Created attachment 145774 [details] logs So. I have big problems with nouveau driver or implementation of nouveau driver (by Debian). I emailed debian-x@lists.debian.org team but have heard nothing back. Straight out the box, some graphical features do not work / are buggy / might hang the system - depending on which DE you use and what eye-candy it has enabled by default. XFCE/LXQT appear unaffected yet KDE Plasma is producing many graphical anomalies/artefacts and gdm did hang. In every DE and version of Debian since Jessie, 360 degree videos (in browser) just display noise. This behaviour changes when using the proprietary 304xx nvidia driver, but the videos still do not function properly - though you do now see the correct video patchwork the 360deg video is generated from. In trying to exercise this problem, I discovered glxgears works fine, yet glmark2 will always crash with a segmentation fault between tests 2 and 3. With nvidia driver installed, glmark2 fails to start at all and gives this :- libGL error: No matching fbConfigs or visuals found libGL error: failed to load driver: swrast ** GLX does not support GLX_EXT_swap_control or GLX_MESA_swap_control! X Error of failed request: BadValue (integer parameter out of range for operation) Major opcode of failed request: 155 (GLX) Minor opcode of failed request: 24 (X_GLXCreateNewContext) Value in failed request: 0x0 Serial number of failed request: 39 Current serial number in output stream: 40 Might be worth noting that after installation of the proprietary driver apt was reporting:- ldconfig: /usr/lib/x86_64-linux-gnu/libGL.so.1 is not a symbolic link I had a brief fling with Piglit. It crashed however. And when trying to use valgrind it looked like it was going to take over a day. Attached are a Xorg.0.log (should module "nv" be being loaded?) and a dmesg output - both from a Debian Stretch install. I have concurrent installs of Buster and Wheezy I can fire up if it will help. The issue might not exist (or is different) in Wheezy. The 360 degree videos after a fresh install look like the other releases after the proprietary driver has been installed, ie I can see the patchwork video rather than just noise.
You say lots of things here... what's the actual problem? That there are non-descript bugs in nouveau that can lead to hangs? That's not news. Is there a specific thing you're trying to make work but doesn't? You mention that xfce works fine - use that. KDE / Gnome will never work with this hardware using nouveau without nuking GL (which you can do by sticking "export LIBGL_ALWAYS_SOFTWARE=1" into your /etc/profile and/or uninstalling nouveau_dri.so).
Created attachment 145775 [details] Xorg.0.log
(In reply to Ilia Mirkin from comment #1) > You say lots of things here... what's the actual problem? That there are > non-descript bugs in nouveau that can lead to hangs? That's not news. > > Is there a specific thing you're trying to make work but doesn't? You > mention that xfce works fine - use that. KDE / Gnome will never work with > this hardware using nouveau without nuking GL (which you can do by sticking > "export LIBGL_ALWAYS_SOFTWARE=1" into your /etc/profile and/or uninstalling > nouveau_dri.so). I'd very much like to watch 360 degree videos.
It is a feature that used to work on the hardware, and now doesn't, and I have a hunch the reason for the proprietary driver not installing properly is related.
(In reply to Doobz from comment #4) > It is a feature that used to work on the hardware, and now doesn't, and I > have a hunch the reason for the proprietary driver not installing properly > is related. Sorry, I have no idea what a 360 degree video is. Can you tell me more about how one might go about reproducing the issue?
(In reply to Ilia Mirkin from comment #5) > (In reply to Doobz from comment #4) > > It is a feature that used to work on the hardware, and now doesn't, and I > > have a hunch the reason for the proprietary driver not installing properly > > is related. > > Sorry, I have no idea what a 360 degree video is. Can you tell me more about > how one might go about reproducing the issue? Videos you can rotate the viewing direction with the mouse while they play. Search 360 video on YouTube.
(In reply to Doobz from comment #6) > (In reply to Ilia Mirkin from comment #5) > > (In reply to Doobz from comment #4) > > > It is a feature that used to work on the hardware, and now doesn't, and I > > > have a hunch the reason for the proprietary driver not installing properly > > > is related. > > > > Sorry, I have no idea what a 360 degree video is. Can you tell me more about > > how one might go about reproducing the issue? > > Videos you can rotate the viewing direction with the mouse while they play. > Search 360 video on YouTube. How are you playing these back? Chrome blacklists nouveau for accel, I doubt the nv4b you have provides high enough GL for firefox to do anything too useful...
