Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[v0.10] Fixes some problems with monitor hotplug #2951

Merged
merged 12 commits into from
Feb 20, 2024

Conversation

metalefty
Copy link
Member

@metalefty metalefty commented Feb 20, 2024

Merge #2942 into v0.10.

@metalefty metalefty marked this pull request as draft February 20, 2024 15:14
This fixes some monitor hotplug issues with non-GFX codepaths.

1) The server_version_message() was working on an out-of-date
   copy of the client_info. As a result, the X server and the
   window manager did not agree on the number of windows
2) As a result of 1), a memory leak was found in the VNC module.
From [MS-RCPBCGR] 3.3.5.12.1:-

> ...The contents of this PDU SHOULD NOT be compressed.
>
> This PDU MUST NOT be sent to a client that has not indicated support for
> it by setting the RNS_UD_CS_SUPPORT_MONITOR_LAYOUT_PDU flag (0x0040)
> in the earlyCapabilityFlags field of the Client Core Data (section
> 2.2.1.3.2).

Also, 2.2.12.1 specifies the source channel must be zero.

In testing, a compressed monitor layout PDU causes mstsc.exe
to exit with a protocol error.
This stores what kind of resizing (if any) can be achieved with
a Deactivation-Reactivation sequence.
This commit DOES NOT compile.

This change alters these module interface calls:-
1) mod_server_monitor_resize() (Call from xrdp to module). Updated.
2) server_reset() (Call from module to xrdp). Replaced.

The mod_server_monitor_resize() call is updated :-
1) to allow a monitor list to be passed in for a multimon resize
2) with an 'in_progress' return value which tells the caller whether or
   not to expect a callback.

The server_reset() call served two purposes up until now:-
1) To allow a module to resize a single monitor session. There
   is no way to request a multi-monitor resize from the module
2) (with bpp == 0) To signal to the mm resize state machine that
   a server screen resize hsa finished.

This is split into two calls:-
1) client_monitor_resize() to allow a mdule to request a
   multimon resize.
2) server_monitor_resize_done(). This is called by a module
   when a resize is completed.
Neutrinordp module now compiles with updated monitor resize
interface
Significant updates for the VNC module:-
1) Support for the new API calls allowing both server and client
   multi-monitor resizes.
2) The s member variable of the vnc_screen_layout structure is no longer
   dynamically allocated.
3) The module server_width and server_height member variables are
   removed as these are just duplicating server_layout.total_width and
   server_layout.total_height.
4) When the server screens are resized, there is no need to restart the
   entire resize state machine as we already know at this point that
   the server supports resizing.
Input message 300 to xorgxrdp which communicated a screen size
has been replaced with message 302 which sends a list of monitors
to xorgxrdp. This simplifies the initialisation and resize logic
somewhat, but a compatible version of xorgxrdp must be used
Clear egfx_up as soon as the channel starts to close so that
xrdp_mm_draw_dirty() doesn't send data to a closed channel.
This prevents valgrind errors when resizing the screen to
a larger size on GFX systems.
In GFX mode, if we're using xorgxrdp, frame updates are send directly
from the client, bypassing the screen buffer in xrdp_mm.

This commit only allows the xrdp_mm screen buffer to be invalidated
if the painter has drawn into it since the module was loaded. This
prevents the unused (and invalid) frame buffer being pushed to the client
in GFX mode with xorgxrdp.
@metalefty metalefty marked this pull request as ready for review February 20, 2024 15:20
@metalefty metalefty merged commit c3cb855 into neutrinolabs:v0.10 Feb 20, 2024
13 checks passed
@metalefty metalefty deleted the v0.10-monitor-hotplug branch February 20, 2024 16:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants