tlinden/crypt--pwsafe3

Perl module to Read and write Passwordsafe v3 files

Tom von Dein

I am a network and systems engineer currently working at FI-TS and this is my github profile.

My current most liked language is go, previously this was perl. I used to program in a bunch of other languages like C, python or ruby. I do almost all my work with emacs, for which I wrote a couple of emacs modes.

The projects I am currently working on are:

• Golsky: a golang implementation of the game of life, supports RLE patterns, saving and loading, drawing, zooming and is controllable using key bindings.
• Openquell: a 2D puzzle game using ebitengine, arche and LDTK. Not (yet) opensource. Play it online
• kageviewer: a live viewer for shaders written in Kage, similar to glslviewer
• gfn: a cli tool to generate fantasy names for games and stories
• kleingebäck: a tool to backup kleinanzeigen.de ads.
• yadu: a go log/slog handler using a mix of log lines and yaml.
• rpnc: a commandline and interactive programmable reverse polish notation calculator
• tablizer: a console table manipulation tool
• ephemerup: a ephemeral file upload system with rest API and commandline cli

The most fun projects I did in the past:

• pretty curved privacy: a tool to encrypt and decrypt data at rest using eliptic curves for the commandline.
• diceware: a diceware password generator, which I use regularly.
• udpxd: a udp proxy supporting v4/v6 conversion.
• twenty4: a self-made cryptographic algorithm I created just for fun.

There are a couple of perl modules I am still maintaining:

• Config::General: a widely used configuration module.
• Data::Validate::Struct: can be used to validate recursive data structures.
• Crypt::PWSafe3: a module to read and write passwordsafe.net files.
• leo: a commandline interface to https://dict.leo.org.

I also maintain (and use daily) the FreeBSD jail management tool jaildk. Another fun shell script is goupdater, which I use to install and update all those go binaries which are avaliable as direct downloads.

And finally these are some emacs modes I maintain (and mostly use):

• novel-mode: a simple mode to use emacs as a comfortable screen reader.
• mark-copy-yank-things-mode: Mark, Copy and Yank Things Emacs Minor Mode.
• autoscratch: Automatically switch major mode from scratch.
• viking-mode: Kill first, ask later - an emacs mode for killing things quickly.

My current emacs configuration can be found here: dot-emacs.

crypt--pwsafe3's Issues

Some CPAN testers reports for your consideration

I got those for Passwd::Keyring::PwSafe3 but they seem to be caused by your module. Not sure whether they are sufficient to diagnose but mayhaps…

1. Windows problems with temporary file (both practically identical)

http://www.cpantesters.org/cpan/report/c6375bfe-6c15-1014-b28c-64d78ddf5304
http://www.cpantesters.org/cpan/report/3df707a4-6bfe-1014-8823-09d9f6238fb6

(before going deeper: I'd consider using File::Temp instead of your current method, mayhaps the name turns too long, or there is some collision)

2. Integrity failure

http://www.cpantesters.org/cpan/report/0df0824a-9721-11e4-b221-9e126cbd7f71

(this is quite ugly, looks like first save crashed but later on things worked properly)

I believe Crypt::CBC has removed the blocksize option from new

I saw this when running a unit test. I believe the change happened between Crypt::CBC 2.33 and Crypt::CBC 3.00 from 2021.
It's possible I'm doing something weird to trigger this that other people don't see. Or it's possible that not many people have upgraded Crypt::CBC recently.

I've created a bug report for Crypt::CBC with the same information. https://rt.cpan.org/Ticket/Display.html?id=145836

Error: 'blocksize' is not a recognized argument at /home/runner/work/tel/tel/local/lib/perl5/Crypt/CBC.pm line 312.
 at /home/runner/work/tel/tel/local/lib/perl5/Crypt/PWSafe3.pm line 260.

Error while reading database with non-ascii (Polish) characters

I created small passwordsafe database (using passwordsafe GUI) with various Polish characters in group or user names. When I was previously using the module (~6 months ago) database was read just fine by Crypt::PWSafe3, unfortunately, no more. Now I get:

'<' allowed only after types sSiIlLqQjJfFdDpP( in unpack at /home/marcink/perl_patches/Crypt--PWSafe3/blib/lib/Crypt/PWSafe3/Field.pm line 79.

