vglass: replace surfman/input-server with vglass #1382

eric-ch · 2020-12-15T21:07:10Z

In this changeset:

Retire Surfman related projects (surfman, libsusrfman, surfman-plugins, input_server, libdmbus, shutdown-screen, xenfb2, gem-foreign/i915-foreign patches, libedid);
Retire the SELinux definitions associated with the aforementioned projects;
Retire the IDL exclusively associated with the aforementioned projects;
Include the vglass layer: https://gitlab.com/vglass/meta-vglass as a dependency, note this inherits meta-qt5 as a dependency;
Amend the necessary configuration to use the required vGlass PV drivers where necessary (ivc, pv-display-helper, openxtfb in UIVM)

General notes:

dom0 memory is bumped to 768M. This has been changed by recommandation, yet it is suspected keeping 512M should be enough with the backported patch for gntdev (9b24f64).

vGlass notes:

This submission currently relies on openxtfb as a PV framebuffer driver in UIVM. An alternative is available (glassdrm), and will be the object of later submissions.
IVC (libivc2) in dom0 is not the same component as IVC (libivc) in service-VMs (stubdomains included).
HVM w/o stubdomain guests are not supported by the current vGlass stack. There is no graphic support for this use-case due to IVC2/IVC limitations at the moment.
Monitor hotplug issues have been reported and are being investigted, they can and should be reported in the vGlass gitlab group: gitlab.com/groups/vglass
There is a known limitation with Windows PV Tools when using the emulated mouse. This changeset includes the work-around to let Windows HVM w/ stubdomain guests create and use a PV device.

Windows PV tools:

vGlass PV tools interface with Xen Windows PV drivers 9.0, the builder provides a built binary: https://openxt.ainfosec.com/OpenXT/auto/windows/9.0.0/;
The bundled installer also includes a port of OpenXT Windows PV USB drivers on Xen Windows PV drivers 9.0;
There is no PV audio drivers available at the moment;

Builds and test:

There is a dedicated builder for this changeset: https://openxt.ainfosec.com/OpenXT/auto/master/master-custom-vglass/

Depends on:

eric-ch · 2020-12-18T19:32:19Z

v2: Rebase on top of master. No change.

jandryuk

Noticed two avcs when a usb-c dock triggered display and usb hotplug. This was a local build - are the public ones running in permissive?

vusb.te should gain
glass_dbus_chat(vusbd_t)
It used to have input_server_dbus_chat(vusbd_t).

dbus-daemon[1142]: avc:  denied  { send_msg } for msgtype=method_call interface=com.citrix.xenclient.input member=get_focus_domid dest=com.citrix.xenclient.input spid=1299 tpid=1388 scontext=system_u:system_r:vusbd_t:s0 tcontext=system_u:system_r:glass_t:s0 tclass=dbus permissive=0

disman.te should gain
disman_dbus_chat(udev_t)
So that udev -> 99-disman-hotplug.rules -> /usr/bin/disman-hotplug.sh -> dbus-send can work. Or disman_script_t needs to be fully implemented as I've indicated elsewhere.

dbus-daemon[1142]: avc:  denied  { send_msg } for msgtype=method_call interface=mil.af.secureview.disman member=hotplug dest=mil.af.secureview.disman spid=3726 tpid=1395 scontext=system_u:system_r:udev_t:s0-s0:c0.c1023 tcontext=system_u:system_r:disman_t:s0 tclass=dbus permissive=0

conf/machine/xenclient-uivm.conf

jandryuk · 2021-01-04T13:34:24Z

recipes-security/refpolicy/refpolicy-mcs-2.%/policy/modules/services/disman.fc

+/etc/init\.d/disman		--	gen_context(system_u:object_r:disman_initrc_exec_t,s0)
+
+/usr/bin/disman			--	gen_context(system_u:object_r:disman_exec_t,s0)
+/usr/bin/disman-hotplug.sh	--	gen_context(system_u:object_r:disman_exec_t,s0)


Swith this to disman_script_exec_t:
/usr/bin/disman-hotplug.sh -- gen_context(system_u:object_r:disman_script_exec_t,s0)

... and also \ that ..

