Created attachment 231508 [details] 510.47.03 It's a "production branch release", so, presumably, we want it. I'll note here that https://download.nvidia.com/XFree86/Linux-x86_64/510.47.03/README/primerenderoffload.html works since 495, so this update brings us (partial?) Optimus support.
Created attachment 231509 [details] legacy-470.103.01
Created attachment 231510 [details] legacy-470.103.01 (Sorry. Wrong field.) Here's a patch introducing x11/nvidia-driver-470 and x11/linux-nvidia-libs-470 for your convenience. 470 is expected to become a legacy branch for Kepler GPUs (495 already doesn't support them), although, curiously enough, I don't see an actual statement from Nvidia anywhere. Eh, whatever.
Not sure how to debug this, but in an optimus configuration I cannot get it to work (I have a 960M GTX) I have this error when trying to setup dual gpu (modesetting + nvidia) [ 41.114] (II) NVIDIA GLX Module 510.47.03 Mon Jan 24 22:35:33 UTC 2022 [ 41.115] (II) NVIDIA: The X server supports PRIME Render Offload. [ 41.116] (EE) [ 41.116] (EE) Backtrace: [ 41.118] (EE) 0: /usr/local/bin/Xorg (?+0x0) [0x41e27a] [ 41.121] (EE) unw_get_proc_name failed: no unwind info found [-10] [ 41.121] (EE) 1: /lib/libthr.so.3 (?+0x0) [0x80092ce0e] [ 41.122] (EE) unw_get_proc_name failed: no unwind info found [-10] [ 41.122] (EE) 2: /lib/libthr.so.3 (?+0x0) [0x80092c3cf] [ 41.124] (EE) 3: ? (?+0x0) [0x7ffffffff003] [ 41.124] (EE) [ 41.124] (EE) Segmentation fault at address 0x0 [ 41.124] (EE) Fatal server error: [ 41.124] (EE) Caught signal 11 (Segmentation fault). Server aborting [ 41.124] (EE) [ 41.124] (EE) Currently I have no issue with the version 495.46, not sure if I messed up the patching through
(In reply to Thibault Payet from comment #3) When exactly does it crash? On start?
Yes it crash at start
(In reply to Thibault Payet from comment #5) I'd check if nvidia-modeset.ko was properly (re)loaded.
(In reply to Alex S from comment #6) I did after a restart, and nvidia-modeset is set to be loaded with kld_list, I will try 510.39.01 to bisect it a little more
(In reply to Thibault Payet from comment #7) It's not like that helps us much. Anyone else wants to test this? (I don't own an Optimus laptop.)
Weirdly enough if I remove any Xorg conf to load nvidia, and thus only using modesetting driver for X, I am still able to run some vulkan application such as vkcube-xcb
Not that surprising. Negotiating OpenGL stuff with Xorg (GLX) is more of a tradition than anything else. EGL might work as well.
(In reply to Alex S from comment #8) Does FreeBSD version of this driver really have support for Optimus? FreeBSD version of README [1] doesn't contain the section you mentioned at Description. [1] http://us.download.nvidia.com/XFree86/FreeBSD-x86_64/510.47.03/README/index.html And unfortunately, version 510.47.03 of this driver segfaults for me, too, on non-Optimus configuration. (Disabling internal GPU on firmware configuration.) Using Quadro P1000. (snip) [ 190.032] (II) Module ABI versions: [ 190.032] X.Org ANSI C Emulation: 0.4 [ 190.032] X.Org Video Driver: 24.1 [ 190.032] X.Org XInput driver : 24.1 [ 190.032] X.Org Server Extension : 10.0 [ 190.033] (--) PCI:*(1@0:0:0) 10de:1cbb:17aa:2262 rev 161, Mem @ 0xa3000000/16777216, 0x60000000/268435456, 0x70000000/33554432, I/O @ 0x00002000/128, BIOS @ 0x????????