(In reply to Ilia Mirkin from comment #7) > (In reply to Doobz from comment #6) > > (In reply to Ilia Mirkin from comment #5) > > > (In reply to Doobz from comment #4) > > > > It is a feature that used to work on the hardware, and now doesn't, and I > > > > have a hunch the reason for the proprietary driver not installing properly > > > > is related. > > > > > > Sorry, I have no idea what a 360 degree video is. Can you tell me more about > > > how one might go about reproducing the issue? > > > > Videos you can rotate the viewing direction with the mouse while they play. > > Search 360 video on YouTube. > > How are you playing these back? Chrome blacklists nouveau for accel, I doubt > the nv4b you have provides high enough GL for firefox to do anything too > useful... Just with the default browser. Firefox and maybe Iceweael. I have a laptop with a GF 8200 gpu, and it does not exhibit this bug. I don't believe it has any h/w acceleration enabled tbh and didn't display a desktop after trying to install the 340xx driver, but that's probably a subject for a different bug report.
So I went ahead and installed Sabayon. Same behaviour - garbled 360 degree videos in Firefox. I did try flipping layers.acceleration.disabled = true in about:config but it made no difference. Does nouveau need any firmware for my card? I thought it didn't, but looking here it appears it might. https://nouveau.freedesktop.org/wiki/VideoAcceleration/
(In reply to Doobz from comment #9) > Does nouveau need any firmware for my card? I thought it didn't, but looking > here it appears it might. > https://nouveau.freedesktop.org/wiki/VideoAcceleration/ It would need it for hypothetical VP1 support. However there is no software support for the VP1 decoding engine, and according to all reports, it didn't make anything actually faster (on Windows). The MPEG decoding engine should work fine (via XvMC ... IIRC VDPAU didn't quite work, and XvMC is a better match anyways). This is all for accelerated video decoding, which is not in play for the videos you're talking about.
I don't know if you were able to deduce anything tangible from the valgrind log. It crossed my mind that glmark2 might by default not run on such old hardware. Also, the 360 degree video could be a bug in Firefox, like you mentioned. I can download the garbled videos and watch successfully in vlc.
Though I struggle to imagine a bug in FF can manifest itself differently on different gpu architectures.
I didn't notice these in dmesg before :~# dmesg | grep fail [ 0.701087] acpi PNP0A08:00: _OSC failed (AE_NOT_FOUND); disabling ASPM [ 0.836978] pci 0000:01:00.0: BAR 6: failed to assign [mem size 0x00020000 pref] And checking the desktop ( integrated GeForce 7100 ) I get [ ..] acpi PNP0A08:00: _OSC failed (AE_SUPPORT); disabling ASPM Same, but slightly different.
Ok, so there's this. https://www.linuxquestions.org/questions/linux-kernel-70/kernel-fails-to-assign-memory-to-pcie-device-4175487043/ I don't know if you can tell, but I'm a little out of my depth here. ;)
Quick update:- glmark2 runs fine on geforce 8200 laptop But on the geforce 7100 (NV63) desktop it produces segmentation fault at same point as on my 7600GO laptop. I can try to upgrade from mesa 18.3.6 to 19.2.1-1 if you think it's worth doing.
(In reply to Doobz from comment #15) > Quick update:- > > glmark2 runs fine on geforce 8200 laptop G80+ support for nouveau is markedly better than pre-G80 (aka GeForce 8000 is the cutoff). The hardware changed substantially for DX10, and has little in common with its DX9 predecessors. > > But on the geforce 7100 (NV63) desktop it produces segmentation fault at > same point as on my 7600GO laptop. > > I can try to upgrade from mesa 18.3.6 to 19.2.1-1 if you think it's worth > doing. Depending on the issue, it may or may not be worth it. Is there a backtrace from the crash?
(In reply to Ilia Mirkin from comment #16) > (In reply to Doobz from comment #15) > > Quick update:- > > > > glmark2 runs fine on geforce 8200 laptop > > G80+ support for nouveau is markedly better than pre-G80 (aka GeForce 8000 > is the cutoff). The hardware changed substantially for DX10, and has little > in common with its DX9 predecessors. > > > > > But on the geforce 7100 (NV63) desktop it produces segmentation fault at > > same point as on my 7600GO laptop. > > > > I can try to upgrade from mesa 18.3.6 to 19.2.1-1 if you think it's worth > > doing. > > Depending on the issue, it may or may not be worth it. Is there a backtrace > from the crash? Are backtraces (only) produced if "core dumped" is reported? If so then probably not.. Can look into getting one if it's possible.
