Bug 94677

Summary: xorg-server 1.18.2: missing mouse pointer when coming back from screen lock
Product: xorg Reporter: Denis Dupeyron <calchan>
Component: Driver/intelAssignee: Chris Wilson <chris>
Status: RESOLVED FIXED QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: medium CC: 8, andrey.vihrov, andyrtr, bugs.freedesktop.org, bugzilla, ckrzen, giantcone, hanno, jefdriesen, jefferym, koterpillar, michel, mvoorhis, nicolasmaia, sitsofe, thomas-lange2, tobias.pal, unhammer+dill
Version: unspecified   
Hardware: Other   
OS: All   
i915 platform: i915 features:
Description Flags
dmesg and Xorg.0.log
xf86-video-intel/tools/cursor output none

Description Denis Dupeyron 2016-03-23 20:50:26 UTC
Created attachment 122506 [details]
dmesg and Xorg.0.log

I am a Gentoo user and have the exact same problem as a Debian user which he describes at [1]. His words:

"The mouse pointer simply disappears when I want to log into the system after
locking the screen. After restarting the Xserver, everything backs to normal,
but when I lock the screen again, I won't see the pointer after login. I'm
using lightdm as DM.

I can get the mouse pointer back also when I switch to TTY using ctl-alt-f1 and
then to xsession via ctrl-alt-f7."

The issue appeared between 1.18.1 and 1.18.2, so I bisected. Reverting commit b04767c84deafc44993723add4b1c5163fc11711 by Michel Dänzer solves this particular issue.


[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=818325
Comment 1 Michel Dänzer 2016-03-24 08:41:31 UTC
Does this also happen using the modesetting driver instead of the intel driver?
Comment 2 Denis Dupeyron 2016-03-24 17:04:05 UTC
(In reply to Michel Dänzer from comment #1)
> Does this also happen using the modesetting driver instead of the intel
> driver?

Good catch. I tried and it does not happen with the modesetting driver.
Comment 3 Chris Wilson 2016-03-25 21:02:58 UTC
First machine I looked at worked fine. Could you please run xf86-video-intel/tools/cursor and cat /sys/kernel/debug/dri/0/i915_display_info when the cursor is invisible coming back from screenlock?
Comment 4 Denis Dupeyron 2016-03-26 17:19:19 UTC
Created attachment 122577 [details]

This is a cat of /sys/kernel/debug/dri/0/i915_display_info when the cursor is invisible. The before and after screenlock diff is this:

--- before	2016-03-26 11:12:18.744047285 -0600
+++ after	2016-03-26 11:13:28.722867306 -0600
@@ -6,11 +6,11 @@
 	fb: 50, pos: 0x0, size: 1280x800
 	encoder 26: type: LVDS-26, connectors:
 		connector 25: type: LVDS-1, status: connected, mode:
-		id 0:"1280x800" freq 60 clock 82060 hdisp 1280 hss 1332 hse 1396 htot 1672 vdisp 800 vss 803 vse 806 vtot 818 type 0x48 flags 0xa
-	cursor visible? yes, position (365, 188), size 64x64, addr 0x00816000, active? yes
+		id 0:"1280x800" freq 60 clock 82060 hdisp 1280 hss 1332 hse 1396 htot 1672 vdisp 800 vss 803 vse 806 vtot 818 type 0x0 flags 0xa
+	cursor visible? no, position (0, 0), size 0x0, addr 0x00000000, active? no
 	No scalers available on this platform
 	--Plane id 22: type=PRI, crtc_pos=   0x   0, crtc_size=1280x 800, src_pos=0.0000x0.0000, src_size=1280.0000x800.0000, format=XR24 little-endian (0x34325258), rotation=0 (0x00000001)
-	--Plane id 23: type=CUR, crtc_pos= 365x 188, crtc_size=  64x  64, src_pos=0.0000x0.0000, src_size=64.0000x64.0000, format=AR24 little-endian (0x34325241), rotation=0 (0x00000001)
+	--Plane id 23: type=CUR, crtc_pos=   0x   0, crtc_size=   0x   0, src_pos=0.0000x0.0000, src_size=0.0000x0.0000, format=N/A, rotation=0 (0x00000001)
 	underrun reporting: cpu=yes pch=no 
 Connector info
Comment 5 Andrey Vihrov 2016-04-19 20:10:21 UTC
Created attachment 123068 [details]
xf86-video-intel/tools/cursor output

To add to the previous comment, running xf86-video-intel/tools/cursor when the pointer is invisible prints a correct pointer image and dimensions.
Comment 6 nicolasmaia 2016-04-27 10:44:49 UTC
For reference: duplicate bug #95132
Comment 7 Michel Dänzer 2016-04-27 14:03:47 UTC
*** Bug 95132 has been marked as a duplicate of this bug. ***
Comment 8 Sitsofe Wheeler 2016-05-02 05:55:50 UTC
I'm seeing this too with XUbuntu 16.04. Here's a potential downstream bug https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/1568604 .
Comment 9 Tim Starling 2016-08-09 05:09:55 UTC
I'm affected by this bug and I'm trying to figure out if there's anything I can do to help get it fixed. For example, I see the cursor-related functions in sna_display.c already have a fair amount of debug logging. Would it help if I compiled the Intel driver with --enable-debug=full and then captured the log while the bug was occurring?
Comment 10 Thomas Lange 2016-08-10 19:28:42 UTC
(In reply to Tim Starling from comment #9)
> I'm affected by this bug and I'm trying to figure out if there's anything I
> can do to help get it fixed.

This bug is already fixed (at least for me). It went away simply by upgrading my Arch Linux installation. But I cannot say for sure which package was the culprit.

Package versions:
xorg-server: 1.18.4
xf86-video-intel: 2.99.917+691+ga77397a
lightdm: 1.18.2
Comment 11 Tim Starling 2016-08-11 05:58:13 UTC
I looked into this with gdb, with the stock packages from Ubuntu 16.04 since I have a reliable reproduction with that version. Specifically, xserver-xorg-core	2:1.18.3-1ubuntu2.3 and xserver-xorg-video-intel 2:2.99.917+git20160325-1ubuntu1.

The cursor is actually restored successfully after unlock, but then it disappears again when sna_mode_disable_secondary_planes() is called, specifically during the LOCAL_IOCTL_MODE_SETPLANE ioctl. 

After this, there is no chance of the cursor being restored, since sna_crtc->cursor remains non-null, so further attempts to restore it hit the "skipping cursor already show on CRTC" case.

So, there's a good chance this bug was fixed by f1c757e4518f6835bbff6c940269a5c6be75f202, since the commit message claims to modify sna_mode_disable_secondary_planes() to "ignore the secondary cursor planes".
Comment 12 Tim Starling 2016-08-11 23:55:28 UTC
Confirmed by patching the xserver-xorg-video-intel package source with that commit and also its dependency 00a3adaf43640b9aaa84b8cb98c1f2f227686689 "sna: Record all sprite planes reported by the kernel".

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.