Yes. Good self-review 🙂

recipes-security/refpolicy/refpolicy-mcs-2.%/policy/modules/services/disman.te

jandryuk

These can go in separately and immediately, I think:
d9eac76
e12d717

conf/machine/xenclient-dom0.conf

jandryuk · 2021-01-04T13:57:10Z

recipes-core/images/xenclient-uivm-image.bb

@@ -96,6 +96,7 @@ IMAGE_INSTALL += "\
    anthy \
    matchbox-keyboard \
    matchbox-keyboard-im \
+    kernel-module-openxtfb \


This seems like it's in the wrong commit 761aed3

It is indeed.

jandryuk · 2021-01-04T13:59:16Z

conf/machine/xenclient-dom0.conf

@@ -21,3 +19,5 @@ KERNEL_MODULE_AUTOLOAD += " \
    psmouse \
    hid-multitouch \
 "
+
+PREFERRED_PROVIDER_virtual/libivc = "libivc2"


I was hoping that dependencies could handle this, but libpvglass depends on virtual/libivc, so we can't just make vglass depend on libvc2 and qemu depend on libivc, right?

Yes that is why I ended up defining a prefered per-machine provider, which in turns comes from libpvbackendhelper/libpvdisplayhelper, aka recipe libpvglass, exporting the same public headers, OE rightfully reports one can overwrite the other if they are provided by two different recipes.

recipes-extended/qemu-dm/qemu-dm.inc

jandryuk · 2021-01-04T14:05:26Z

recipes-kernel/linux/5.4/linux-openxt_5.4.64.bb

+
+do_configure_append_openxt-installer(){
+    # vglass supports radeon and nouveau DRM.
+    kernel_conf_variable DRM_RADEON m


We can just update all the relevant defconfigs.

Agreed. It was somewhat convenient to isolate the changes initially.
Just for documentation: With #1383 it would become configuration fragments.

jandryuk · 2021-01-04T14:14:04Z

recipes-extended/xen/files/libxl-xenmgr-support.patch

     {
         libxl__device_console console;
         libxl__device device;
 -        libxl_device_vkb vkb;
+
+        for (i = 0; i < d_config->num_vkbs; i++) {


Instead of adding this here, this hunk can be dropped to re-instate the original code

xenclient-oe/recipes-extended/xen/files/libxl-xenmgr-support.patch

Line 78 in 4f925c4

- libxl__device_add(gc, domid, &libxl__vkb_devtype, &vkb);

That is assuming 1 vkb is acceptable, but I think only one is added with vglass.

This was a long time ago and I'm trying to remember why the vkbs ended up like this. libxl-xenmgr-support.patch split apart the vfb and vkb creation in xl_parse.c because I think we wanted 2 vkbs (one for keyboard and one for mouse). I think you're right about vglass not needing 2 vkbs so we could drop the loop in xl_parse.c, but the UIVM still wants vfb=0, and vkb=1, so the initialization needs to remain split.

The d_config->vkbs[] array already has initialized vkbs from the code in xl_parse.c, so we're just calling the libxl__device_add() in this hunk instead of initializing another new vkb, like line 78 in libxl-xenmgr-support.patch.

Yes, input server used two vkbs because a single one can only send either relative or absolute events.

UIVM is PVH, so it doesn't go through this code path and vkbs can be added separately anyway. My suggested change would mean every HVM gets a vkb xenstore entry regardless of config.vkbd setting. vglass actually providing a backend seems to be conditioned on the config.vglass-enabled setting.

The previously mentioned xenmgr change to use vkb=[] (https://github.com/jandryuk/manager/commits/vkb-fixup) may mean I'm missing something for the case without it.

vkb=[0|1] and vfb=[0|1] are colliding with xl.cfg(5) definitions for vkb=[ "VKB_SPEC_STRING", "VKB_SPEC_STRING", ...] and vfb=[ "VFB_SPEC_STRING", "VFB_SPEC_STRING", ...] which, it sounds like, we could use directly without removing the other PV display support like @jandryuk points out.

Windows HVM guests and UIVM by default boot with vfb=0 and vkb=1 when using vglass.