/65536 [ 190.033] (II) "glx" will be loaded. This was enabled by default and also specified in the config file. [ 190.033] (II) LoadModule: "dbe" [ 190.033] (II) Module "dbe" already built-in [ 190.033] (II) LoadModule: "glx" [ 190.033] (II) Loading /usr/local/lib/xorg/modules/extensions/libglx.so [ 190.037] (II) Module glx: vendor="X.Org Foundation" [ 190.037] compiled for 1.20.13, module version = 1.0.0 [ 190.037] ABI class: X.Org Server Extension, version 10.0 [ 190.037] (II) LoadModule: "extmod" [ 190.037] (II) Module "extmod" already built-in [ 190.037] (II) LoadModule: "nvidia" [ 190.038] (II) Loading /usr/local/lib/xorg/modules/drivers/nvidia_drv.so [ 190.041] (II) Module nvidia: vendor="NVIDIA Corporation" [ 190.041] compiled for 1.6.99.901, module version = 1.0.0 [ 190.041] Module class: X.Org Video Driver [ 190.041] (II) LoadModule: "libinput" [ 190.041] (II) Loading /usr/local/lib/xorg/modules/input/libinput_drv.so [ 190.045] (II) Module libinput: vendor="X.Org Foundation" [ 190.045] compiled for 1.20.13, module version = 0.30.0 [ 190.045] Module class: X.Org XInput Driver [ 190.045] ABI class: X.Org XInput driver, version 24.1 [ 190.045] (II) NVIDIA dlloader X Driver 510.47.03 Mon Jan 24 22:37:17 UTC 2022 [ 190.045] (II) NVIDIA Unified Driver for all Supported NVIDIA GPUs [ 190.046] (--) Using syscons driver with X support (version 2.0) [ 190.046] (--) using VT number 9 [ 190.047] (II) Loading sub module "fb" [ 190.047] (II) LoadModule: "fb" [ 190.047] (II) Loading /usr/local/lib/xorg/modules/libfb.so [ 190.047] (II) Module fb: vendor="X.Org Foundation" [ 190.048] compiled for 1.20.13, module version = 1.0.0 [ 190.048] ABI class: X.Org ANSI C Emulation, version 0.4 [ 190.048] (II) Loading sub module "wfb" [ 190.048] (II) LoadModule: "wfb" [ 190.048] (II) Loading /usr/local/lib/xorg/modules/libwfb.so [ 190.048] (II) Module wfb: vendor="X.Org Foundation" [ 190.048] compiled for 1.20.13, module version = 1.0.0 [ 190.048] ABI class: X.Org ANSI C Emulation, version 0.4 [ 190.048] (II) Loading sub module "ramdac" [ 190.048] (II) LoadModule: "ramdac" [ 190.048] (II) Module "ramdac" already built-in [ 190.049] (WW) VGA arbiter: cannot open kernel arbiter, no multi-card support [ 190.049] (**) NVIDIA(0): Depth 24, (--) framebuffer bpp 32 [ 190.049] (==) NVIDIA(0): RGB weight 888 [ 190.049] (==) NVIDIA(0): Default visual is TrueColor [ 190.049] (==) NVIDIA(0): Using gamma correction (1.0, 1.0, 1.0) [ 190.049] (**) NVIDIA(0): Option "UBB" "True" [ 190.049] (**) NVIDIA(0): Option "RenderAccel" "True" [ 190.049] (**) NVIDIA(0): Option "AllowGLXWithComposite" "True" [ 190.049] (**) NVIDIA(0): Option "UseEdidDpi" "DP-2" [ 190.049] (**) NVIDIA(0): Option "AddARGBGLXVisuals" "True" [ 190.049] (**) NVIDIA(0): Option "AllowEmptyInitialConfiguration" "True" [ 190.049] (**) NVIDIA(0): Enabling RENDER