I mean like run glmark in gdb, and when it crashes, do a "bt". Make sure you have symbols available.
Not managed to do the backtrace yet, but thought I'd attach dmesg from 7100-based desktop. Had it sitting cycling gl screensavers for a while. Lots of these:- [ 45.639455] nouveau 0000:00:10.0: bus: MMIO write of 00540001 FAULT at 00b000 And some of these:- [ 825.282398] perf: interrupt took too long (5006 > 4996), lowering kernel.perf_event_max_sample_rate to 39750 In other news, same machine hard locked as soon as I tried to play an local video file.
Created attachment 145794 [details] dmesg output from NV63 GF7100 desktop
b000 range is the media range. Just remove dri/libvdpau_nouveau.so -- I'm sure that's not helping. And any similar va-api thing.
Created attachment 145795 [details] Backtrace glmark2 segfault nv4b mesa1836 Let me know if I messed up. And thanks for taking the time to look into this.
Looks like a crash in glmark2 itself. Probably some unexpected condition...
Turns out opensuse tumbleweed has mesa 19.1.7 and a packaged glmark2 Framerate much slower than on previous kernel/mesa. Only getting ~130 vs >200 previously. Identical graphical glitch on laptop and desktop with Suse Welcome window. The only shared dmesg oddities are [ 0.202564] acpi PNP0A08:00: _OSC failed (AE_NOT_FOUND); disabling ASPM [ ] acpi PNP0A08:00: _OSC failed (AE_SUPPORT); disabling ASPM ..and the hpet warnings on both of them [ 1670.140263] CE: hpet increased min_delta_ns to 30172 nsec I will attach both suse dmesg logs.
Created attachment 145823 [details] dmesg from Acer5685 GeForce 7600GO, opensuse Mesa 19.1.7
Created attachment 145824 [details] dmesg from HP nvidia7100 opensuse mesa19.1.7
I've been looking at the interrupts, and pretty sure something is amiss. nvkm does tick away quite rapidly 50+/sec, but it's by no means the worst. cat /proc/interrupts CPU0 CPU1 0: 884041 0 IO-APIC 2-edge timer 1: 943 0 IO-APIC 1-edge i8042 8: 0 1 IO-APIC 8-edge rtc0 9: 0 71944 IO-APIC 9-fasteoi acpi 12: 0 407479 IO-APIC 12-edge i8042 14: 15003 0 IO-APIC 14-edge ata_piix 15: 0 0 IO-APIC 15-edge ata_piix 16: 4422 0 IO-APIC 16-fasteoi yenta, uhci_hcd:usb5 17: 0 0 IO-APIC 17-fasteoi mmc0, mmc1 18: 0 184 IO-APIC 18-fasteoi uhci_hcd:usb4 19: 0 0 IO-APIC 19-fasteoi uhci_hcd:usb3, i801_smbus 20: 0 2 IO-APIC 20-fasteoi firewire_ohci 23: 0 3 IO-APIC 23-fasteoi uhci_hcd:usb1, ehci_hcd:usb2 29: 89581 51828 PCI-MSI 524288-edge nvkm 30: 0 518 PCI-MSI 442368-edge snd_hda_intel:card0 31: 54496 31631 PCI-MSI 2621440-edge iwl3945 32: 0 3 PCI-MSI 2097152-edge enp4s0 NMI: 7 8 Non-maskable interrupts LOC: 117697 732513 Local timer interrupts SPU: 0 0 Spurious interrupts PMI: 7 8 Performance monitoring interrupts IWI: 157032 144987 IRQ work interrupts RTR: 0 0 APIC ICR read retries RES: 99916 105307 Rescheduling interrupts CAL: 43819 44884 Function call interrupts TLB: 93370 93058 TLB shootdowns TRM: 0 0 Thermal event interrupts THR: 0 0 Threshold APIC interrupts DFR: 0 0 Deferred Error APIC interrupts MCE: 0 0 Machine check exceptions MCP: 9 9 Machine check polls This was after 45mins. I was chasing interrupt issues on windows on this laptop. Never got it sussed. It was fine in safe mode, but not booted normally. Looking at the desktop(hp gf7100), it has no massive number on line 0 timer, and no acpi ones either. Nvkm is ticking at similar rate to laptop, but there is one line going up faster. Ill leave it on for a similar time and post it here for comparison.
