Bug 1496253
| Summary: | Spice/QXL cannot support gnome3 with multi-monitors as it requires hardware acceleration | |||
|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 7 | Reporter: | Frank DeLorey <fdelorey> | |
| Component: | mesa | Assignee: | Dave Airlie <airlied> | |
| Status: | CLOSED ERRATA | QA Contact: | Desktop QE <desktop-qa-list> | |
| Severity: | high | Docs Contact: | ||
| Priority: | high | |||
| Version: | 7.0 | CC: | airlied, bsanford, cfergeau, dblechte, debarshir, djasa, fdelorey, fziglio, jwright, lsurette, marcandre.lureau, mtessun, rduda, rstrode, salmy, srevivo, tpelka, vanhoof, ycui | |
| Target Milestone: | pre-dev-freeze | |||
| Target Release: | --- | |||
| Hardware: | Unspecified | |||
| OS: | Unspecified | |||
| Whiteboard: | ||||
| Fixed In Version: | Doc Type: | If docs needed, set a value | ||
| Doc Text: | Story Points: | --- | ||
| Clone Of: | ||||
| : | 1667301 1667318 (view as bug list) | Environment: | ||
| Last Closed: | 2018-10-30 08:01:02 UTC | Type: | Bug | |
| Regression: | --- | Mount Type: | --- | |
| Documentation: | --- | CRM: | ||
| Verified Versions: | Category: | --- | ||
| oVirt Team: | Virt | RHEL 7.3 requirements from Atomic Host: | ||
| Cloudforms Team: | --- | Target Upstream Version: | ||
| Embargoed: | ||||
| Bug Depends On: | ||||
| Bug Blocks: | 1547138, 1667301, 1667318 | |||
|
Description
Frank DeLorey
2017-09-26 19:21:49 UTC
Customer is using 3 monitors resolutions of 2560x1440, 1440x2560 and 1440x2560. Gnome shell is consuming high CPU in the VMs in RHEV. They are unable to use their VMs in VDI effectively due to Gnome taking resources which should be available to other applications. Additionally they are unable to use 3 monitors effectively in a triple head display (xinerama, not mirrored) with resolutions of 2560x1440, 1440x2560 and 1440x2560. Update from the customer: Most recent comment: On 2017-09-27 08:48:52, Thomas, Greg commented: "This issue is very easy to reproduce. Install RHEL 7.4 Workstation on a new VM hosted by RHEV 3.6. Login using Gnome, the default. Set the display properties to use 2 or more monitors running at least 1920x1080 (and for an additional test, 2560x1440) Open windows to browse the file system. Open a terminal window. Watch top to see gnome-shell CPU running at 100% No special programs needed. No 3D apps. We merely want to move standard OS windows around the desktop. All our VMs are set to Single PCI, either 2 or 4 monitors (note our biggest demands are 3 monitors, not 4). We only changed the Single PCI setting when prompted by support earlier in the ticket. We run all the above resolutions under RHEL6 and Gnome2 without issue. When moving to RHEL7 the issues were immediately apparent. MATE works for 2 monitors, not 3. Besides not being a supported desktop manager, since it doesn't work for 3 monitors, we can't use it for all users. KDE works for the moment but hardware acceleration requirements are in the pipeline and isn't a long term solution since we are a yum update away from breaking the setup. Greg top - 08:31:09 up 1 day, 22:04, 13 users, load average: 2.83, 1.84, 0.96 Tasks: 245 total, 4 running, 241 sleeping, 0 stopped, 0 zombie %Cpu(s): 43.3 us, 4.9 sy, 0.2 ni, 51.0 id, 0.5 wa, 0.0 hi, 0.2 si, 0.0 st KiB Mem : 10070864 total, 687756 free, 4534780 used, 4848328 buff/cache KiB Swap: 3145724 total, 3145676 free, 48 used. 4827820 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 14939 jeff 20 0 3745828 1.473g 50652 S 119.6 15.3 190:08.99 gnome-shell 14856 jeff 20 0 1945740 562028 97712 S 21.6 5.6 18:53.47 chrome 14376 root 20 0 659028 295216 15376 R 16.9 2.9 41:54.21 X 1616 jeff 20 0 1396464 438284 36284 R 13.3 4.4 0:13.49 chrome 14050 jeff 20 0 1089820 