acceleration [ 190.049] (**) NVIDIA(0): Option "MetaModes" "nvidia-auto-select +0+0" [ 190.049] (**) NVIDIA(0): Option "RegistryDwords" "EnableBrightnessControl=1" [ 190.049] (**) NVIDIA(0): Enabling 2D acceleration [ 190.049] (II) Loading sub module "glxserver_nvidia" [ 190.049] (II) LoadModule: "glxserver_nvidia" [ 190.049] (II) Loading /usr/local/lib/xorg/modules/extensions/libglxserver_nvidia.so [ 190.066] (II) Module glxserver_nvidia: vendor="NVIDIA Corporation" [ 190.066] compiled for 1.6.99.901, module version = 1.0.0 [ 190.066] Module class: X.Org Server Extension [ 190.066] (II) NVIDIA GLX Module 510.47.03 Mon Jan 24 22:35:33 UTC 2022 [ 190.066] (II) NVIDIA: The X server supports PRIME Render Offload. [ 190.774] (--) NVIDIA(0): Valid display device(s) on GPU-0 at PCI:1:0:0 [ 190.774] (--) NVIDIA(0): DFP-0 [ 190.774] (--) NVIDIA(0): DFP-1 [ 190.774] (--) NVIDIA(0): DFP-2 [ 190.774] (--) NVIDIA(0): DFP-3 (boot) [ 190.774] (--) NVIDIA(0): DFP-4 [ 190.774] (--) NVIDIA(0): DFP-5 [ 190.774] (II) NVIDIA(0): NVIDIA GPU Quadro P1000 (GP107GL-A) at PCI:1:0:0 (GPU-0) [ 190.774] (--) NVIDIA(0): Memory: 4194304 kBytes [ 190.774] (--) NVIDIA(0): VideoBIOS: 86.07.63.00.76 [ 190.774] (II) NVIDIA(0): Detected PCI Express Link width: 16X [ 190.774] (--) NVIDIA(GPU-0): DFP-0: disconnected [ 190.774] (--) NVIDIA(GPU-0): DFP-0: Internal DisplayPort [ 190.774] (--) NVIDIA(GPU-0): DFP-0: 1440.0 MHz maximum pixel clock [ 190.774] (--) NVIDIA(GPU-0): [ 190.774] (--) NVIDIA(GPU-0): DFP-1: disconnected [ 190.774] (--) NVIDIA(GPU-0): DFP-1: Internal DisplayPort [ 190.774] (--) NVIDIA(GPU-0): DFP-1: 1440.0 MHz maximum pixel clock [ 190.774] (--) NVIDIA(GPU-0): [ 190.775] (--) NVIDIA(GPU-0): DFP-2: disconnected [ 190.775] (--) NVIDIA(GPU-0): DFP-2: Internal TMDS [ 190.775] (--) NVIDIA(GPU-0): DFP-2: 165.0 MHz maximum pixel clock [ 190.775] (--) NVIDIA(GPU-0): [ 190.775] (--) NVIDIA(GPU-0): Sharp (DFP-3): connected [ 190.775] (--) NVIDIA(GPU-0): Sharp (DFP-3): Internal DisplayPort [ 190.775] (--) NVIDIA(GPU-0): Sharp (DFP-3): 1440.0 MHz maximum pixel clock [ 190.775] (--) NVIDIA(GPU-0): [ 190.775] (--) NVIDIA(GPU-0): DFP-4: disconnected [ 190.775] (--) NVIDIA(GPU-0): DFP-4: Internal DisplayPort [ 190.775] (--) NVIDIA(GPU-0): DFP-4: 1440.0 MHz maximum pixel clock [ 190.775] (--) NVIDIA(GPU-0): [ 190.775] (--) NVIDIA(GPU-0): DFP-5: disconnected [ 190.775] (--) NVIDIA(GPU-0): DFP-5: Internal TMDS [ 190.775] (--) NVIDIA(GPU-0): DFP-5: 165.0 MHz maximum pixel clock [ 190.775] (--) NVIDIA(GPU-0): [ 190.778] (II) NVIDIA(0): Validated MetaModes: [ 190.778] (II) NVIDIA(0): "nvidia-auto-select+0+0" [ 190.778] (II) NVIDIA(0): Virtual screen size determined to be 3840 x 2160 [ 190.782] (--) NVIDIA(0): DPI set to (286, 288); computed from "UseEdidDpi" X config [ 190.782] (--) NVIDIA(0): option [ 190.782] (**) NVIDIA(0): Enabling 32-bit ARGB GLX visuals. [ 190.782] (II) NVIDIA: Reserving 24576.00 MB of virtual memory for indirect memory [ 190.782] (II) NVIDIA: access. [ 190.806] (II) NVIDIA(0): Setting mode "nvidia-auto-select+0+0" [ 191.934] (==) NVIDIA(0): Disabling shared memory