The existing code used to produce an empty output in situations where a
TLV_IE_Collection would be parsed from a single TLV only with some
additional trailing padding:
>>> from pySim.utils import h2b
>>> from pySim.ts_31_102 import EF_CSGT
>>> t = EF_CSGT.Csgt_TLV_Collection()
>>> t.from_tlv(h2b('8906810300666f6fff'))
[TextCsgType(foo)]
>>> t.to_dict()
[]
This was caused by an early return (actually returning the decoded
result) but *without updating self.children*.
Change-Id: I1c84ccf698c6ff7e7f14242f9aaf7d15ac2239f4
These files are mostly related to CSG (Closed Subscriber Group)
in the context of HomeNodeB (HNB), aka femtocells.
Change-Id: Ie57963381e928e2c1da408ad46549a780056242a
Now that we have support for the UCS-2 encoding as per TS 102 221 Annex A,
we can start to make use of it from various file constructs.
As some specs say "Either 7-bit GSM or UCS-2" we also introduce
a related automatic GsmOrUcs2Adapter and GsmOrUcs2String class.
Change-Id: I4eb8aea0a13260a143e2c60fca73c3c4312fd3b2
TS 102 221 Annex A defines three variants of encoding UCS-2 characters
into byte streams in files on UICC cards: One rather simplistic one, and
two variants for optimizing memory utilization on the card.
Let's impelement a construct "Ucs2Adapter" class for this.
Change-Id: Ic8bc8f71079faec1bf0e538dc0dfa21403869c6d
The human representation of a PLMN is usually MCC-MNC like 262-01
or 262-001. Let's add a PlmnAdapter for use within construct, so we
can properly decode that.
Change-Id: I96f276e6dcdb54a5a3d2bcde5ee6dbaf981ed789
We used __subclasses__(), but this only returns the immediate
subclasses and not all further/nested subclasses. Instead, we must
use the pySim.utils.all_subclasses() function to really get all of them.
The hack to use the method signature of the constructor to determine if
it's an intermediate class didn't work, as even GlobbalPlatformISDR
has a optional argument for non-default AIDs. So let's introduce an
explicit class attribute for that purpose.
Change-Id: I7fb1637f8f7a149b536c4d77dac92736c526aa6c
This adds some first test data for the new unitdata driven test cases
for the TLV encoder/decoder.
It also fixes a bug in the ts_102_221.FileDescriptor decoder for BER-TLV
structured files which was found and fixed while introducing the test
data.
Related: OS#6317
Change-Id: Ief156b7e466a772c78fb632b2fa00cba2eb1eba5
While we do have the _test_de_encode data driven tests for file
definitions, we don't yet have something similar for derived classes of
BER_TLV_IE. This means that TLVs used outside of the filesystem context
(for example, decoding the SELECT/STATUS response, but also eUICC and
other stuff) do not yet have test coverage.
This commit just adds the related test code, but no test data yet.
Related: OS#6317
Change-Id: Ied85f292bb57fde11dc188be84e3384dc3ff1601
the to_dict() method generates a {class_name: value} dictionary,
for both the nested and non-nested case. However, before this patch,
the from_dict() method expects a plain list of child IE dicts
in the nested case. This is illogical.
Let's make sure from_dict always expectes a {class_name: value} dict
for both nested and non-nested situations.
Change-Id: I07e4feb3800b420d8be7aae8911f828f1da9dab8
We recently introduced a pySim.utils.build_construct() wrapper around
the raw call of the construct.build() method. So far, this wrapper
was only used from pySim.tlv, but let's also use it from
pySim.filesystem.
Basically, whenever we use parse_construct(), we should use
build_construct() as the inverse operation.
Change-Id: Ibfd61cd87edc72882aa66d6ff17861a3e918affb
on some systems, the output would otherwise contain colored status
words, which in turn mean the test otuput no longer matches the expected
output.
Change-Id: Icb700f6e85a285748e00367a398975aa5e75dec5
The context is some opaque dictionary that can be used by the
constructs; let's allow the caller of parse_construct, from_bytes,
from_tlv to specify it.
