
   Please read README file to see how to run it.

Benchmark

/* below is test no fs daemon using ublio library*/
$ uname -a
QNX localhost 6.3.2 2006/03/16-14:19:50EST x86pc x86
#ext2fuse /home/ext2_256M.fs /ext2
#ntfs-3g /home/ntfs_256M.fs /ntfs
#mount -t qnx4 /home/qnx4_256M.fs /qnx4

/* default 8k record testing*/
# rw -t 128M -k /ext2/rw.tmp
Sequential File Write/Read Benchmark
OS:      QNX 6.3.2 x86pc
Filesys: disk, fsys
Config:  128MiB file, 8KiB record, fd, malloc
Create:      10 msec
Write:     7783 msec,  475 usec/write(),  89% CPU, 16.83 MiB/sec
Read:      9538 msec,  582 usec/read(),   72% CPU, 13.74 MiB/sec
# rw -t 128M -k /ntfs/rw.tmp
Sequential File Write/Read Benchmark
OS:      QNX 6.3.2 x86pc
Filesys: disk, fsys
Config:  128MiB file, 8KiB record, fd, malloc
Create:       4 msec
Write:     8774 msec,  535 usec/write(), 100% CPU, 14.93 MiB/sec
Read:      7164 msec,  437 usec/read(),   61% CPU, 18.29 MiB/sec
# rw -t 128M -k /qnx4/rw.tmp
Sequential File Write/Read Benchmark
OS:      QNX 6.3.2 x86pc
Filesys: ST340014A, UDMA5, qnx4, 1% full
Config:  128MiB file, 8KiB record, fd, fsync, malloc
Create:      11 msec
Write:     5624 msec,  343 usec/write(),  43% CPU, 23.30 MiB/sec
Read:      3812 msec,  232 usec/read(),   58% CPU, 34.38 MiB/sec

FAQ:

1. Can I compile them on Linux or Windows HOST with QNX development toools installed?
   Answer: You can run ". ./run.sh" to make it on Linux or Windows host and there is an example in ntfs-3g or fuse. And you'd better not to mix QNX' makefile and GNU's makefile together so I create nto-cpu-endian for QNX multi arch compiling and cpu-endian for GNU multi arch compiling so put run.sh to corresponding directory. If want to switch between GNU makefile.gnu and QNX makefile.qnx, you can run "make -f Makefile.gnu distclean" to clean it with GNU makefile.

Resources:

1. FUSE project
   
2. ntfs-3g project