Virtual Filesystem in Perl with Fuse - perl

Question

Anybody help me make a virtual file system in Perl.
Very simple, 2 depth level, as
/subdir
subdir-l2
file2.txt
/file1.txt
I try use Fuse.pm, but not understand how create subdir level. I create %files hash, and if go to subdir, recreate it with new records. It's for test only.
#!/usr/bin/env perl
use strict;
use warnings;
use utf8;
use Fuse;
use POSIX qw(ENOENT EISDIR EINVAL);
my (%files) = (
'.' => {
type => 0040,
mode => 0755,
ctime => 1490603721
},
subdir => {
type => 0040,
mode => 0755,
ctime => 1490603721
},
"file1.txt" => {
type => 0100,
mode => 0755,
ctime => 1490603721
}
);
sub filename_fixup {
my ($file) = shift;
$file =~ s,^/,,;
$file = '.' unless length($file);
return $file;
}
sub getdir {
my $tmp = shift;
if ($tmp eq '/') {
return (keys %files),0;
} else {
(%files) = (
'.' => {
type => 0040,
mode => 0755,
ctime => 1490603721
},
# /subdir/subdir-l2
"subdir-l2" => {
type => 0040,
mode => 0755,
ctime => 1490603721
} ,
# /subdir/a-l2.file
"file2.txt" => {
cont => "File 'al2'.\n",
type => 0100,
mode => 0755,
ctime => 1490603721
}
);
return (keys %files),0;
}
}
sub getattr {
my ($file) = filename_fixup(shift);
$file =~ s,^/,,;
$file = '.' unless length($file);
return -ENOENT() unless exists($files{$file});
my ($size) = exists($files{$file}{cont}) ? length($files{$file}{cont}) : 0;
$size = $files{$file}{size} if exists $files{$file}{size};
my ($modes) = ($files{$file}{type}<<9) + $files{$file}{mode};
my ($dev, $ino, $rdev, $blocks, $gid, $uid, $nlink, $blksize) = (0,0,0,1,0,0,1,1024);
my ($atime, $ctime, $mtime);
$atime = $ctime = $mtime = $files{$file}{ctime};
return ($dev,$ino,$modes,$nlink,$uid,$gid,$rdev,$size,$atime,$mtime,$ctime,$blksize,$blocks);
}
Fuse::main(
mountpoint => "/tmp/123",
getdir => \&getdir,
getattr => \&getattr,
);
one level mount fine, but if go to deeper i get
?????????? ? ? ? ? ? file2.txt
?????????? ? ? ? ? ? subdir-l2

Answer 1

I'm really not a regular user of the Fuse module, neither of FUSE system. Tinkered with this issue out of pure curiosity. Thus, although I can't explain in very much details how to use the plain Fuse module to achieve your goal, I have a working code that does create the wanted filesystem (at least on my system, and seems that it is capable of creating any arbitrary filesystem tree), and I can explain how I got this code working.
So first of all I discovered the Fuse::Simple module on CPAN.
Its SYNOPSIS shows that it provides a really simple API to the Fuse module for creating arbitrary filesystems from a hash structure. Its source code isn't that huge, so I just created 'listing.pl' script file and copied there most of the functions (except fserr that caused a Modification of a read-only value exception), put the main sub contents out, so they will be the main script's flow, hardcoded the filesystem structure ($fs var), and made some little adjustments here and there (like declare vars with my to prevent exceptions), and finally got the filesystem mounted, with all directories listed and files readable. So this is the code I got at last:
#!/usr/bin/env perl
use strict;
use warnings;
use diagnostics;
use Carp;
use Fuse;
use Errno qw(:POSIX); # ENOENT EISDIR etc
use Fcntl qw(:DEFAULT :mode); # S_IFREG S_IFDIR, O_SYNC O_LARGEFILE etc.
use Switch;
my $debug = 0;
my %codecache = ();
my $ctime = time();
my $uid = $>;
my $gid = $) + 0;
my $fs = {
"file1.txt" => "File 1 contents",
"subdir" => {
"subdir-l2" => {},
"file2.txt" => "File 2 contents"
}
};
# some default args
my %args = (
"mountpoint" => "listing",
"debug" => $debug,
"fuse_debug" => 0,
"threaded" => 0,
"/" => $fs
);
# the default subs
my %fs_subs = (
"chmod" => \&fs_not_imp,
"chown" => \&fs_not_imp,
"flush" => \&fs_flush,
"fsync" => \&fs_not_imp,
"getattr" => \&fs_getattr,
"getdir" => \&fs_getdir,
"getxattr" => \&fs_not_imp,
"link" => \&fs_not_imp,
"listxattr" => \&fs_not_imp,
"mkdir" => \&fs_not_imp,
"mknod" => \&fs_not_imp,
"open" => \&fs_open,
"read" => \&fs_read,
"readlink" => \&fs_readlink,
"release" => \&fs_release,
"removexattr" => \&fs_not_imp,
"rmdir" => \&fs_not_imp,
"rename" => \&fs_not_imp,
"setxattr" => \&fs_not_imp,
"statfs" => \&fs_statfs,
"symlink" => \&fs_not_imp,
"truncate" => \&fs_truncate,
"unlink" => \&fs_not_imp,
"utime" => sub{return 0},
"write" => \&fs_write,
);
# except extract these ones back out.
$debug = delete $args{"debug"};
$args{"debug"} = delete( $args{"fuse_debug"} ) || 0;
delete $args{"/"};
# add the functions, if not already defined.
# wrap in debugger if debug is set.
for my $name (keys %fs_subs) {
my $sub = $fs_subs{$name};
# $sub = wrap($sub, $name) if $debug;
$args{$name} ||= $sub;
}
Fuse::main(%args);
sub fetch {
my ($path, #args) = #_;
my $obj = $fs;
for my $elem (split '/', $path) {
next if $elem eq ""; # skip empty // and before first /
$obj = runcode($obj); # if there's anything to run
# the dir we're changing into must be a hash (dir)
return ENOTDIR() unless ref($obj) eq "HASH";
# note that ENOENT and undef are NOT the same thing!
return ENOENT() unless exists $obj->{$elem};
$obj = $obj->{$elem};
}
return runcode($obj, #args);
}
sub runcode {
my ($obj, #args) = #_;
while (ref($obj) eq "CODE") {
my $old = $obj;
if (#args) { # run with these args. don't cache
delete $codecache{$old};
print "running $obj(",quoted(#args),") NO CACHE\n" if $debug;
$obj = saferun($obj, #args);
} elsif (exists $codecache{$obj}) { # found in cache
print "got cached $obj\n" if $debug;
$obj = $codecache{$obj}; # could be undef, or an error, BTW
} else {
print "running $obj() to cache\n" if $debug;
$obj = $codecache{$old} = saferun($obj);
}
if (ref($obj) eq "NOCACHE") {
print "returned a nocache() value - flushing\n" if $debug;
delete $codecache{$old};
$obj = $$obj;
}
print "returning ",ref($obj)," ",
defined($obj) ? $obj : "undef",
"\n" if $debug;
}
return $obj;
}
sub saferun {
my ($sub, #args) = #_;
my $ret = eval { &$sub(#args) };
my $died = $#;
if (ref($died)) {
print "+++ Error $$died\n" if ref($died) eq "ERROR";
return $died;
} elsif ($died) {
print "+++ $died\n";
# stale file handle? moreorless?
return ESTALE();
}
return $ret;
}
sub nocache {
return bless(\ shift, "NOCACHE"); # yup, utter abuse of bless :-)
}
sub dump_open_flags {
my $flags = shift;
printf " flags: 0%o = (", $flags;
for my $bits (
[ O_ACCMODE(), O_RDONLY(), "O_RDONLY" ],
[ O_ACCMODE(), O_WRONLY(), "O_WRONLY" ],
[ O_ACCMODE(), O_RDWR(), "O_RDWR" ],
[ O_APPEND(), O_APPEND(), "|O_APPEND" ],
[ O_NONBLOCK(), O_NONBLOCK(), "|O_NONBLOCK" ],
[ O_SYNC(), O_SYNC(), "|O_SYNC" ],
[ O_DIRECT(), O_DIRECT(), "|O_DIRECT" ],
[ O_LARGEFILE(), O_LARGEFILE(), "|O_LARGEFILE" ],
[ O_NOFOLLOW(), O_NOFOLLOW(), "|O_NOFOLLOW" ],
) {
my ($mask, $flag, $name) = #$bits;
if (($flags & $mask) == $flag) {
$flags -= $flag;
print $name;
}
}
printf "| 0%o !!!", $flags if $flags;
print ")\n";
}
sub accessor {
my $var_ref = shift;
croak "accessor() requires a reference to a scalar var\n"
unless defined($var_ref) && ref($var_ref) eq "SCALAR";
return sub {
my $new = shift;
$$var_ref = $new if defined($new);
return $$var_ref;
