When the event is reset in transport_check_fds xfreerdp doesn't work and
consumes 100% CPU (see #2790). On windows this is require otherwise the
CPU consumption is 100% there.
This quick fix only resets the event on windows. It's a working approach
but definitely not the final solution.
The rdpContext gets an event which will
get set if an error occoured in a channel.
If a thread or a void callback has to report an
error it will get signaled by this system.
winsock.h pulls in a lot of defines and dependencies that are not
required and partially unwanted in winpr's core (for parts that are not
related to network). In order to get rid of this dependency and have an
independent defines for extended winpr functions the WINPR_FD_* defines
are used internally (and for exposed functions). Where required, like in
WSAEventSelect, the FD_* is mapped to WINPR_FD_*.
The token buffer size during authentication was constructed
from the wrong buffer size. These sizes are equal in case of
local account logins but differ with domain accounts.
When using NULL credentials (current context)
the server state machine did not send back the
required authentication token.
On client side erroneous checks prevented sending
the appropriate public key.
From MSDN, it looks same as CreateEvent(NULL, FALSE, FALSE, NULL):
The WSACreateEvent function creates a manual-reset event object with an initial state of nonsignaled. The event object is unnamed.
However they are not really equivalent. When we use normal event, the WSAEventSelect still works but the event appears to be 'auto-reset'.
This patch contains:
* checks for malloc return value + treat callers;
* modified malloc() + ZeroMemory() to calloc();
* misc fixes of micro errors seen during the code audit:
** some invalid checks in gcc.c, also there were some possible
integer overflow. This is interesting because at the end the data are parsed
and freed directly, so it's a vulnerability in some kind of dead code (at least
useless);
** fixed usage of GetComputerNameExA with just one call, when 2 were used
in misc places. According to MSDN GetComputerNameA() is supposed to return
an error when called with NULL;
** there were a bug in the command line parsing of shadow;
** in freerdp_dynamic_channel_collection_add() the size of array was multiplied
by 4 instead of 2 on resize
rdp_recv_message_channel_pdu always read the rdp security header
even if it was already previously read (which is the case if rdp
security is active)
This caused malfunctions and disconnects when heartbeat or bandwidth
autodetect packets were sent/received in rdp security mode.
Credit goes to @MartinHaimberger for identifying the broken code
part.
If no entry for the <host> <port> combination
was found in the v2 file and there is a legacy file
check if a matching <host> entry can be found.
In case there is a matching entry and the <fingerprint>
also matches, create a new entry in the v2 file using the
current port.
Exposing lzcnt in crt.h might causes compiler errors (redefinition) with
recent versions of gcc (>=4.9) when winpr is included in other projects.
As lzcnt isn't part of crt according to MSDN and also shouldn't be
exported by default it was moved to intrin.h.
The related test was also moved to the top level directory of winpr.
The grid is composed of 64x64 blocks and should not be smaller
than the surface. If width or height were not a multiple of 64
the previous rounding resulted in a grid smaller than the surface.
- handle WAIT_TIMEOUT result as error in async transport thread
if an INFINITE timeout was specified in WaitForMultipleObjects
- fix mfreerdp's async transport handling to not use
freerdp_get_event_handles/freerdp_check_event_handles if async
transport is activated
It's unclear why this option would be necessary, and it causes problems
when people do not match it to their toolchain and CFLAGS.
To set the float abi, either use a toolchain with an appropriate default
or set the float-abi option in the CFLAGS environment variable.
This should resolve#2586.
1)
Added missing checks for CreateEvent which also required the
following related changes:
- changed freerdp_context_new API to BOOL
- changed freerdp_peer_context_new API to BOOL
- changed pRdpClientNew callback to BOOL
- changed pContextNew callback to BOOL
- changed psPeerAccepted callback to BOOL
- changed psPeerContextNew callback to BOOL
2)
Fixed lots of missing alloc and error checks in the
changed code's neighbourhood.
3)
Check freerdp_client_codecs_prepare result to avoid segfaults
caused by using non-initialized codecs.
4)
Fixed deadlocks in x11 caused by missing xf_unlock_x11() calls
in some error handlers
5)
Some fixes in thread pool:
- DEFAULT_POOL assignment did not match TP_POOL definition
- don't free the pool pointer if it points to the static DEFAULT_POOL
- added error handling and cleanup in InitializeThreadpool
When a hdc is created no initial or default objects are created
therefore can the first call of gdi_SelectObject return NULL.
Because of this checking the return value of gdi_SelectObject failed
for newly create hdc causing errors (disconnects).