pixmaps [ 191.934] (==) NVIDIA(0): Backing store enabled [ 191.934] (==) NVIDIA(0): Silken mouse enabled [ 191.934] (==) NVIDIA(0): DPMS enabled [ 191.935] (II) Loading sub module "dri2" [ 191.935] (II) LoadModule: "dri2" [ 191.935] (II) Module "dri2" already built-in [ 191.935] (II) NVIDIA(0): [DRI2] Setup complete [ 191.935] (II) NVIDIA(0): [DRI2] VDPAU driver: nvidia [ 191.935] (II) Initializing extension Generic Event Extension [ 191.935] (II) Initializing extension SHAPE [ 191.935] (II) Initializing extension MIT-SHM [ 191.936] (II) Initializing extension XInputExtension [ 191.936] (II) Initializing extension XTEST [ 191.936] (II) Initializing extension BIG-REQUESTS [ 191.937] (II) Initializing extension SYNC [ 191.937] (II) Initializing extension XKEYBOARD [ 191.937] (II) Initializing extension XC-MISC [ 191.937] (II) Initializing extension SECURITY [ 191.938] (II) Initializing extension XFIXES [ 191.938] (II) Initializing extension RENDER [ 191.938] (II) Initializing extension RANDR [ 191.939] (II) Initializing extension COMPOSITE [ 191.939] (II) Initializing extension DAMAGE [ 191.939] (II) Initializing extension MIT-SCREEN-SAVER [ 191.939] (II) Initializing extension DOUBLE-BUFFER [ 191.940] (II) Initializing extension RECORD [ 191.940] (II) Initializing extension DPMS [ 191.940] (II) Initializing extension Present [ 191.940] (II) Initializing extension DRI3 [ 191.941] (II) Initializing extension X-Resource [ 191.941] (II) Initializing extension XVideo [ 191.941] (II) Initializing extension XVideo-MotionCompensation [ 191.941] (II) Initializing extension GLX [ 191.941] (II) Initializing extension GLX [ 191.941] (II) Indirect GLX disabled. [ 191.942] (II) GLX: Another vendor is already registered for screen 0 [ 191.942] (II) Initializing extension XFree86-VidModeExtension [ 191.942] (II) Initializing extension XFree86-DGA [ 191.942] (II) Initializing extension XFree86-DRI [ 191.942] (II) Initializing extension DRI2 [ 191.942] (II) Initializing extension NV-GLX [ 191.943] (II) Initializing extension NV-CONTROL [ 191.943] (II) Initializing extension XINERAMA (snip non-video related part) [ 192.965] (--) NVIDIA(GPU-0): DFP-0: disconnected [ 192.965] (--) NVIDIA(GPU-0): DFP-0: Internal DisplayPort [ 192.965] (--) NVIDIA(GPU-0): DFP-0: 1440.0 MHz maximum pixel clock [ 192.965] (--) NVIDIA(GPU-0): [ 192.965] (--) NVIDIA(GPU-0): DFP-1: disconnected [ 192.965] (--) NVIDIA(GPU-0): DFP-1: Internal DisplayPort [ 192.965] (--) NVIDIA(GPU-0): DFP-1: 1440.0 MHz maximum pixel clock [ 192.965] (--) NVIDIA(GPU-0): [ 192.965] (--) NVIDIA(GPU-0): DFP-2: disconnected [ 192.965] (--) NVIDIA(GPU-0): DFP-2: Internal TMDS [ 192.965] (--) NVIDIA(GPU-0): DFP-2: 165.0 MHz maximum pixel clock [ 192.965] (--) NVIDIA(GPU-0): [ 192.965] (--) NVIDIA(GPU-0): Sharp (DFP-3): connected [ 192.965] (--) NVIDIA(GPU-0): Sharp (DFP-3): Internal DisplayPort [ 192.965] (--) NVIDIA(GPU-0): Sharp (DFP-3): 1440.0 MHz maximum pixel clock [ 192.965] (--) NVIDIA(GPU-0): [ 192.965] (--) NVIDIA(GPU-0): DFP-4: disconnected [ 192.965] (--) NVIDIA(GPU-0): DFP-4: Internal DisplayPort [ 