With @jandryuk changes the changes from libxl-xenmgr-support.patch related to vkb and vfb are dropped. Are there still concerns regarding this?

recipes-security/refpolicy/refpolicy-mcs-2.%/policy/modules/services/glass.fc

recipes-security/refpolicy/refpolicy-mcs-2.%/policy/modules/services/ivcd.fc

eric-ch · 2021-01-07T20:59:44Z

v2

Address refpolicy comments and typos (e,g \.);
xen-fbfront is no longer in the modprobe configuration, it will be loaded depending on the xenbus probe. xenbus.wait_nonessentials is removed since no longer needed;
vkb and vfb flags in xenmgr are deprecated, amend the libxl patch in consequence (see relevant commits in vglass: introduce RPCs used by vglass and remove deprecated ones idl#41 and vglass: deprecate surfman logics and wire up vglass manager#180). An upgrade script is still needed;
Prune the kernel defconfigs using bitbake -c savedefconfig to remove auto-selected configurations;
Transfer kernel configuration from the linux-openxt recipe to the defconfig themselves;
sort out cross-commit changes

eric-ch · 2021-01-08T15:08:41Z

Build: $mcv-9

jandryuk · 2021-01-08T15:50:17Z

This looks good.
a43a418 got re-introduced?
c77f7d5 should maybe be its own PR?

eric-ch · 2021-01-08T15:55:30Z

@jandryuk a43a418 is a mistake on my part, I must have rebased before the other PR went in. Agreed for c77f7d5.

eric-ch · 2021-01-08T16:07:23Z

v3

Rebase on master
Removed c77f7d5 for separate submission.

jandryuk · 2021-01-13T19:43:04Z

I've seen vglass segfault 4 times.
This is Eric's build. Looks like it was from hot unplugging a USB-C Dock with an external display.

Dec 16 16:37:25.486943 kernel:[ 3267.244797] glass[1433]: segfault at 0 ip 00007ff88ba1c10f sp 00007ffd47b71280 error 4 in libfs_wm.so.1.0.0[7ff88ba0b000+1f000]

The December segfault is Eric's build and the January ones are my own builds:

Dec 18 13:54:31.235419 kernel:[ 4169.988263] glass[1295]: segfault at a0 ip 00007f3b6a02c570 sp 00007ffea13b7528 error 4 in libpthread-2.31.so[7f3b6a028000+f000]
Jan  8 20:22:00.205688 kernel:[ 1019.965024] glass[1408]: segfault at a0 ip 00007f32e6a95570 sp 00007fffab7649e8 error 4 in libpthread-2.31.so[7f32e6a91000+f000]
Jan 13 19:27:04.764638 kernel:[16805.510069] glass[1373]: segfault at a0 ip 00007f6d7dcbb570 sp 00007fff73002aa8 error 4 in libpthread-2.31.so[7f6d7dcb7000+f000]

These occur randomly when shutting down a Windows 10 HVM without PV graphics drivers.

If you ssh into dom0, you can manual restart vglass, but uivm doesn't display - it doesn't reconnect its graphics. X, surf and xterm windows are still running seemingly none the wiser. Restarting uivm restores the display. The inittab line makes this tricky, but If you sshargo into uivm, you can kill X, rmmod and modprobe openxtfb, and then restart X to get the display restored.

eric-ch · 2021-01-13T23:06:36Z

I have also had glass crash when plugging an external display (DVI port, but likely irrelevant) while a windows guest with PV tools is running, but UIVM is in focus:

[75046.121497] glass[1717]: segfault at 0 ip 0000000000000000 sp 00007faaf3013418 error 14 in glass[400000+6000]
[75046.121537] Code: Bad RIP value.

eric-ch · 2021-01-22T15:48:36Z

OpenXT: Build: $mcqv-11
Windows tools: Build: $9.0.0-5

eric-ch · 2021-01-26T23:39:19Z

V4

Rebase on master
Add new dependency to xenmgr to support the new RPC

eric-ch · 2021-04-15T20:33:02Z

V5

Rebase on master

eric-ch · 2021-04-26T14:56:04Z

OpenXT: Build: $master-custom-vglass-18
WinTools: Build: $windows-tools-9-0-0-13

dpsmith

No Rb since I haven't honestly reviewed the latest but willing to ACK merging

replaced by openxtfb or xenfb. Signed-off-by: Eric Chanudet <[email protected]>

Deprecated, vGlass can handle splash-screen without help from the kernel. Signed-off-by: Eric Chanudet <[email protected]>

