The previous code was assuming that the host name used for doing AAD was
ServerHostname parameter. But when you connect directly to Azure hosts you most
likely connect by IP and use short name for the AAD host, so you need to be able
to give ServerHostname=<IP of host> and AadServerHostname=<shortname>.
When connecting to child session(or a VM), we have CredSSP/Nego/ActivationSequence
instead of Nego/CredSSP/ActivationSequence. Before the patch, we were not applying
flags read in the NegoResponse to the settings and that was leading to FreeRDP
thinking that the remote server was not supporting EXTENDED_CLIENT_DATA_SUPPORTED.
That was later preventing some GCC blocks to be sent, and make the server unhappy
and freeze the connection.
* use enum types for settings getter/setter to have compiler check if
keys are valid
* generate enum types from struct at compile time
* CMake option WITH_OPAQUE_SETTINGS to hide rdpStruct implementation
from external projects
Adds support for server-side remote credential guard in NLA. When enabled that allows
the remote user to connect without shipping credentials in TSCred packets. Instead
it will send his TGT encoded with a TGS from the remote server. This way the server
is able to populate that TGT in a local credential cache without knowing the user's
password.
The patch only treats the NLA part and does not contain the associated RDPEAR channel
that allows to have the complete interaction to retrieve new access tokens.
The RDP client sends TSCredentials so that the server can reuse these credentials. This
patch stores these values in the peer's rdpSettings in the corresponding fields.
It handles TSPasswordCreds for user/domain/password connection but also TSSmartCardCreds
when the users has connected with a smartcard.
Under windows you can connect to a child session by requesting a named pipe to
the local server, and then do some RDP on this named pipe.
The protocol is like for /vmconnect with CredSSP, then Nego and then the "normal"
workflow for a connection. For CredSSP we force the usage of NTLM for the Negociate
SSPI, and the credentials are empty.
This patch prepares the reading of the dynamic channel version so that next we
can take in account this to take advantage of advanced features in last versions
(compressions or priorities).
The patch also implement notifying the VCM event when the dynamic channel becomes ready
so that users of FreeRDP can just do calls to WTSVirtualChannelManagerGetDrdynvcState
when the channel event is set (no blind calls).
Treat the case where the publicIpAddress is there but empty.
Don't try RDSTLS if the password has not been been provided.
When RDSTLS is what we will do, disengage all the other security kinds.
the field for delta rectangles/points/... are only transmitted if they
changed from the previous order of the same type. So keep the original
value and update only if a new one is read.
This patch moves the ARM configuration before starting the connection process, so
that we can do some provisioning of the FreeRDP settings with the items retrieved
from Azure.
Most notably that allows us to connect directly using RDSTLS security.
update initial state transitions according to [MS-RDPBCGR]
the diagram is misleading, some of the text below ambigious, but
1.3.1.1 Connection Sequence phase 10 description lists the dependencies
of server initiated messages.
C requires prototypes or compilers will complain about them missing. Our
library entry points do not have such, therefore add the macro
FREERDP_ENTRY_POINT which declares the function prototype automatically
before the function.
There were some fields sent without considering endianness, making the
clients fail when parsing those fields.
Use Data_Write_XXX functions so the endiannes won't affect the byte order
and the clients will read properly the fields.
All clients can implement their own callbacks for
certificate/credential/smartcard/... but there is a common (default)
implementation for all clients.
with the new setting FreeRDP_UseCommonStdioCallbacks it is now possible
to force these over the client implementation provided ones
The AAD and AVD authentication mechanisms both need an OAuth2 token.
They only differ in the provided arguments, so unify the callbacks into
a single one with variable argument lists.
- Move responsibility for obtaining access tokens to clients
- Add function for getting access tokens for AVD
- Get correct server hostname during AVD setup
- Add utility function for doing http requests
This code fixes an issue where some settings were lost when copying
settings around. I.e. the device count was first set but then calling
`freerdp_settings_set_pointer_len` caused the device count to be reset
to 0 because we passed `NULL` for the data value.
This means the count got lost upon the first copy and further copies
also resulted in losing the array data (because the count was set to 0).
This PR fixes this issue by first resetting the array and afterwards
setting the correct count value.
On some platforms (i.e. Mac) the BOOL type might be narrower than the
`RedirectionFlags` type leading to an overflow when just assigning the
logical AND result to the setting. This fix makes sure that the value is
correctly set to either `TRUE` or `FALSE`.
This PR creates a new /gateway:type:arm transport.
It depends on CJSON
The arm transport is not a transport by itself but is responsible
for getting the websocket endpoint from a configuration url derived from
the configured gateway host in the rdpw file or cmdline
If the reserved filed is not 0 the request PDU seems to contain
some extra data. Two bytes of 0 (probably a version field) followed by a
JSON payload (not null terminated, until the end of the packet. There
seems to be no dedicated length field)
* factor out most websocket specific code parts into websocket.c
* create wst.c (Websocket Transport) as gateway transport implementation
* introduce GatewayUrl setting that holds the websocket url
* introduce GatewayHttpExtAuthBearer that holds the HTTP Bearer
* GatewayHttpExtAuthBearer can be used by both rdg and wst
the new (optional) callback CheckPeerAcceptRestrictions is used to check
for server implementation specific connection requirements before
accepting a client.