142148 37344 S 5.3 1.4 24:47.73 chrome 10538 jeff 20 0 1200108 244844 45412 S 2.3 2.4 4:07.56 chrome 4578 jeff 20 0 1145008 174276 47712 S 1.7 1.7 2:38.94 chrome 15217 jeff 20 0 1154596 174960 23048 S 1.7 1.7 2:37.11 chrome Hi Frank, (In reply to Frank DeLorey from comment #5) > top - 08:31:09 up 1 day, 22:04, 13 users, load average: 2.83, 1.84, 0.96 > Tasks: 245 total, 4 running, 241 sleeping, 0 stopped, 0 zombie > %Cpu(s): 43.3 us, 4.9 sy, 0.2 ni, 51.0 id, 0.5 wa, 0.0 hi, 0.2 si, 0.0 > st > KiB Mem : 10070864 total, 687756 free, 4534780 used, 4848328 buff/cache > KiB Swap: 3145724 total, 3145676 free, 48 used. 4827820 avail Mem > > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > > 14939 jeff 20 0 3745828 1.473g 50652 S 119.6 15.3 190:08.99 > gnome-shell > 14856 jeff 20 0 1945740 562028 97712 S 21.6 5.6 18:53.47 chrome > > 14376 root 20 0 659028 295216 15376 R 16.9 2.9 41:54.21 X > > 1616 jeff 20 0 1396464 438284 36284 R 13.3 4.4 0:13.49 chrome > > 14050 jeff 20 0 1089820 142148 37344 S 5.3 1.4 24:47.73 chrome > > 10538 jeff 20 0 1200108 244844 45412 S 2.3 2.4 4:07.56 chrome > > 4578 jeff 20 0 1145008 174276 47712 S 1.7 1.7 2:38.94 chrome > > 15217 jeff 20 0 1154596 174960 23048 S 1.7 1.7 2:37.11 chrome the CPU usage of gnome-shell looks wrong to me, as already discussed in some mail threads. Did you already open a BZ for gnome-shell to investigate this, as I am not able to reproduce this in my virtual environments with multimonitor setups. Not even with a slow remote connection. In addition you can disable most 3D effects in gnome shell via the Tweak Tool. Still all the rendering is done with opengl, so 3D software rendering is still in the game. Please link the two BZs together if you opened the gnome-shell one. Thanks! Martin (In reply to Martin Tessun from comment #10) > Hi Frank, > > (In reply to Frank DeLorey from comment #5) > > top - 08:31:09 up 1 day, 22:04, 13 users, load average: 2.83, 1.84, 0.96 > > Tasks: 245 total, 4 running, 241 sleeping, 0 stopped, 0 zombie > > %Cpu(s): 43.3 us, 4.9 sy, 0.2 ni, 51.0 id, 0.5 wa, 0.0 hi, 0.2 si, 0.0 > > st > > KiB Mem : 10070864 total, 687756 free, 4534780 used, 4848328 buff/cache > > KiB Swap: 3145724 total, 3145676 free, 48 used. 4827820 avail Mem > > > > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > > > > 14939 jeff 20 0 3745828 1.473g 50652 S 119.6 15.3 190:08.99 > > gnome-shell > > 14856 jeff 20 0 1945740 562028 97712 S 21.6 5.6 18:53.47 chrome > > > > 14376 root 20 0 659028 295216 15376 R 16.9 2.9 41:54.21 X > > > > 1616 jeff 20 0 1396464 438284 36284 R 13.3 4.4 0:13.49 chrome > > > > 14050 jeff 20 0 1089820 142148 37344 S 5.3 1.4 24:47.73 chrome > > > > 10538 jeff 20 0 1200108 244844 45412 S 2.3 2.4 4:07.56 chrome > > > > 4578 jeff 20 0 1145008 174276 47712 S 1.7 1.7 2:38.94 chrome > > > > 15217 jeff 20 0 1154596 174960 23048 S 1.7 1.7 2:37.11 chrome > > the CPU usage of gnome-shell looks wrong to me, as already discussed in some > mail threads. Did you already open a BZ for gnome-shell to investigate this, > as I am not able to reproduce this in my virtual environments with > multimonitor setups. Not even with a slow remote connection. > > In addition you can disable most 3D effects in gnome shell via the Tweak > Tool. Still all the rendering is done with opengl, so 3D software rendering > is still in the game. > > Please link the two BZs together if you opened the gnome-shell one. > > Thanks! > Martin The spice-server is the wrong place for this BZ in any case. Let' close it and find the reproducer first based on comment #6 Just for clarification of my tests (I am writing this update from a RHEL 7.4 VM having 3 displays 1920x1080 (don't have larger monitors) with a slow remote connection (VPN).