Also, when decoding a TLV_IE_Collection, pass the decode results of
existing siblings via the construct.
Change-Id: I021016aaa09cddf9d36521c1a54b468ec49ff54d
The EF.CFIS definition is not identical to EF.ADN, so we cannot recycle
the EF.ADN class to decode EF.CFIS.
Change-Id: Idcab35cbe28332e3c8612bcb90226335b48ea973
There are some pretty intricate rules about how GSM and E-UTRAN are
encoded, let's make sure we fully support both as per 3GPP TS 31.102
Release 17. As part of this, switch to a sorted list of access technologies,
in order to have a defined order. This makes comparing in unit tests
much easier. However, it also means that we need to sort the set
when printing the list of AcT in pySim-read to generate deterministic
output.
Change-Id: I398ac2a2527bd11e9c652e49fa46d6ca8d334b88
Without this the diff between expected and actual output is truncated
and one instead reads the following output:
Diff is 844 characters long. Set self.maxDiff to None to see it.
We actually want to see the full diff to see what's not matching.
Change-Id: I6e89705061454191b6db1255de7fe549ad720800
This driver description we add to the code is automatically added to the
respective user manual sections.
Change-Id: I8807bfb11f43b167f1321d556e09ec5234fff629
Let's avoid copy+pasting print statements everywhere. The instances
do already have a __str__ method for the purpose of printing their name in a
generic way.
Change-Id: I663a9ea69bf7e7aaa6502896b6a71ef692f8d844
Opening PC/SC readers by index/number is very error-prone as the order
is never deterministic in any system with multiple (hot-plugged, USB)
readers. Instead, let's offer the alternative of specifying a regular
expression to match the reader name (similar to remsim-bankd).
Change-Id: I983f19c6741904c1adf27749c9801b44a03a5d78
It's odd that the individual transport driver specifies their argparse
options but then the core transport part evaluates them individually.
This means we cannot add new options within a transport.
Let's pass the Namespace instance into the constructor of the
specific transport to improve this.
Change-Id: Ib977007dd605ec9a9c09a3d143d2c2308991a12c
In the previous commit we've stopped using those functions from modern
pySim-shell code. Hence, the only remaining user is the legacy tools,
so we can move the code to the legacy module.
Change-Id: I6f18ccb36fc33bc204c01f9ece135676510e67ec
We've recently introduced IPv{4,6}Adapter construct classes and can
switch to this instead of using the old imperative encoder/decoder
functions {enc,dec}_addr_tlv().
Aside from code cleanup, this also means we now support the IPv6 address
type in EF.PCSCF.
Change-Id: I4d01ccfe473a8a80fbee33fdcbd8a19b39da85ac
This can happen if there's a file with invalid encoding on the card,
such as a tag followed by all-ff. Let's gracefully ignore it and
return zero bytes as response.
Change-Id: Ic44557368a6034dbf4bb021ab23a57927c22def0
The implementation of the methods select and select_file of class
RuntimeLchan is a bit complex. We access the card directly in several
places which makes it difficult to track the state changes. We should
clean this up so that we call self.rs.card.select_adf_by_aid/
self.scc.select_file from a single place only.
This means that the method select uses the method select_file. This
results in a much cleaner implementation. We also should take care
that the important states that we track (selected_file, selected_adf,
etc.) are updated by a single private method. Since the update always
must happen after a select _select_post is a good place to do this.
Related: OS#5418
Change-Id: I9ae213f3b078983f3e6d4c11db38fdbe504c84f2
The method build_select_path_to uses the internal file system tree model
to find the path to a given file. This works the same for applications
(ADF) as it works for normal files (EF/DF). However, an application can
be selected anytime from any location in the filesystem tree. There is
no need to select a specific path leading to that application first.
This means that if there is an ADF somewhere in the resulting
inter_path, we may clip everything before that ADF.
Related: OS#5418
Change-Id: I838a99bb47afc73b4274baecb04fff31abf7b2e2
We use a trick to probe a file (that does not exist in the local file
model yet). Let's explain further how that works, in particular why we
do not have to upate any state if probing fails.
Change-Id: I2a8af73654251d105af8de1c17da53dfa10dc669
Related: OS#5418
The metaclass uese the 'nested' attribute, while the existing code
accidentially used the 'children' attribute. The latter is used
by instances for actual child classes, while the Class/nested
attribute is for the list of classes whose instancse could be potential
children.
Change-Id: I968bd84d074dcdcec37d99be5d3d4edac9c35a0c
The length value "of course" is a hex value, don't use %02u but %02x
This fixes any eUICC command with a Lc > 10 bytes.
Change-Id: I1e1efbfb9916fc43699602cc889cf4b3d42736f2
When we traverse the file system, we may also end up selecting
applications (ADF), which do not support an USIM/ISIM like file system.
This will leave us without the ability to select the MF (or any other
file) again. The only way out is to select the ISIM or USIM application
again to get the access to the file system again.
Change-Id: Ia2fdd65f430c07acb1afdaf265d24c6928b654e0
Related: OS#5418
The encoding was missing a "CHOICE" container and missed the
fact that the refreshFlag presence is mandatory for enable+disable.
Change-Id: I12e2b16b2c1b4b01dfad0d1fb485399827f25ddc
If a TLV was elementary (no nested IEs), and it had only a single
integer content whose value is 0, we erroneously encoded that as
zero-length TLV (len=0, no value part):
>>> rf = pySim.euicc.RefreshFlag(decoded=0);
>>> rf.to_bytes()
b''
>>> rf.to_tlv()
b'\x81\x00'
After this change it is correct:
>>> rf = pySim.euicc.RefreshFlag(decoded=0);
>>> rf.to_bytes()
b'\x00'
>>> rf.to_tlv()
b'\x81\x01\x00'
Change-Id: I5f4c0555cff7df9ccfc4a56da12766d1bf89122f
One of the most important properties of the RuntimeLchan are the
selected_file/adf properties. Let's reformat the code so that those
properties are more pronounced.
Change-Id: I4aa028f66879b7d6c2a1cd102cda8d8ca5ff48b1
Related: OS#5418
When we are in the constructor of RuntimeState, we may/must access the
card object directly. Let's explain why, since it may not be immediately
obvious.
Change-Id: I01f74d5f021d46679d1c9fa83fb8753382b0f88f
Related: OS#5418
The constructor of the RuntimeState object selects the MF befor it does
some other steps. However it does this through the _scc object of the
card object. This method is before we had lchan abstraction, so we
should now use the lchan object like in all other places.
Related: OS#5418
Change-Id: I9a751c0228c77077e3fabb50a9a68e4489e7151c
If we have a list of dicts, and we flatten that into a dict: Only
do that if there are no dicts with duplocate key values in the list,
as otherwise we will loose information during the transformation.
Change-Id: I7f6d03bf323a153f3172853a3ef171cbec8aece7
Closes: OS#6288
Ever since commit 30de9fd8ab in July
we are (properly) using snake_case names in the from_dict (to become
bijective with to_dict). This code was not updated by accident,
creating an exception when using the `aram_get_config`
Change-Id: If216b56b38ab17d13896074aa726278b9ba16923
Related: OS#6119
An ADF may or may not support a file system. For example ADF.ARA-M does
not have any filesystem support, which means the SELECT we may use from
this ADF is limited and an can only select a different application. To
know about this in advance let's add a flag that we set when we
instantiate an ADF.
Change-Id: Ifd0f7c34164685ea18d8a746394e55416fa0aa66
Related: OS#5418
So far we implemented only one round of "Send the APDU, get SW=61xx,
call GET RESPONSE". This permitted us to receive only data up to 256
bytes.
Let's extend that to doing multiple rounds, concatenating the result.
This allows us to obtain arbitrary-length data from the card.
See Annex C.1 of ETSI TS 102 221 for examples showing multiple 61xx
iterations.
Change-Id: Ib17da655aa0b0eb203c29dc92690c81bd1300778
Closes: OS#6287
Harald Welte
Philipp Maier