Replaced with vglass. So long Surface Manager. Signed-off-by: Eric Chanudet <[email protected]>

Signed-off-by: Eric Chanudet <[email protected]>

Replaced by vglass. Signed-off-by: Eric Chanudet <[email protected]>

Signed-off-by: Eric Chanudet <[email protected]>

https://www.openembedded.org/wiki/Styleguide Add quotes and use OE variables. Remove deprecated Rpc. Signed-off-by: Eric Chanudet <[email protected]>

The only consumers of dmbus were surfman and input-server. Both are deprecated, dmbus can follow. Signed-off-by: Eric Chanudet <[email protected]>

remove switcher.patch remove xenmou.patch switcher.patch also added the configuration for libargo, this is now done in a new patch (argo-configure.patch). Fix conflicts and refresh relevant patches on top of the above. Signed-off-by: Eric Chanudet <[email protected]>

xen-translate.patch is no longer required and was already a hack as surfman should have dealt with gpfn instead of mfns. workaround-nehalem-igd-vtd.patch is a work-around with security implications for platform 10+ years old. vglass no longer rely on the gem-foreign which is the root cause of this. Signed-off-by: Eric Chanudet <[email protected]>

libivc2/ivcdaemon is only suited for dom0. The Linux stubdomain should not add QT or C++ runtime dependencies to keep the footprint manageable, hence use the ivc module through libivc. Signed-off-by: Eric Chanudet <[email protected]>

Refresh the patch-queue for fuzz introduced by vglass changes in it. Signed-off-by: Eric Chanudet <[email protected]>

Apparently a wrapper script around depmod. Unused and forgotten. Signed-off-by: Eric Chanudet <[email protected]>

bitbake -c savedefconfig, which runs make savedefconfig, which in turns takes into account automatically selected configurations. No functional changes, this only removes kernel configuration statement that would be selected anyway by the remaining configuration or generated by the kernel build-system. Signed-off-by: Eric Chanudet <[email protected]>

Add vglass components to relevant images and configurations. - enable meta-vglass layer - Add vglass and disman to dom0 - Pass kernel modules recommended for vglass. Note, some of the kernel configuration only handles drivers in images that do not leverage vglass, but are now supported with vglass. Signed-off-by: Eric Chanudet <[email protected]>

inherit the module-signing bbclass for out-of-tree modules. Signed-off-by: Eric Chanudet <[email protected]>

xenfb2 is now deprecated. Remove xf86 fbdev driver customization supporting randr1.2 interfaces for modesetting to a higher resolution framebuffer. vGlass provides the same feature in openxtfb associated xf86 driver: xf86-video-openxtfb. Signed-off-by: Eric Chanudet <[email protected]>

Add openxtfb to UIVM. vGlass provides an xf86 Xorg driver for the PV framebuffer openxtfb. Signed-off-by: Eric Chanudet <[email protected]>

Signed-off-by: Eric Chanudet <[email protected]>

openxtfb is the default framebuffer driver in UIVM with vGlass. Autoload the driver. Signed-off-by: Eric Chanudet <[email protected]>

allow-pat-cacheattrs-for-all-domains is only used by xenfb2/surfman that would foreign map pages in the guest and have the GPU scan them, hence required the guest to set the cache policies to WT or WC. PV guests were prevented from doing this. Note that since UIVM was PVH, the cache policy was changed in surfman anyway. allow-stubdoms-cacheattr-control.patch handles the same issue from the stubdomain perspective. Since the i915-foreign map is no more, this is no longer required. foreign-batch-cacheattr.patch is no longer required since the cache policy is no longer modified by the display stack. Signed-off-by: Eric Chanudet <[email protected]>

