Philipp Gesang [Fri, 31 Mar 2017 13:01:20 +0000]
pass encryption context to tarfile
This supersedes the individual parameters.
Philipp Gesang [Fri, 31 Mar 2017 12:31:19 +0000]
do not require parameter version with each encrypted object
When encrypting, stick to the parameter version specified in the
ctor. Despite the format allowing for on-the-fly adjustment of
encryption parameters, there is no real world scenario yet in
which this might be desired. Thus, remove this prerequisite as it
only encumbers the signature of ``.next()`` with cruft.
Philipp Gesang [Tue, 28 Mar 2017 15:44:31 +0000]
first draft for making the encryption layer independent
WIP.
In Deltatar, we cannot use the ctor itself to set up the
encryption because it is neutral wrt. reading / writing.
Only once one of the entry points:
- ``.list_backup()``,
- ``.create_full_backup()``, ``.create_diff_backup()``, or
- ``.restore_backup()``.
are invoked do we know what the object’s intended use is.
Thus we hook the encryption handler somewhere in there.
Philipp Gesang [Tue, 28 Mar 2017 13:12:04 +0000]
change extension for encrypted files
Philipp Gesang [Tue, 28 Mar 2017 12:28:06 +0000]
return collected fixed iv parts from .close() when encrypting
Provisional implementation for dumping the IVs in the info file
that will be superseded once the crypto context can be provided
to tarfile by the user.
Philipp Gesang [Mon, 27 Mar 2017 15:43:27 +0000]
automate iv fixed-part generation
The crypto context keeps track of the used IV fixed parts
so they can eventually be included in the info file. A new
fixed part is created in the ctor, then for every time the
counter wraps.
“Wrapping” resets the counter to 2 since 1 is globally reserved
for the info file.
Philipp Gesang [Mon, 27 Mar 2017 15:22:13 +0000]
properly enter/leave encryption sections
Philipp Gesang [Mon, 27 Mar 2017 14:06:18 +0000]
explicitly constuct zlib headers
Get rid of the unnecessary literal byte strings. Commit
5fdff89f4d9fa28e6b210d40d389680072651eb7
introduced headers for additional blocks, omitting the “original
file name” field that gzip set by default.
Philipp Gesang [Mon, 27 Mar 2017 12:10:51 +0000]
unify zlib initialization
Philipp Gesang [Mon, 27 Mar 2017 09:11:00 +0000]
apply compression if compressor is present
Philipp Gesang [Fri, 24 Mar 2017 13:46:59 +0000]
overhaul pre- and post-crypto sync
In order to handle ``_Stream.close()`` well later, we need to
write the header of the last object before the tar info is
injected. To allow the padding and zlib finalization in .close(),
this cannot be performed when the actual file contents are
written but has to be suspended until we are certain no data will
be written to the current crypto object.
Philipp Gesang [Fri, 24 Mar 2017 11:20:45 +0000]
implement encryption passthrough mode
Philipp Gesang [Fri, 24 Mar 2017 10:27:45 +0000]
unify common operations between encryption and decryption
Philipp Gesang [Fri, 24 Mar 2017 10:12:29 +0000]
adjust versioned encryption parameters
Prepare for revised versioning: Both the encryption mechanism and
the KDF may be specified in the versioning.
Philipp Gesang [Thu, 23 Mar 2017 10:48:59 +0000]
extend open_index() API for info file handling
In fact, backup_python’s “info file” is just another “index file”
to deltatar. Conceptually they’re quite different though
especially regarding encryption. To allow requesting an info
flavored index file, add a parameter to communicate with the
crypto layer.
Philipp Gesang [Thu, 23 Mar 2017 09:43:25 +0000]
start payload encryption counter at 2
As per the spec, a file counter of 1 is reserved for the info
file.
Philipp Gesang [Tue, 21 Mar 2017 14:32:32 +0000]
track encryption state
Philipp Gesang [Tue, 21 Mar 2017 14:20:44 +0000]
reinit crypto for objects other than files
Invoke the new-item handler to force a new crypto object for
directories and anything else as well.