Since all types of HGDIOBJECT are handled and the return value of
gdi_SelectObject isn't used the recently added checks were removed
again.
rfx_process_message_sync:
- simplified the check if the header messages got processed
rfx_process_message_tileset:
- ObjectPool_Take result was not checked
- fail if TS_RFX_TILE block type is not CBT_TILE
- CreateThreadpoolWork result was not checked
- post decoding loop code segfaulted in error case
rfx_decoder_tile_new:
- missing malloc check
rfx_message_free:
- segfault protection
rfx_write_message_tileset:
- segfault protection
- removed some unneeded null checks for free()
- fixed a memory leak in shadow_client
- removed rfx_compose_message_header from API
Changed the following functions to BOOL, check the result
where they are called and handle failures:
- rfx_compose_message
- rfx_compose_message_header
- rfx_write_tile
- rfx_write_message_tileset
- rfx_write_message_frame_begin
- rfx_write_message_region
- rfx_write_message_frame_end
- rfx_write_message
rfx_process_message:
- check memory allocation failures
- verify protocol-conform order of data messages to prevents memory
leaks caused by repeated allocations
- verify that header messages were parsed/received before the
data messages
- treat unknown rlgr mode as error
- fixed/added error handling
- fixed all callers to check/handle result
rfx_encode_message:
- fixed incorrect usage of realloc
- missing malloc check
- missing check of CreateThreadpoolWork
- correct cleanup on failure (threadpool, memory)
- check rfx_encode_message result
rfx_encode_messages:
- check rfx_split_message result
- correct cleanup on failure
- prevent memory leak on failure
rfx_write_message_context:
- fixed invalid channelId value (must be 0xFF for WBT_CONTEXT)
rfx_process_message_codec_versions:
- fixed invalid read size of codec_version (it is 16bit)
rfx_process_message_channels:
- verify protocol conform channelId value
rfx_process_message_region:
- replaced invalid reallocs with malloc
- read and verify regionType and numTileSets from stream
rfx_process_message_tileset:
- check allocation results
- fixed incorrect usages of realloc
setupWorkers:
- fixed incorrect usages of realloc
rfx_split_message:
- removed dead code
- missing malloc check
rfx_compose_message:
- fixed a memory leak
- check/handle rfx_encode_message result
* top level GDI functions return 0 on error and != 0 otherwise but the
low level functions (16bpp.c, 8bpp.c 32bpp.c) which are called did it
exactly the other way around. Those were adapted.
* change gdi_InvalidateRegion to BOOL and check calls where appropriate
* integrate comments from pull request
Now using nCount as in and out argument.
When called, set nCount to the number of available handles.
This value is checked and an error returned, if not enough
handles are available.
* Though not frequent, it's possible to get TsProxySetupReceivePipe
data of stublength 4 that is actual data. This happens when
header->common.call_id == rpc->PipeCallId &&
!(header->common.pfc_flags & PFC_LAST_FRAG).
This should address GW disconnects that manifest as SSL read errors.
This is a fix for #2399: when there's no variants we should not try to scan them.
I have set the RDP US keyboard for the South African layout, if someone has a better
layout...
Note: we should probably set something that is not zero for other layouts
Change the return type of Stream_Ensure*Capacity from void to BOOL to be
able to detect realloc problems easily. Otherwise the only way to detect
this was to check if the capacity after the call was >= the required
size.
In case Stream_Ensure*Capacity fails the old memory is still available
and need to freed outside.
This commit also adds checks to most calls of Stream_Ensure*Capacity to
check if the call was successful.
Microsoft iOS Remote Desktop Clients eventually send NULL-terminated
hostnames in SNI which is not allowed in the OpenSSL implementation.
Since we're not using SNI this commit adds an OpenSSL TLS extension
debug callback which modifies the SSL context in a way preventing it
from parsing this extension
The X.224 Connection Request PDU might contain an optional cookie or
routing token before the optional RDP Negotiation Request (rdpNegReq).
If present, both of these fields must be terminated by a 0x0D0A
two-byte sequence. It seems that until now FreeRDP has incorrectly
assumed that a token or cookie must always be present.
If that was not the case, FreeRDP was searching for 0x0D0A until it
arrived at the end of the stream which prevented the remaining data
(RDP Negotiation Request, RDP Correlation Info) from being parsed.
tls_disconnect shut down the ssl stream but didn't inform
the BIO(s) about this therefore could happen that a second shut down
was initiated (e.g. in bio_rdp_tls_free) causing rather long delays.
After removing the shut down from tls_disconnect the only thing the
function does is to prepare/send an alert therefore it was renamed to
tls_send_alert.
