| Summary: | Online Upgrade of MGA-5.1 with AMD video (e.g., "Kaveri" APUs) leads to Plasma session failure. | ||
|---|---|---|---|
| Product: | Mageia | Reporter: | Rick Stockton <rickstockton> |
| Component: | RPM Packages | Assignee: | KDE maintainers <kde> |
| Status: | RESOLVED OLD | QA Contact: | |
| Severity: | critical | ||
| Priority: | Normal | CC: | mageia, marja11, rickstockton, tmb |
| Version: | Cauldron | Keywords: | NEEDINFO |
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Source RPM: | CVE: | ||
| Status comment: | |||
|
Description
Rick Stockton
2017-03-16 11:37:44 CET
Rick Stockton
2017-03-16 16:26:41 CET
CC:
(none) =>
rickstockton
Rick Stockton
2017-03-16 16:28:14 CET
CC:
(none) =>
tmb
Rick Stockton
2017-03-16 16:32:38 CET
CC:
(none) =>
mageia, marja11 This bug report replaces bug 20452. The remaining problem is much "smaller" in scope after upgrade to kernel 4.9.15-desktop-1.mga6 (SMP Wed Mar 15 22:14:13 UTC 2017 x86_64). I can run multiple 'Gnome' and 'Ice' sessions in sequence, on X or Wayland; and I can run multiple 'Plasma' sessions among 'Gnome' and 'Ice' sessions, as long as I avoid 'Gnome-Wayland'. I SWAG: It seems that after Gnome-Wayland and return to the session-choosing DM, it seems that X-Server is left in a state which hangs on the next attempt to initiate Qt/KDE. The problem is definitely 'lower down' than Plasma. I don't whether Kwin is trying to use Wayland or X, and I don't know if Kwin/Qt should be responsible for requesting some kind of "reset" to a standard set of X11 options. (Perhaps that job should be done upon initializing Mageia's DM "session chooser", even though just one kind of session can't handle the current behavior.) My hardware and software may also be an important factor: In this bug report, my configuration is AMD A10-Kaveri built-in video, with 8 GPU processors on the CPU die. I run AMDGPU, all default, with no other specifications in Xorg.conf. Early next week, I will also test with a separate video card (Also AMD R7, I can theoretically run 'crossfire' if Linux drivers can support that.) This will allow us to see if the X-Server "issue" occurs only with AMD A-series, or also occurs when AMD physicial cards are used. Next comment: I will now test whether running a "Gnome on Wayland" session, followed by "Gnome on X", restores the Xorg Server to a state in which "Plasma" can run successfully. Weird. Upon reboot, I am unable to reproduce this 'problem' at all. Closing as 'WORKSFORME'. Status:
NEW =>
RESOLVED New 'amdgpu' driver: The 'plasma problem is back, and much worse. (Gnome-Wayland is spectacularly fast, but I hate hate the DE.) After choosing session "Plasma" leaves "runlevel 5" in a deadly loop of partial X-failure and some kind of ineffective X-restart. (Mixed with console messages "CIFS VS: cifs_mount failed w/return code = -13"). No idea why why the "restored" Xorg is re-trying Plasma immediately, rather than restoring an un-hung DM. On my PC, the rapid re-appearance of the "CIFS" messages prevents root login of VT-4. I am always forced to crash the box. I have given up, until my physical R7 arrives for comparision testing. (Re-open, this is no longer "works-for-me". I am not marking with unusual severity or priority, because I'm unable (and tired) of working on it - until that card arrives. Because the problem seems entirely due to the new amdgpu driver stack, this might be upstream. Resolution:
WORKSFORME =>
(none) I note for next week: there is no "crossfire support" within the current Linux Drivers, it will be with one OR the other. Summary:
A "plasma session" request will hang the X11 Server if run AFTER a Gnome-Wayland session. =>
AMD-A8 'Kaveri" integrated GPUs: A "plasma session" request w/ driver amdgpu (1.3.0) pushes X-Server into some kind of "hard loop". These problems occur only with 'UPGRADE' from MGA-5.1. I have just done a complete "install" from STA-2 "live" KDE/Plasma, and the compositing desktop works. (At least, to the extent that Plamsa/KF-5 supports "features" of KDE-4: it last lost some capabilities.) Upgrade of my 5.1 from configuration from the "complete" STA-2 DVD fails pretty hard (unhandled error result, at an early stage of "upgrading"). That's a different problem.Online Upgrade completes "successfully, but experiences THIS problem after the install. SWAG: Perhaps something in Plasma 5.x use of qtScreen() (perhaps combined with bad parameters to X, e.g. xrandr), causes the screen to be hung or "lost" within kwin. When we do a "fresh install", the parameter strings get written fresh. And KDE4-compatible RandR options work - because Kwin4 did a lot more calls to X11 directly. Kwin5, going through qt5 for more things.... And that's why Gnome was mostly immune - it's built on GTK, rather than qt. 'Summary' is revised. Summary:
AMD-A8 'Kaveri" integrated GPUs: A "plasma session" request w/ driver amdgpu (1.3.0) pushes X-Server into some kind of "hard loop". =>
Online Upgrade of MGA-5.1 with AMD video (e.g., "Kaveri" APUs) leads to Plasma session failure. (In reply to Rick Stockton from comment #5) > These problems occur only with 'UPGRADE' from MGA-5.1. > > I have just done a complete "install" from STA-2 "live" KDE/Plasma, and the > compositing desktop works. (At least, to the extent that Plamsa/KF-5 > supports "features" of KDE-4: it last lost some capabilities.) > > Upgrade of my 5.1 from configuration from the "complete" STA-2 DVD fails > pretty hard (unhandled error result, at an early stage of "upgrading"). > That's a different problem.Online Upgrade completes "successfully, but > experiences THIS problem after the install. > > SWAG: Perhaps something in Plasma 5.x use of qtScreen() (perhaps combined > with bad parameters to X, e.g. xrandr), causes the screen to be hung or > "lost" within kwin. When we do a "fresh install", the parameter strings get > written fresh. And KDE4-compatible RandR options work - because Kwin4 did a > lot more calls to X11 directly. Kwin5, going through qt5 for more things.... > And that's why Gnome was mostly immune - it's built on GTK, rather than qt. > > 'Summary' is revised. This bug report confuses me. Is comment #5 + the new summary "Online Upgrade of MGA-5.1 with AMD video (e.g., "Kaveri" APUs) leads to Plasma session failure." all that needs to be known? Or which information in the previous comments is still valid and needed? Severity:
normal =>
critical We will assume the problem was not reproducible, or has been fixed in one of the updates we have released for the reporter's distribution. Users who have experienced this problem are encouraged to upgrade to the latest update of our distribution, and if this issue turns out to still be reproducible in the latest update, please reopen this bug with additional information. Closing as OLD. Status:
REOPENED =>
RESOLVED |