Philipp Gesang [Tue, 21 Mar 2017 14:11:32 +0000]
track data handled in crypto
Philipp Gesang [Tue, 21 Mar 2017 12:33:16 +0000]
first draft of revised encryption layer
WIP. Tested for encryption (no compression) only, and only for
TOC listings (-l).
Decryption is handled entirely by the stream, traversing the
archive entry wise relying only one the header informations.
Encryption requires poking the _Stream thingy to initiate a new
crypto entry.
Philipp Gesang [Tue, 21 Mar 2017 12:13:00 +0000]
simplify password save and retrieval
The password must be available for the entire time of the
decryption since it might be necessary to recalculate the key on
account of different salt or parameters of some object.
Philipp Gesang [Mon, 20 Mar 2017 16:35:45 +0000]
implement null-kdf to speed up testing
With “parameter version” zero, the KDF consists only of a trivial
string derived from the password so as to reduce runtime. (SCRYPT
takes about 48 seconds here with our parameters.)
Philipp Gesang [Mon, 20 Mar 2017 15:42:51 +0000]
adapt tag handling in decryption
Philipp Gesang [Mon, 20 Mar 2017 14:02:03 +0000]
fix encrypted read logic for begin/end at entry boundaries
Philipp Gesang [Mon, 20 Mar 2017 11:07:08 +0000]
retrieve and save GCM tag from object header
Philipp Gesang [Mon, 20 Mar 2017 10:48:38 +0000]
delay kdf until parameters are available from header
When decrypting, initialize the key immediately if parameters and
salt are being passed to the ctor. Otherwise, just save the
passphrase in the object and run the KDF when ``.next()`` is
passed the required bits as part of a PDTCRYPT header.
Philipp Gesang [Fri, 17 Mar 2017 16:19:10 +0000]
move ct length bookkeeping into encryptor
This saves us unreliable calculations over the _Stream progress
in tarfile.py.
Philipp Gesang [Fri, 17 Mar 2017 15:02:31 +0000]
move tag back into the header
Since we seek back to write the final header it makes little
sense to append the tag to the ciphertext regardless.
Philipp Gesang [Fri, 17 Mar 2017 10:22:44 +0000]
initially write dummy header during encryption
Philipp Gesang [Thu, 16 Mar 2017 16:39:57 +0000]
create crypto header in .next()
Saves us from exposing the IV to the stream.
Philipp Gesang [Thu, 16 Mar 2017 16:06:28 +0000]
redo transition between objects in crypto layer
When encrypting, the ciphertext size isn’t known beforehand.
Likewise, the file name isn’t available when initializing the
decryption of a file.
Philipp Gesang [Thu, 16 Mar 2017 15:40:31 +0000]
pass salt between index and archive
Memoization of the scrypt params requires static storage because
both the index file and the archive each have an encryption
context.
Philipp Gesang [Thu, 9 Mar 2017 15:41:57 +0000]
revise crypto context interface
Fold key handling and encryption into a common context “class”.
The context takes care of the counter, iv, keys etc. It has one
entry point (ctor) for each direction (read, write → decrypt,
encrypt) and provides hooks for transitioning to the next item.
Header and tag handling remain accessible independent of the
context since tarfile operates on the archive stream and file
objects.
Philipp Gesang [Mon, 6 Mar 2017 15:51:13 +0000]
redo stream decryption
When decrypting, the size of the encrypted object is known, as is
the length of the appended authentication tag; there is no
ambiguity regarding the end of an object. Thus the old string
matching logic with its linear search behavior can go.
Philipp Gesang [Fri, 3 Mar 2017 16:57:37 +0000]
supersede encryption type by encryption parameters
WIP
Philipp Gesang [Thu, 2 Mar 2017 14:41:42 +0000]
polish up backup.py arguments
These are about the only high-level clues regarding its
functionality so it may as well be presented right.