Surfman used VBE and other BIOS call extensions to handle early boot resolution. This is no longer required. QEMU: - vbe-xt-extensions.patch: Added VBE calls to configure resolution taken by Surfman from the host EDID while being compatible with the GPU scan and make sure this was used from early compatible resolutions (set by vgabios). Used by vgabios/vbe-extensions.path and vgabios/vbe-edid-interface.patch - vga-spinlock: Work-around for vgabios/vbe-xenvesa.patch presumably. int10h was used after bootstrap to pass up modsetting information to the defunct xenvesa driver. It is unclear how this could have raced. Used by vgabios/vga-spinlock.patch, vgabios/vga-shadown-bda.patch, vgabios/vbe-xenvesa.patch. - vga-shadow-bda.patch: implement IO-ports to store VGABIOS data otherwise lost after early bootstrap as it resides in BDA. VGABIOS: - vbe-edid-interface.path: Use VBE extensions implemented in qemu/vbe-xt-extensions.patch to handle compatible resolutions for Surfman. - vbe-extensions.patch: In the line of vbe-edid-interface.patch, except these calls are not defined in the VBE interface. - vbe-xenvesa.patch: Pass a Surfman provided EDID through an int10h IO-port. The EDID being stored in BDA, this depends on vga-shadow-bda.patch as it would be used after the BDA is reclaimed by the guest OS. - vga-shadow-bda.patch: Allow some BDA stored info to still be reachable post bootstrap. - vga-spinlock.patch: spinlock to avoid presumed int10h races in emulated guests. Signed-off-by: Eric Chanudet <[email protected]>

openxtfb depends on: - FB_SYS_FILLRECT - FB_SYS_COPYAREA - FB_SYS_IMAGEBLIT - FB_SYS_FOPS - FB_DEFERRED_IO - FB_MODE_HELPERS In order to select the above configuration, UIVM is built with XEN_FBDEV_FRONTEND as it shares and selects most of these. Both drivers will therefor load creating /dev/fb0 and /dev/fb1 by default. This is undesirable as it makes the XOrg configuration more complicated, so avoid loading xen-fbfront. This could be handled differently, e.g, in the xf86 driver. Signed-off-by: Eric Chanudet <[email protected]>

modules.dep is present, modprobe can be used. Add ivc and use modprobe for xen-argo. Signed-off-by: Eric Chanudet <[email protected]>

libedid was used by Surfman to emulate EDID values. Signed-off-by: Eric Chanudet <[email protected]>

vGlass has higher memory requirements. This increase is motivated as vGlass would otherwise fail to drive UIVM with ENOMEM when opening the grant-refs used by the channels with UIVM and other guests. Signed-off-by: Eric Chanudet <[email protected]>

Backport upstream commit: b3f7931f5c61 xen/gntdev: switch from kcalloc() to kvcalloc() When mapping guest framebuffer, gntdev can reach order 5 allocation sizes (1920x1080 display only). Using non physically contiguous buffers makes it easier on dom0's memory. Signed-off-by: Eric Chanudet <[email protected]>

SELinux module policies for ivcdaemon, glass and disman components of vglass. Signed-off-by: Eric Chanudet <[email protected]>

Remove hunks related to display_depth and display_res passed via XenStore. No components were using these initially. These might be an artefact from before libdmbus. Signed-off-by: Eric Chanudet <[email protected]>

vfb=[0|1] and vkb=[0|1] only really create the base object for each structure in order to get xl to create the base xenstore nodes. This is redundant with vkb=[ "VKB_SPEC_STRING", "VKB_SPEC_STRING", ...] and vfb=[ "VFB_SPEC_STRING", "VFB_SPEC_STRING", ...] which should be used instead. Note: vfb is constantly set to 0 in xenmgr with vglass, should '1' be useful somehow it would need to be handled via a new VFB_SPEC_STRING entry to have the base vfb structure initialized correctly. Signed-off-by: Eric Chanudet <[email protected]>