Still I don't see any exhaustive gnome-shell CPU usage in top:
top - 14:38:28 up 40 min, 2 users, load average: 0.41, 0.45, 0.28
Tasks: 185 total, 1 running, 184 sleeping, 0 stopped, 0 zombie
%Cpu(s): 2.6 us, 0.4 sy, 0.0 ni, 97.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem : 3873476 total, 1432696 free, 1395376 used, 1045404 buff/cache
KiB Swap: 4063228 total, 4063228 free, 0 used. 2118860 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
2533 mtessun 20 0 2639056 490908 48028 S 7.0 12.7 2:03.25 gnome-shell
1273 root 20 0 496808 146620 32200 S 3.0 3.8 0:40.55 X
3549 mtessun 20 0 2210608 280624 117612 S 0.7 7.2 0:37.30 firefox
107 root 20 0 0 0 0 S 0.3 0.0 0:00.73 kworker/2:1
751 root 20 0 50104 1208 848 S 0.3 0.0 0:00.20 spice-vdagentd
1104 root 20 0 1564404 34368 10148 S 0.3 0.9 0:06.23 python
3186 mtessun 20 0 707228 23172 13964 S 0.3 0.6 0:01.13 gnome-terminal-
3289 mtessun 20 0 157876 2352 1564 R 0.3 0.1 0:01.53 top
3632 mtessun 20 0 1966368 166824 93356 S 0.3 4.3 0:30.43 Web Content
1 root 20 0 193704 6832 4056 S 0.0 0.2 0:02.59 systemd
2 root 20 0 0 0 0 S 0.0 0.0 0:00.01 kthreadd
3 root 20 0 0 0 0 S 0.0 0.0 0:00.00 ksoftirqd/0
5 root 0 -20 0 0 0 S 0.0 0.0 0:00.00 kworker/0:0H
6 root 20 0 0 0 0 S 0.0 0.0 0:00.09 kworker/u128:0
7 root rt 0 0 0 0 S 0.0 0.0 0:00.03 migration/0
$ xrandr --listmonitors
Monitors: 3
0: +*Virtual-0 1920/508x988/261+0+0 Virtual-0
1: +Virtual-1 1920/508x988/261+1920+0 Virtual-1
2: +Virtual-2 1920/508x1015/269+3840+0 Virtual-2
$
Joe Wright from the Desktop SBR has been working this. I am not sure if he has opened a BZ for this yet. I know he was waiting for some data from the customer. Regards, Frank This bug is getting pretty long and confusing. The title says that 3D hardware acceleration is needed. If is needed QXL is designed for 2D and won't never support 3D as would conflicting with Virgl project leading to duplicate effort (for instance having double drivers). If the problem, on the other hand, is something specific to QXL that makes the solution bad this is different. Just to remove QXL from the picture, how are the performance using VGA? If they are more or less the same the issue is software rendering in Mesa and we need a 3D card. If with VGA they are much better than there's something QXL specific and we can work on it. I've put the shm patches into the mesa 7.6 packages, hopefully solving this, however if there are still issues once 7.6 releases please open a new bug and reference this one. Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHSA-2018:3059 This bug was backed out in 7.6.z, not sure if I should reopen or clone |