> cat /proc/interrupts CPU0 CPU1 0: 8 0 IO-APIC 2-edge timer 1: 0 438 IO-APIC 1-edge i8042 8: 1 0 IO-APIC 8-edge rtc0 9: 0 0 IO-APIC 9-fasteoi acpi 12: 161488 0 IO-APIC 12-edge i8042 14: 0 0 IO-APIC 14-edge pata_amd 15: 0 0 IO-APIC 15-edge pata_amd 19: 0 2 IO-APIC 19-fasteoi firewire_ohci 20: 92448 0 IO-APIC 20-fasteoi nvkm 21: 0 9323 IO-APIC 21-fasteoi ahci[0000:00:0e.0] 22: 143237 0 IO-APIC 22-fasteoi ehci_hcd:usb1, snd_hda_intel:card0 23: 0 0 IO-APIC 23-fasteoi ohci_hcd:usb2 27: 803 2879 PCI-MSI 245760-edge enp0s15 NMI: 124 113 Non-maskable interrupts LOC: 426946 389908 Local timer interrupts SPU: 0 0 Spurious interrupts PMI: 124 113 Performance monitoring interrupts IWI: 148823 138382 IRQ work interrupts RTR: 0 0 APIC ICR read retries RES: 6984 6476 Rescheduling interrupts CAL: 2349 1588 Function call interrupts TLB: 565 358 TLB shootdowns TRM: 0 0 Thermal event interrupts THR: 0 0 Threshold APIC interrupts DFR: 0 0 Deferred Error APIC interrupts MCE: 0 0 Machine check exceptions MCP: 10 10 Machine check polls HYP: 0 0 Hypervisor callback interrupts HRE: 0 0 Hyper-V reenlightenment interrupts HVS: 0 0 Hyper-V stimer0 interrupts ERR: 1 Given that there's no usb devices plugged in I'm a little suspicious of 22: 143237 0 IO-APIC 22-fasteoi ehci_hcd:usb1,
So.. My laptop's interrupt issues disappear when booting with acpi=off. So did nouveau however. Have tried a few other boot parameters, including acpi_apic_instance=2 as suggested in dmesg, but no joy. Could these ACPI warnings be hinting at something? [ 20.414011] i801_smbus 0000:00:1f.3: SMBus using PCI interrupt [ 20.416862] ACPI Warning: SystemIO range 0x0000000000001028-0x000000000000102F conflicts with OpRegion 0x0000000000001000-0x000000000000107F (\PMIO) (20190703/utaddress-204) [ 20.416870] ACPI: If an ACPI driver is available for this device, you should use it instead of the native driver [ 20.416873] ACPI Warning: SystemIO range 0x00000000000011B0-0x00000000000011BF conflicts with OpRegion 0x0000000000001180-0x00000000000011BB (\GPIO) (20190703/utaddress-204) [ 20.416878] ACPI: If an ACPI driver is available for this device, you should use it instead of the native driver [ 20.416879] ACPI Warning: SystemIO range 0x0000000000001180-0x00000000000011AF conflicts with OpRegion 0x0000000000001180-0x00000000000011BB (\GPIO) (20190703/utaddress-204) [ 20.416883] ACPI: If an ACPI driver is available for this device, you should use it instead of the native driver [ 20.416884] lpc_ich: Resource conflict(s) found affecting gpio_ich
Sorry to spam, but I did notice partially visible text in the garbled stuff FF was displaying instead of video. Like there is some memory overlap/confusion somewhere.