Philipp Gesang [Thu, 2 Mar 2017 13:54:59 +0000]
include offending mode string in exception
Be a little more informative about the cause considering this
exception is passed on to the user as-is.
Philipp Gesang [Thu, 2 Mar 2017 13:40:39 +0000]
ensure octal format is fed an integer
Fix a “type” error that seems to be enforced by Python 3:
File "/src/python/python-delta-tar/deltatar/tarfile.py", line 220, in itn
s = bytes("%0*o" % (digits - 1, n), "ascii") + NUL
TypeError: %o format: an integer is required, not float
Philipp Gesang [Thu, 2 Mar 2017 13:36:43 +0000]
display backup.py usage if no action was specified
Getting no feedback if the invocation had no effect is rather
unseemly. Print the help message instead.
Philipp Gesang [Tue, 28 Feb 2017 16:45:54 +0000]
handle reading and formatting of tags
Philipp Gesang [Tue, 28 Feb 2017 16:01:07 +0000]
unit test scrypt wrapper
Philipp Gesang [Tue, 28 Feb 2017 15:34:34 +0000]
unit test bogus header data
Philipp Gesang [Tue, 28 Feb 2017 14:58:20 +0000]
unit test auth tag handling
Philipp Gesang [Tue, 28 Feb 2017 14:45:16 +0000]
unit test crypto handling of data spanning multiple chunks
Philipp Gesang [Tue, 28 Feb 2017 14:17:04 +0000]
unit test header handling
Philipp Gesang [Tue, 28 Feb 2017 13:36:19 +0000]
add basic wrapper for GCM handling with python-cryptography
This currently requires our yet unmerged contribution:
https://github.com/pyca/cryptography/pull/3421
Philipp Gesang [Tue, 28 Feb 2017 13:34:59 +0000]
add unit test module for encryption layer
Philipp Gesang [Fri, 24 Feb 2017 10:18:18 +0000]
remove key length parameter wherever feasible
Since we’re using fixed AES-128 everywhere, the revised version
no longer offers adjustable key length.
Philipp Gesang [Fri, 24 Feb 2017 09:50:03 +0000]
make tarfile.py error out on invalid crypto modes and combos
The tarfile stream ctor will simply gloss over encryption
requested by the caller unless it happens to exactly match the
string (!) “aes”. Furthermore, with non-gzip compression the
encryption will be ignored altogether.
Instead of deceiving the user about the encryption being applied,
have the ctor fail immediately on invalid combinations.
Philipp Gesang [Thu, 23 Feb 2017 15:34:19 +0000]
init crypto support v2
Implements header reading and writing as well as PoC encryption
wrappers.
WIP
Thomas Jarosch [Mon, 2 Apr 2018 11:31:18 +0000]
Merge branch 'skip-symlinks'
Thomas Jarosch [Mon, 2 Apr 2018 11:15:30 +0000]
RestoreHelper: Prevent endless loop if both indexes contain a list:// entry
This completes the control flow refactoring from Philipp
in commit
7273719cca856677d25102d805f6f96e36173731
Thomas Jarosch [Mon, 2 Apr 2018 10:48:59 +0000]
Document restore code index handling
Philipp Gesang [Mon, 7 Nov 2016 09:00:32 +0000]
ignore all symlinks
Don’t delay the creation of symlinks but suppress it entirely.
The rationale is that extraction with deltatar will only ever
operate on inputs whose symlinks are dereferenced upon archive
creation. Thus valid archives will not contain symlinks at all.
Also, it would appear that deltatar assumes paths of objects
inside a tarball are unique. If the tarball contains ultiple
objects with the same path, it will extract only the first one it
encounters and ignore the rest. This means that it would take at
least two successive backups to perform a symlink attack, the
first one planting the link and the second writing over the
location. This is prevented by the current mitigation strategy
(and by the --unlink option of other tar utilities).
Philipp Gesang [Fri, 4 Nov 2016 16:00:59 +0000]
add unit test for overwriting symlinks
Currently, we implement the behavior of GNU Tar: Subsequent files
in an archive override previous ones, which is also true of
symlinks.