192.965] (--) NVIDIA(GPU-0): DFP-4: 1440.0 MHz maximum pixel clock [ 192.965] (--) NVIDIA(GPU-0): [ 192.966] (--) NVIDIA(GPU-0): DFP-5: disconnected [ 192.966] (--) NVIDIA(GPU-0): DFP-5: Internal TMDS [ 192.966] (--) NVIDIA(GPU-0): DFP-5: 165.0 MHz maximum pixel clock [ 192.966] (--) NVIDIA(GPU-0): [ 192.966] (--) NVIDIA(GPU-0): DFP-0: disconnected [ 192.966] (--) NVIDIA(GPU-0): DFP-0: Internal DisplayPort [ 192.966] (--) NVIDIA(GPU-0): DFP-0: 1440.0 MHz maximum pixel clock [ 192.966] (--) NVIDIA(GPU-0): [ 192.966] (--) NVIDIA(GPU-0): DFP-1: disconnected [ 192.966] (--) NVIDIA(GPU-0): DFP-1: Internal DisplayPort [ 192.966] (--) NVIDIA(GPU-0): DFP-1: 1440.0 MHz maximum pixel clock [ 192.966] (--) NVIDIA(GPU-0): [ 192.966] (--) NVIDIA(GPU-0): DFP-2: disconnected [ 192.966] (--) NVIDIA(GPU-0): DFP-2: Internal TMDS [ 192.966] (--) NVIDIA(GPU-0): DFP-2: 165.0 MHz maximum pixel clock [ 192.966] (--) NVIDIA(GPU-0): [ 192.967] (--) NVIDIA(GPU-0): Sharp (DFP-3): connected [ 192.967] (--) NVIDIA(GPU-0): Sharp (DFP-3): Internal DisplayPort [ 192.967] (--) NVIDIA(GPU-0): Sharp (DFP-3): 1440.0 MHz maximum pixel clock [ 192.967] (--) NVIDIA(GPU-0): [ 192.967] (--) NVIDIA(GPU-0): DFP-4: disconnected [ 192.967] (--) NVIDIA(GPU-0): DFP-4: Internal DisplayPort [ 192.967] (--) NVIDIA(GPU-0): DFP-4: 1440.0 MHz maximum pixel clock [ 192.967] (--) NVIDIA(GPU-0): [ 192.967] (--) NVIDIA(GPU-0): DFP-5: disconnected [ 192.967] (--) NVIDIA(GPU-0): DFP-5: Internal TMDS [ 192.967] (--) NVIDIA(GPU-0): DFP-5: 165.0 MHz maximum pixel clock [ 192.967] (--) NVIDIA(GPU-0): [ 192.968] (--) NVIDIA(GPU-0): DFP-0: disconnected [ 192.968] (--) NVIDIA(GPU-0): DFP-0: Internal DisplayPort [ 192.968] (--) NVIDIA(GPU-0): DFP-0: 1440.0 MHz maximum pixel clock [ 192.968] (--) NVIDIA(GPU-0): [ 192.968] (--) NVIDIA(GPU-0): DFP-1: disconnected [ 192.968] (--) NVIDIA(GPU-0): DFP-1: Internal DisplayPort [ 192.968] (--) NVIDIA(GPU-0): DFP-1: 1440.0 MHz maximum pixel clock [ 192.968] (--) NVIDIA(GPU-0): [ 192.968] (--) NVIDIA(GPU-0): DFP-2: disconnected [ 192.968] (--) NVIDIA(GPU-0): DFP-2: Internal TMDS [ 192.968] (--) NVIDIA(GPU-0): DFP-2: 165.0 MHz maximum pixel clock [ 192.968] (--) NVIDIA(GPU-0): [ 192.968] (--) NVIDIA(GPU-0): Sharp (DFP-3): connected [ 192.968] (--) NVIDIA(GPU-0): Sharp (DFP-3): Internal DisplayPort [ 192.968] (--) NVIDIA(GPU-0): Sharp (DFP-3): 1440.0 MHz maximum pixel clock [ 192.968] (--) NVIDIA(GPU-0): [ 192.968] (--) NVIDIA(GPU-0): DFP-4: disconnected [ 192.968] (--) NVIDIA(GPU-0): DFP-4: Internal DisplayPort [ 192.968] (--) NVIDIA(GPU-0): DFP-4: 1440.0 MHz maximum pixel clock [ 192.968] (--) NVIDIA(GPU-0): [ 192.968] (--) NVIDIA(GPU-0): DFP-5: disconnected [ 192.968] (--) NVIDIA(GPU-0): DFP-5: Internal TMDS [ 192.968] (--) NVIDIA(GPU-0): DFP-5: 165.0 MHz maximum pixel clock [ 192.968] (--) NVIDIA(GPU-0): [ 193.412] (EE) [ 193.412] (EE) Backtrace: [ 193.414] (EE) 0: /usr/local/bin/Xorg (OsInit+0x38a) [0x42851a] [ 193.414] (EE) unw_get_proc_name failed: no unwind info found [-10] [ 193.414] (EE) 1: /lib/libthr.so.3 (?