Some good news - irq 0 timer interrupts fixed by booting with processor.nocst=1
..and the acpi ones with acpi_mask_gre=0x19. There's a bunch of other one's to try when I'm bored.. :) Back to the 360 degree vid thing, On a fresh boot into debian, the mess coming out of Firefox included the bottom half of what appeared to be a suse filebrowser window I had open the day before! I only just remembered to look in .xsession.errors. It is indeed logged there along with some others. Cleaned up .xsessions from fresh suse tumbleweed on NV63 :- /usr/bin/iceauth: creating new authority file /run/user/1000/ICEauthority (xfce4-session:1436): xfce4-session-WARNING **: 09:47:57.705: xfsm_manager_load_session: Something wrong with /home/doobz/.cache/sessions/xfce4-session-debian:0, Does it exist? Permissions issue? gnome-keyring-daemon: insufficient process capabilities, insecure memory might get used ** Message: 09:47:57.797: couldn't access control socket: /run/user/1000/keyring/control: No such file or directory (xfsettingsd:1522): xfsettingsd-WARNING **: 09:48:05.059: Property "/Gdk/WindowScalingFactor" has an unsupported type "guint". ** (wrapper-2.0:1545): WARNING **: 09:48:06.348: No outputs have backlight property (wrapper-2.0:1550): GLib-GIO-CRITICAL **: 09:48:06.704: g_file_new_for_path: assertion 'path != NULL' failed (wrapper-2.0:1550): GLib-GIO-CRITICAL **: 09:48:06.704: g_file_monitor_file: assertion 'G_IS_FILE (file)' failed (wrapper-2.0:1550): GLib-GObject-WARNING **: 09:48:06.704: invalid (NULL) pointer instance (wrapper-2.0:1550): GLib-GObject-CRITICAL **: 09:48:06.705: g_signal_connect_data: assertion 'G_TYPE_CHECK_INSTANCE (instance)' failed (wrapper-2.0:1550): Gtk-WARNING **: 09:48:06.705: Attempting to add a widget with type GtkToggleButton to a container of type XfcePanelPlugin, but the widget is already inside a container of type XfcePanelPlugin, please remove the widget from its existing container first. gnome-keyring-daemon: insufficient process capabilities, insecure memory might get used (wrapper-2.0:1543): Gtk-WARNING **: 09:48:16.508: Negative content width -3 (allocation 1, extents 2x2) while allocating gadget (node button, owner GtkToggleButton) (wrapper-2.0:1545): Gtk-WARNING **: 09:48:16.801: Negative content width -3 (allocation 1, extents 2x2) while allocating gadget (node button, owner PowerManagerButton) (wrapper-2.0:1544): Gtk-WARNING **: 09:48:16.810: Negative content width -3 (allocation 1, extents 2x2) while allocating gadget (node button, owner PulseaudioButton) (wrapper-2.0:1550): Gtk-WARNING **: 09:48:16.822: Negative content width -3 (allocation 1, extents 2x2) while allocating gadget (node button, owner GtkToggleButton) gnome-keyring-daemon: insufficient process capabilities, insecure memory might get used [Child 2152, MediaDecoderStateMachine #1] WARNING: Decoder=7f3103930800 Decode error: NS_ERROR_DOM_MEDIA_FATAL_ERR (0x806e0005) - RefPtr<mozilla::MozPromise<RefPtr<mozilla::MediaTrackDemuxer::SamplesHolder>, mozilla::MediaResult, true> > mozilla::MediaSourceTrackDemuxer::DoGetSamples(int32_t): manager is detached.: file /home/abuild/rpmbuild/BUILD/firefox-69.0.3/dom/media/MediaDecoderStateMachine.cpp, line 3309 [Child 2152, MediaDecoderStateMachine #1] WARNING: Decoder=7f310IA_FATAL_ERR (0x806e0005) - RefPtr<mozilla::MozPromise<RefPtr<mozilla::MediaTrackDemuxer::SamplesHolder>, mozilla::MediaResult, true> > mozilla::MediaSourceTrackDemuxer::DoGetSamples(int32_t): manager is detached.: file /home/abuild/rpmbuild/BUILD/firefox-69.0.3/dom/media/MediaDecoderStateMachine.cpp, line 3309 [Parent 1979, Gecko_IOThread] WARNING: pipe error (61): Connection reset by peer: file /home/abuild/rpmbuild/BUILD/firefox-69.0.3/ipc/chromium/src/chrome/common/ipc_channel_posix.cc, line 358 Invalid MIT-MAGIC-COOKIE-1 keyInvalid MIT-MAGIC-COOKIE-1 key
-- GitLab Migration Automatic Message -- This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/xorg/driver/xf86-video-nouveau/issues/510.
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.