Changes to Xenmgr require the split from Hackage. Signed-off-by: Eric Chanudet <[email protected]>

eric-ch · 2021-05-24T21:44:30Z

V6

Rebase on master

eric-ch · 2021-05-25T17:56:37Z

Build: $master-custom-vglass-20

jandryuk · 2021-05-28T13:31:56Z

Merging soon

eric-ch force-pushed the vglass branch from 5dbec36 to 4f925c4 Compare December 18, 2020 19:31

jandryuk reviewed Jan 4, 2021

View reviewed changes

eric-ch mentioned this pull request Jan 4, 2021

qemu: small log output improvments #1384

Merged

crogers1 reviewed Jan 6, 2021

View reviewed changes

recipes-security/refpolicy/refpolicy-mcs-2.%/policy/modules/services/glass.fc Outdated Show resolved Hide resolved

crogers1 reviewed Jan 6, 2021

View reviewed changes

recipes-security/refpolicy/refpolicy-mcs-2.%/policy/modules/services/ivcd.fc Outdated Show resolved Hide resolved

eric-ch force-pushed the vglass branch from 4f925c4 to 01d77a9 Compare January 7, 2021 20:46

eric-ch force-pushed the vglass branch from 01d77a9 to 34387d1 Compare January 8, 2021 16:06

eric-ch force-pushed the vglass branch from 34387d1 to 4e2a95d Compare January 26, 2021 23:26

eric-ch force-pushed the vglass branch from 4e2a95d to 619f8b2 Compare April 15, 2021 20:31

dpsmith approved these changes Apr 28, 2021

View reviewed changes

jandryuk approved these changes May 4, 2021

View reviewed changes

Eric Chanudet added 6 commits May 20, 2021 17:24

xenfb2: retire deprecated pv driver

dd6c87b

replaced by openxtfb or xenfb. Signed-off-by: Eric Chanudet <[email protected]>

fbtap: Retire splash-screen tool

a824a0c

Deprecated, vGlass can handle splash-screen without help from the kernel. Signed-off-by: Eric Chanudet <[email protected]>

surfman: retire vm display manager

ec2ebbb

Replaced with vglass. So long Surface Manager. Signed-off-by: Eric Chanudet <[email protected]>

surfman: remove SELinux policy

0c66f04

Signed-off-by: Eric Chanudet <[email protected]>

inputserver: retire project

8f82bbb

Replaced by vglass. Signed-off-by: Eric Chanudet <[email protected]>

inputserver: remove selinux policy definitions

12213df

Signed-off-by: Eric Chanudet <[email protected]>

Eric Chanudet added 25 commits May 20, 2021 17:24

xenmgr: recipe style and format

36f7bc3

https://www.openembedded.org/wiki/Styleguide Add quotes and use OE variables. Remove deprecated Rpc. Signed-off-by: Eric Chanudet <[email protected]>

dmbus: retire the project

f2e6def

The only consumers of dmbus were surfman and input-server. Both are deprecated, dmbus can follow. Signed-off-by: Eric Chanudet <[email protected]>

libivc: version preference in images

2c85f96

libivc2/ivcdaemon is only suited for dom0. The Linux stubdomain should not add QT or C++ runtime dependencies to keep the footprint manageable, hence use the ivc module through libivc. Signed-off-by: Eric Chanudet <[email protected]>

qemu-dm: refresh patch-queue

9c67f4f

Refresh the patch-queue for fuzz introduced by vglass changes in it. Signed-off-by: Eric Chanudet <[email protected]>

update-module: retire recipe

3bfc716

Apparently a wrapper script around depmod. Unused and forgotten. Signed-off-by: Eric Chanudet <[email protected]>

vglass: inherit module-signing

ad00c30

inherit the module-signing bbclass for out-of-tree modules. Signed-off-by: Eric Chanudet <[email protected]>

uivm: switch to openxtfb

81d94b0

Add openxtfb to UIVM. vGlass provides an xf86 Xorg driver for the PV framebuffer openxtfb. Signed-off-by: Eric Chanudet <[email protected]>

xorg-conf: amend UIVM configuration for openxtfb

714bdec

Signed-off-by: Eric Chanudet <[email protected]>

openxtfb: autoload openxtfb in UIVM

78d7a87

openxtfb is the default framebuffer driver in UIVM with vGlass. Autoload the driver. Signed-off-by: Eric Chanudet <[email protected]>