You can pull test-case (sample database and test file which reads it) from
Mekk@cfbfcd6

Note: I test on Ubuntu 13.04 Linux, perl 5.14.2

t/run.t started to fail

See subject. Some pointers:

• date-sorted list of reports: http://matrix.cpantesters.org/?dist=Crypt-PWSafe3%201.22;reports=1#sl=7,1
• a random fail report: http://www.cpantesters.org/cpan/report/f6581ee4-8414-11eb-a434-19407f78d25b

Failed install from CPAN (new file missing in manifest)

Attempt to

cpan Crypt::PWSafe3

ends up in

Running make for T/TL/TLINDEN/Crypt-PWSafe3-1.13.tar.gz
(...)
t/run.t .. 1/? 
   Failed test 'use Crypt::PWSafe3;'
   at t/run.t line 18.
   Tried to use 'Crypt::PWSafe3'.
   Error:  Can't locate Crypt/PWSafe3/PasswordPolicy.pm in @INC 

Likely reason: this file (PasswordPolicy.pm) is missing in the MANIFEST (so was not packaged to .tar.gz and was not distributed).

random usage in stretchpw()

I was investigating why something I wrote was hanging after a few runs and I discoverted it was because I was running out of entropy. I further tracked this to using Crypt::Random and saw that Bytes::Random::Secure seems to be the better option.

However, I was still confused why it needed to generate a random number just to open an existing file. I looked at how things worked and saw that read() was calling stretchpw() which had this line near the end:

$passwd = $this->random(64);

This isn't returned from the sub and doesn't seemed to be referenced anywhere. Commenting the line out seems to fix my issue, but might create problems in other places I don't know about. If the intention is to overwrite the plaintext password in memory can you just zero it instead of using random?

I think my only alternative for my program would be to require Bytes::Random::Secure if they use Crypt::PWSafe3, since Crypt::Random can block silently in places without the user knowing why.

A few problems from cpantesters reports

I started distributing my wrapper around Crypt::PWSafe3 ( http://search.cpan.org/~mekk/Passwd-Keyring-PWSafe3-0.20/lib/Passwd/Keyring/PWSafe3.pm - sources at https://bitbucket.org/Mekk/perl-keyring-pwsafe3 ) together with some test suite, and I got some cpantesters reports which seem related more to your module than mine, but could also pinpoint my wrong usage of your code. I'd be grateful if you could take a look and comment on those issues.

1. File integrity errors lik e http://www.cpantesters.org/cpan/report/b1ace1e2-33f2-11e2-8219-b9bba290f8f5 (I got a few such reports)

I distribute some sample password file - created with pwsafe GUI on my Linux - with my tests. In many cases it is used properly, but here we have

File integrity check failed (File integrity check failed at /tmp/loop_over_bdir-TsD8Db/Passwd-Keyring-PWSafe3-0.2001-SjyNqH/blib/lib/Passwd/Keyring/PWSafe3.pm line 277.
) at /tmp/loop_over_bdir-TsD8Db/Passwd-Keyring-PWSafe3-0.2001-SjyNqH/blib/lib/Passwd/Keyring/PWSafe3.pm line 277.

I am not sure whether it happens on first read, or on save after updates. My first doubt is whether it is at all sensible to distribute example file (= expect the format to be cross-platform).

2. Ugly undef warnings http://www.cpantesters.org/cpan/report/ecd74cfc-3452-11e2-a056-9b38e258596b

Leaving apart the exact meaning of tests, do you have any idea what are those undef warnings about?

3. Unpack errors http://www.cpantesters.org/cpan/report/d3f7c37a-3410-11e2-8c57-bbca67d21516

Ditto.

I am not sure whether those are problems in your code at all, but I lack the knowledge to interpret them by myself, so would be glad if you could make some comments.

Suspcious use of Crypt::Random

The following code block looks questionable, after discovering that the default of Crypt::Random is to ensure the highest bit of every generated number is 1.

Which in practice means:

my $i = makerandom( Size => 256, Strength => 1 );

$i can never be an integer lower than 2 ** 255

eg:

use Crypt::Random qw( makerandom );

my $yes = 0; 
my $no = 0;

for ( 1..10000 ) {
  my $bits = makerandom( Size => 256, Strength => 1 );
  if ( $bits < 2 ** 255 ) {
    $yes++
  } else {
    $no++;
  }
}
printf qq[yes: %d, no: %d\n], $yes, $no;
# yes: 0, no: 10000

https://metacpan.org/pod/Crypt::Random#Uniform-0-||-1

When Uniform => 1, the distribution is from 0 .. 2**Size , not 2**(Size-1)+1 .. 2**Size

use Crypt::Random qw( makerandom );

my $yes = 0; 
my $no = 0;

for ( 1..10000 ) {
  my $bits = makerandom( Size => 256, Strength => 1, Uniform => 1);
  if ( $bits < 2 ** 255 ) {
    $yes++
  } else {
    $no++;
  }
}
printf qq[yes: %d, no: %d\n], $yes, $no;
# yes: 5079, no: 4921

It also seems odd that you call the return value of this function $bytes, when that function returns an integer, and doing substr() on an integer is uh. Weird.

https://metacpan.org/source/TLINDEN/Crypt-PWSafe3-1.22/lib/Crypt/PWSafe3.pm#L60-64