Philipp Gesang [Fri, 4 Nov 2016 14:32:13 +0000]
rectify delayed symlink restoration
Again, GNU tar serves as the model for safe behavior: We now
check whether the placeholder file exists and if it is indeed the
one we created earlier.
Since deltatar does not allow including symlinks in the backup,
the unit tests invoke tarfile functionality directly to add some
symlinks to an existing backup.
Philipp Gesang [Fri, 4 Nov 2016 10:59:34 +0000]
add unit test tracking behavior wrt symlinks
Philipp Gesang [Thu, 3 Nov 2016 16:02:54 +0000]
fix calls to deprecated function in deltatar.py
Fixes warning mandating “.warning()” over “.warn()”:
DeprecationWarning: The 'warn' method is deprecated, use 'warning' instead
Philipp Gesang [Thu, 3 Nov 2016 13:31:14 +0000]
delay only absolute symlinks and those pointing to parent dirs
Only apply the symlink hook on those with fishy targets. Internal
symlinks need not be contained so they can be applied as-is.
Philipp Gesang [Thu, 3 Nov 2016 11:02:15 +0000]
implement delayed symlink creation
Introduce a hook in ``extract()`` to invoke a callback if a
symlink is encountered in the archive. The implementation is
modeled after GNU Tar.
This is a v2 attempt on the symlink extraction problematic. The
first version simply ``unlink(2)`` all files before extraction
which is a less efficient albeit more robust strategy.
Philipp Gesang [Mon, 31 Oct 2016 16:44:42 +0000]
avoid crash in test helper due to fp division
As a matter of fact, ``randomint()`` accepts only int-ishly typed
values, not floats. Consequently, integer division is the way to
go.
Philipp Gesang [Wed, 2 Nov 2016 16:42:33 +0000]
simplify control flow in RestoreHelper methods
Make the control flow more obvious. The code in question was
introduced with commit
ea6d3c3e… but did not make sense back then
either because cur_index which is the constant $1$ was compared
to the literal constant $0$:
+ cur_index = 1
+ while cur_index < len(self._data):
+ data = self._data[cur_index]
…
+ if cur_index == 0:
This bogus test was since removed but the convoluted ``while``
loop survived. Instead, access index 1 only once using an integer
literal.
Thomas Jarosch [Mon, 4 Jul 2016 10:13:39 +0000]
Increase version to 1.5
Thomas Jarosch [Mon, 4 Jul 2016 09:49:06 +0000]
Code review done, comment changes only
Thomas Jarosch [Mon, 4 Jul 2016 09:48:28 +0000]
Remove dead code
cur_index is always >= 1 in this code path
Thomas Jarosch [Mon, 4 Jul 2016 09:48:16 +0000]
Remove code duplication
Thomas Jarosch [Thu, 30 Jun 2016 08:03:40 +0000]
Don't use exception handling for normal control flow
-> Replace buf.index() with buf.find().
Unwinding the stack is expensive and we were
even doing it for the default code path.
Christian Herdtweck [Fri, 24 Jun 2016 07:15:56 +0000]
add file_crypt.py to scripts in setup.py
Christian Herdtweck [Thu, 23 Jun 2016 16:03:03 +0000]
created tool to encrypt/decrypt files using aes128 with compression
Thomas Jarosch [Thu, 23 Jun 2016 12:17:46 +0000]
Increase version to 1.4
Lots of little fixes and improvements.
Christian Herdtweck [Thu, 23 Jun 2016 12:33:40 +0000]
appease pylint
Christian Herdtweck [Wed, 22 Jun 2016 15:14:15 +0000]
fix error found by pylint
Thomas Jarosch [Thu, 23 Jun 2016 10:31:11 +0000]
Rename design document so pylint3 doesn't pick it up
Thomas Jarosch [Thu, 23 Jun 2016 08:08:16 +0000]
Implement cache for pwd.getpwuid() and grp.getgrgid()
Those functions always parse /etc/passwd and we
look up the owner for each file we backup.