stubdomain: load IVC module

763ec4f

modules.dep is present, modprobe can be used. Add ivc and use modprobe for xen-argo. Signed-off-by: Eric Chanudet <[email protected]>

libedid: retire defunct project

a4d659f

libedid was used by Surfman to emulate EDID values. Signed-off-by: Eric Chanudet <[email protected]>

dom0: bump memory to 768M

e0b801f

vGlass has higher memory requirements. This increase is motivated as vGlass would otherwise fail to drive UIVM with ENOMEM when opening the grant-refs used by the channels with UIVM and other guests. Signed-off-by: Eric Chanudet <[email protected]>

vglass: policy for vglass components

21a012c

SELinux module policies for ivcdaemon, glass and disman components of vglass. Signed-off-by: Eric Chanudet <[email protected]>

libxl: remove display_{res,depth} xenstore nodes

eded591

Remove hunks related to display_depth and display_res passed via XenStore. No components were using these initially. These might be an artefact from before libdmbus. Signed-off-by: Eric Chanudet <[email protected]>

xenmgr: add build dependency on split

401e525

Changes to Xenmgr require the split from Hackage. Signed-off-by: Eric Chanudet <[email protected]>

eric-ch force-pushed the vglass branch from 619f8b2 to 401e525 Compare May 24, 2021 21:43

jandryuk merged commit 8522fd8 into OpenXT:master May 28, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

vglass: replace surfman/input-server with vglass #1382

vglass: replace surfman/input-server with vglass #1382

eric-ch commented Dec 15, 2020 •

edited

Loading

eric-ch commented Dec 18, 2020

jandryuk left a comment

jandryuk Jan 4, 2021

eric-ch Jan 4, 2021

jandryuk Jan 4, 2021

jandryuk left a comment

jandryuk Jan 4, 2021

eric-ch Jan 4, 2021

jandryuk Jan 4, 2021

eric-ch Jan 4, 2021

jandryuk Jan 4, 2021

eric-ch Jan 5, 2021

jandryuk Jan 4, 2021

crogers1 Jan 6, 2021

jandryuk Jan 6, 2021

eric-ch Jan 7, 2021 •

edited

Loading

eric-ch Jan 8, 2021

eric-ch commented Jan 7, 2021 •

edited

Loading

eric-ch commented Jan 8, 2021

jandryuk commented Jan 8, 2021

eric-ch commented Jan 8, 2021

eric-ch commented Jan 8, 2021

jandryuk commented Jan 13, 2021

eric-ch commented Jan 13, 2021

eric-ch commented Jan 22, 2021 •

edited

Loading

eric-ch commented Jan 26, 2021

eric-ch commented Apr 15, 2021

eric-ch commented Apr 26, 2021

dpsmith left a comment

eric-ch commented May 24, 2021

eric-ch commented May 25, 2021 •

edited

Loading

jandryuk commented May 28, 2021

vglass: replace surfman/input-server with vglass #1382

vglass: replace surfman/input-server with vglass #1382

Conversation

eric-ch commented Dec 15, 2020 • edited Loading

Depends on:

eric-ch commented Dec 18, 2020

jandryuk left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jandryuk left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eric-ch Jan 7, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eric-ch commented Jan 7, 2021 • edited Loading

v2

eric-ch commented Jan 8, 2021

jandryuk commented Jan 8, 2021

eric-ch commented Jan 8, 2021

eric-ch commented Jan 8, 2021

v3

jandryuk commented Jan 13, 2021

eric-ch commented Jan 13, 2021

eric-ch commented Jan 22, 2021 • edited Loading

eric-ch commented Jan 26, 2021

V4

eric-ch commented Apr 15, 2021

V5

eric-ch commented Apr 26, 2021

dpsmith left a comment

Choose a reason for hiding this comment

eric-ch commented May 24, 2021

V6

eric-ch commented May 25, 2021 • edited Loading

jandryuk commented May 28, 2021

eric-ch commented Dec 15, 2020 •

edited

Loading

eric-ch Jan 7, 2021 •

edited

Loading

eric-ch commented Jan 7, 2021 •

edited

Loading

eric-ch commented Jan 22, 2021 •

edited

Loading

eric-ch commented May 25, 2021 •

edited

Loading