Optionally build the SDL client with Qt WebEngine to create a popup
browser for authentication to AAD. Also change the URL output on the
command line to use the "nativeclient" redirect for easier copy/pasting
of the authorization code.
When a PKCS11 module was provided, the CSP could not be set by command line
arguments, leading to an error when loading the ncrypt module, and an empty
smartcard list.
freerdp_bitmap_planar_context_new() expects flags as first argument not a BOOL,
even if giving FALSE ends with the same result, it makes it more clear.
contrary to '[MS-RDPBCGR] 1.3.9 Connect-Time and Continuous Network
Characteristics Detection' we have seen autodetection reqeusts mixed in
between licensing messages. This relaxes the state machine and allows
handling.
The VirtualChannelChunkSize can only be larger than 1600 Bytes, when
both client and server write that value in their capability set
regardless of the value itself.
Also, Microsoft clients and servers only advertise the capabilities that
are relevant for the other peer, e.g. mstsc only tells the server that
it supports decompressing compressed data from the server, but it does
not advertise, that it is able to compress data for the server.
Additionally, correctly apply the read capabilities after reading them.
The VirtualChannelChunkSize setting refers to the VCChunkSize for static
channels and not to the maximum size for DVC data PDUs.
DVC data PDUs are according to [MS-RDPEDYC] always limited to 1600
Bytes.
When enumerating smartcard certificates we check if we have duplicates
in our certificate list. In case we detect a duplicate we just return
`TRUE` (indicating that we consumed the certificate info) but do not
free the smartcard info instance.
Server side we often see "FreeRDP_ChannelDefArray::len expected to be >= 31,
but have XXX", where XXX is lower than 31.
This patche fixes that, the old code was setting the size of ChannelDefArray to the
number of ChannelCount, which is usually not what we want. We want to keep it to 31
and have ChannelCount indicate how many of these channels are used.
* unify reading of domain and username strings with all the checks
* add handling of (undocumented) padding in [MS-RDPBCGR]
2.2.10.1.1.2 Logon Info Version 2 (TS_LOGON_INFO_VERSION_2)
occurring with windows 11
* move type definition to WinPR as used there too.
* supported keyboard types are defined in
[MS-RDPBCGR] 2.2.1.3.2 Client Core Data (TS_UD_CS_CORE)]
use a enum instead of magic numbers to make code more readable.
Commit 2de7a4c249 introduced major changes
in the gateway authentication code. One of these changes was to decouple
NTLM specific authentication from the gateway code.
However with these changes, gateway authenciation with the old RPC code
stopped working and returned an authentication error. The problem is
that currently `credssp_auth_encrypt` encrypts the given message along
creating a signature.
The old code prevented encryption of the message by specifying
`SECBUFFER_READONLY` on the message buffer. The native Windows SSPI then
leaves this buffer as-is and gateway authentication works again.
This fix only applies to Windows platforms using the native SSPI API.
Interestingly this works on other platforms using the WinPR SSPI so
there seems to be a difference between the implementations (but that's a
topic for another PR).
The attributes xPos and yPos for a Color Pointer Update are confusing,
as they may be confused with the xPos and yPos of the pointer bitmap on
the actual screen.
Rename these attributes to what they actually represent, and that is the
hotspot position.
xPos and yPos are still members of the hotspot. However, hotSpotX and
hotSpotY are much more clearer.
In addition to that, the Large Pointer Update uses the same names for
the hotspot coordinates.
Currently, FreeRDP-based server implementations can do connect-time
autodetection.
However, without having any control over it.
In order to be able to override the default connect-time autodetection
handling, introduce three new states for the state machine of the
connection sequence and two new callbacks for the autodetect handling.
These are:
- CONNECTION_STATE_CONNECT_TIME_AUTO_DETECT_BEGIN
- CONNECTION_STATE_CONNECT_TIME_AUTO_DETECT_IN_PROGRESS
- CONNECTION_STATE_CONNECT_TIME_AUTO_DETECT_END
- OnConnectTimeAutoDetectBegin()
- OnConnectTimeAutoDetectProgress()
The END state is pretty simple: When the autodetection is finished and
the autodetect state is FREERDP_AUTODETECT_STATE_COMPLETE, transition
into the next state of the connection sequence.
The BEGIN state is entered, when capability-wise network autodetection
is available.
In this state, the OnConnectTimeAutoDetectBegin callback is called, the
server implementation may initialize any related handling here.
If the server implementation determines, that no further handling is
required, it can end the autodetection phase by returning
FREERDP_AUTODETECT_STATE_COMPLETE.
If not, and an autodetection request is sent, it returns
FREERDP_AUTODETECT_STATE_REQUEST.
The state machine of the connection sequence will then switch into the
IN_PROGRESS state.