This change is only relevant when creating full backups.
Speed up with ~1.000.000 emails is 11%.
Thomas Jarosch [Tue, 21 Jun 2016 08:01:35 +0000]
Fix 'directory' type when iterating tar archives without index
'dir' is not used anywhere in the code base.
Christian Herdtweck [Mon, 20 Jun 2016 07:43:54 +0000]
use the "& 0xFFFfff" after all crc32 calculations
Thomas Jarosch [Fri, 17 Jun 2016 15:39:13 +0000]
Increase release to 1.3
Also switch group from Intranator to Intra2net
Christian Herdtweck [Fri, 17 Jun 2016 14:02:27 +0000]
improve one more unittest: raise proper assertion instead of failing with non-existent variable
Christian Herdtweck [Fri, 17 Jun 2016 14:01:37 +0000]
adjust unittests in test_deltatar
Christian Herdtweck [Fri, 17 Jun 2016 13:29:19 +0000]
correct a comment, add more info to log message
Christian Herdtweck [Fri, 17 Jun 2016 13:28:34 +0000]
adjust filter_path: also remove trailing os.sep
Christian Herdtweck [Fri, 17 Jun 2016 13:28:09 +0000]
fix strip_base_dir argument for DeltaTar._recursive_walk_dir: check for os.sep
Christian Herdtweck [Fri, 17 Jun 2016 09:59:27 +0000]
simplify DeltaTar._recursive_walk_dir
(had called os.path.isdir and filter_path twice on each file directly
after another)
Christian Herdtweck [Fri, 17 Jun 2016 09:56:24 +0000]
had forgotten a few tarobj.close and os.unlink(temp_file) in new tests
Christian Herdtweck [Fri, 17 Jun 2016 07:31:50 +0000]
change one output, make 2 variables to testing routine arguments
Christian Herdtweck [Fri, 17 Jun 2016 07:31:16 +0000]
use KiB, MiB (factor 1024) instead of KB, MB (factor 1000)
Christian Herdtweck [Wed, 15 Jun 2016 13:07:24 +0000]
fix search for file with impossible size (had forgotten that volume_size is in MB)
Thomas Jarosch [Wed, 15 Jun 2016 12:28:38 +0000]
Increase version to 1.2
Thomas Jarosch [Wed, 15 Jun 2016 12:18:33 +0000]
Merge branch 'fix-compression-size'
The new code will give final tar file sizes
close to the volume size even when using compression.
Christian Herdtweck [Wed, 15 Jun 2016 09:39:41 +0000]
ensure temp file is deleted; add some comments about results
Christian Herdtweck [Wed, 15 Jun 2016 09:19:09 +0000]
added performance test script
Christian Herdtweck [Wed, 15 Jun 2016 09:18:55 +0000]
added minimum file size arg to find_random_files
should make returned files more realistic
Christian Herdtweck [Wed, 15 Jun 2016 07:55:16 +0000]
reduce time wasted on _dbg output: format string only when it is actually printed
Christian Herdtweck [Wed, 15 Jun 2016 07:54:02 +0000]
remove _dbg(str.format(args)) from performance-sensitive loop in addfile
Christian Herdtweck [Wed, 15 Jun 2016 07:53:48 +0000]
added some more comments
Christian Herdtweck [Tue, 14 Jun 2016 10:28:32 +0000]
add unittest that runs one of the many multivolume compression size tests
Christian Herdtweck [Mon, 13 Jun 2016 11:07:08 +0000]
created another test for multivolume compression size
Christian Herdtweck [Tue, 14 Jun 2016 09:59:03 +0000]
changed debug output level of the debug output I added earlier
Christian Herdtweck [Mon, 13 Jun 2016 11:06:38 +0000]
removed some debug output
Christian Herdtweck [Mon, 13 Jun 2016 11:06:28 +0000]
fix ValueError message (otherwise '*' is interpreted as string repetition)