}
}
sub fs_not_imp { return -ENOSYS() }
sub fs_flush {
# we're passed a path, but finding my coderef stuff from a path
# is a bit of a 'mare. flush the lot, won't hurt TOO much.
print "Flushing\n" if $debug;
%codecache = ();
return 0;
}
sub easy_getattr {
my ($mode, $size) = #_;
return (
0, 0, # $dev, $ino,
$mode,
1, # $nlink, see fuse.sourceforge.net/wiki/index.php/FAQ
$uid, $gid, # $uid, $gid,
0, # $rdev,
$size, # $size,
$ctime, $ctime, $ctime, # actually $atime, $mtime, $ctime,
1024, 1, # $blksize, $blocks,
);
}
sub fs_getattr {
my $path = shift;
my $obj = fetch($path);
# undef doesn't actually mean "file not found", it could be a coderef
# file-sub which has returned undef.
return easy_getattr(S_IFREG | 0200, 0) unless defined($obj);
switch (ref($obj)) {
case "ERROR" { # this is an error to be returned.
return -$$obj;
}
case "" { # this isn't a ref, it's a real string "file"
return easy_getattr(S_IFREG | 0644, length($obj));
}
# case "CODE" should never happen - already been run by fetch()
case "HASH" { # this is a directory hash
return easy_getattr(S_IFDIR | 0755, 1);
}
case "SCALAR" { # this is a scalar ref. we use these for symlinks.
return easy_getattr(S_IFLNK | 0777, 1);
}
else { # what the hell is this file?!?
print "+++ What on earth is ",ref($obj)," $path ?\n";
return easy_getattr(S_IFREG | 0000, 0);
}
}
}
sub fs_getdir {
my $obj = fetch(shift);
return -$$obj if ref($obj) eq "ERROR"; # THINK this is a good idea.
return -ENOENT() unless ref($obj) eq "HASH";
return (".", "..", sort(keys %$obj), 0);
}
sub fs_open {
# doesn't really need to open, just needs to check.
my $obj = fetch(shift);
my $flags = shift;
dump_open_flags($flags) if $debug;
# if it's undefined, and we're not writing to it, return an error
return -EBADF() unless defined($obj) or ($flags & O_ACCMODE());
switch (ref($obj)) {
case "ERROR" { return -$$obj; }
case "" { return 0 } # this is a real string "file"
case "HASH" { return -EISDIR(); } # this is a directory hash
else { return -ENOSYS(); } # what the hell is this file?!?
}
}
sub fs_read {
my $obj = fetch(shift);
my $size = shift;
my $off = shift;
return -ENOENT() unless defined($obj);
return -$$obj if ref($obj) eq "ERROR";
# any other types of refs are probably bad
return -ENOENT() if ref($obj);
if ($off > length($obj)) {
return -EINVAL();
} elsif ($off == length($obj)) {
return 0; # EOF
}
return substr($obj, $off, $size);
}
sub fs_readlink {
my $obj = fetch(shift);
return -$$obj if ref($obj) eq "ERROR";
return -EINVAL() unless ref($obj) eq "SCALAR";
return $$obj;
}
sub fs_release {
my ($path, $flags) = #_;
dump_open_flags($flags) if $debug;
return 0;
}
sub fs_statfs {
return (
255, # $namelen,
1,1, # $files, $files_free,
1,1, # $blocks, $blocks_avail, # 0,0 seems to hide it from df?
2, # $blocksize,
);
}
sub fs_truncate {
my $obj = fetch(shift, ""); # run anything to set it to ""
return -$$obj if ref($obj) eq "ERROR";
return 0;
}
sub fs_write {
my ($path, $buf, $off) = #_;
my $obj = fetch($path, $buf, $off); # this runs the coderefs!
return -$$obj if ref($obj) eq "ERROR";
return length($buf);
}
Final word: I didn't try to use the module itself (it's not listed in my distro package repository, and I was too lazy (sorry) to install it by cpanm or other way). But I think that if I'll have to just use FUSE with Perl, I'll probably just use Fuse::Simple instead of Fuse, maybe forking it. I'd use plain Fuse only for my academic research, I think...
Hope this helps.