In the IN_PROGRESS state, any incoming PDU is handled first, then the
OnConnectTimeAutoDetectProgress callback is called.
Like in the BEGIN state, the return value will determine, whether the
state machine of the connection sequence goes into the END state or goes
into (or rather stays) in the IN_PROGRESS state.
The current state of the autodetect API for the server side does not
include all allowed scenarios where the network autodetection can be
used.
This for example includes the connect-time autodetection, as the
related calls are hidden inside FreeRDP, and not exposed as public API.
In order to avoid duplicate send methods, check the state of the
connection sequence.
If the connection sequence is not yet done, use the connect-time request
types.
Otherwise, use the continuous request types.
The Bandwidth Measure Payload PDU is a little special case, as it is
only allowed to be sent during the connection sequence.
To ensure this, add an assertion in its sending method.
Also fix the handling for the Network Characteristics Sync PDU:
Previously, after parsing the PDU data, the read data was just sent
again to the client, which is wrong.
To fix this issue, introduce a callback for this client-to-server PDU,
so that the actual server implementation can hook up its own handling
for this PDU.
Depending on the situation, the server side may want to discard or use
the retrieved data here.
Moreover, decouple the send-handling for the Network Characteristics
Result PDU from the local autodetect variables.
Currently, these variables are shared between the send and receive
methods.
This leads to access problems, where the server side, wants to use a
different thread to send the autodetect PDU, as the receive handler may
receive an autodetect PDU and overwrite these values with possible
nonsense values.
This is especially the case with RTT response PDUs, as the written
netCharAverageRTT and netCharBaseRTT values are only correct, when only
one RTTRequest happens at a time and no stray RTTResponses are received.
rts_read_common_pdu_header is used to determine if the packet was read
successfully. In that case it might fail but there should be no logging
done to not spam it with unnecessary warnings.
TargetCertificate option is set in TARGET_CERTIFICATE_CONTAINER format.
Exposed ELEMENT_TYPE_CERTIFICATE and ENCODING_TYPE_ASN1_DER1 to allow
building it externally in the mentioned format.
Progressive codec was not honoring the threading flags in settings like does
remoteFX, so even when no multiple threads decoding was asked, progressive was
using multiple threads anyway. This patch fixes it.
During the recent changes the possiblity to perform NLA auth using the
current identity was removed.
In case we receive AUTH_NO_CREDENTIALS with NLA we should resume using a
NULL-identity and not abort the connection.
Authenticate and GatewayAuthenticate return FALSE when there are no
credentials supplied. AuthenticateEx indicates connection termination
with that return value. Handle this accordingly
the new AuthenticateEx callback has a default implementation. To not
break old clients only setting Authenticate or GatewayAuthenticate
callbacks prefer the old ones over the new one.
The spec states that the GUID must be sent as a Base64-encoded GUID in
Unicode format. However in the redirection code we read the (correctly
formatted) GUID and convert it to a binary BLOB.
This PR removes the unnecessary conversion which now results in a
correct RDSTLS auth request.
It also removes some dead code in `rdstls_write_data`.
The client tries to connect using RDSTLS only when it has received a
server redirection PDU with LB_PASSWORD_IS_ENCRYPTED flag.
The server exposes RDSTLS on negotiation if it has been configured on settings.
Then authenticates a client using configured credentials from settings:
RedirectionGuid, Username, Domain, Password.
If a client reconnects on redirection process and uses NLA authentication,
the client was using the old password because it wasn't setting
usePassword to false.
With this commit the client will use the new password.
* extract the certificate from the redirection PDU
* if there is a certificate provided accept it if it matches the
redirection target certificate without further user checks
most protocol internal caches do not need to be exposed. this reduces
the public API and allows us to more easily improve/change this during a
release cycle
Too often experimental flags had been used without the user noticing
that. As bug reports are hard to analyze without proper information take
this approach and inform about experimental flags in use by logging
these.
Currently smartcard settings were only updated in the WIN32 code path.
This must be done on all platforms to have the correct settings (i.e.
pkinitArgs) correctly applied.
Windows seems to favor using the legacy Crypto API (CAPI) for
enumerating RSA key containers and only relies on the newer CNG APIs for
ECC keys.
This PR adds support for CAPI key container enumeration on Windows.
The PR also fixes an issue where the CSP was always set to the MS Base
Smart Card Provider during NLA authentication.
This PR adds a few changes so that a client is able to change the
authentication/logon type in the Authentication callback. I.e. if the
client was started without user/domain the authentication callback is
now able to activate smart card logon by setting the SmartcardLogon
setting along with csp/container/reader name.
Currently if the authentication callback returns `FALSE` the utils
function handle this as scenario as no credentials provided (returns
`AUTH_NO_CREDENTIALS)`.
This PR introduces a new `auth_status` called `AUTH_CANCELLED` that is
returned if the authentication callback returns `FALSE`. If the callback
returns `TRUE` and username or password are empty the util function will
continue to return `AUTH_NO_CREDENTIALS`.
THe PR also fixes some incorrect returns in RPC over HTTP gateway code.