+0x0) [0x800955657] [ 193.415] (EE) unw_get_proc_name failed: no unwind info found [-10] [ 193.415] (EE) 2: /lib/libthr.so.3 (?+0x0) [0x800954b3e] [ 193.416] (EE) 3: ? (?+0x0) [0x7ffffffff2d3] [ 193.417] (EE) 4: ? (?+0x0) [0x0] [ 193.418] (EE) 5: /usr/local/bin/Xorg (present_event_notify+0x1961) [0x3916a1] [ 193.418] (EE) 6: /usr/local/bin/Xorg (present_register_complete_notify+0xffd) [0x392ead] [ 193.419] (EE) 7: /usr/local/bin/Xorg (UpdateCurrentTimeIf+0x41d) [0x29a8ed] [ 193.420] (EE) 8: /usr/local/bin/Xorg (InitFonts+0xb8c) [0x2a588c] [ 193.421] (EE) 9: /usr/local/bin/Xorg (_start+0x105) [0x28caa5] [ 193.421] (EE) 10: ? (?+0x0) [0x800462008] [ 193.421] (EE) [ 193.421] (EE) Segmentation fault at address 0x0 [ 193.421] (EE) Fatal server error: [ 193.421] (EE) Caught signal 11 (Segmentation fault). Server aborting [ 193.421] (EE) [ 193.421] (EE) Please consult the The X.Org Foundation support at http://wiki.x.org for help. [ 193.422] (EE) Please also check the log file at "/var/log/Xorg.0.log" for additional information. [ 193.422] (EE) [ 194.527] (EE) Server terminated with error (1). Closing log file.
Just an unsccessful info. Comparing dropped patches in files/ directory and nvidia code (510.47.03), I found nvidia code uses undefined macro LINUX_IOC_INOUT. *After build, grep'ed under and no #define found. /usr/ports/x11/nvidia-driver/work/ /usr/include/ /usr/local/include/ /compat/linux/usr/include/ /compat/linux/opt/ But unfortunately, avoiding use of the macro by the method used on dropped 460-* patches didn't help. *Created subsets of 460-* patches as 510-* patches and hooked them on Makefile. As it didn't help, I've not uploaded them here.
(In reply to Tomoaki AOKI from comment #11) That backtrace looks nonsensical to me. Neither InitFonts nor UpdateCurrentTimeIf call each other, for example. Maybe opening a core file directly with a debugger will get us better result? In general: 1. Linux emulation bits in the kernel driver can't cause this crash; 2. PRIME render offloading support on the kernel side doesn't have anything to do with this crash; 3. The module Xorg loads is not patched in any way by the port.
(In reply to Alex S from comment #13) > That backtrace looks nonsensical to me. Neither InitFonts nor > UpdateCurrentTimeIf call each other, for example. Maybe opening a core file > directly with a debugger will get us better result? Yes, it's confusing. And unfortunately, no core generated. I could not find core even if I've run `bfs / -name "*.core" -type f`as root. Only found kept cores before (unrelated here). Maybe the pre-compiled-by-nvidia binaries handle sigsegv and not at all invoking coredump. > In general: > 1. Linux emulation bits in the kernel driver can't cause this crash; > 2. PRIME render offloading support on the kernel side doesn't have anything to > do with this crash; > 3. The module Xorg loads is not patched in any way by the port. Agreed. But PRIME is not used here. Neither enabled by /etc/X11/xorg.conf nor enabled on BIOS configuration (as disabled internal GPU).
(In reply to Tomoaki AOKI from comment #14) > And unfortunately, no core generated. Start X in a debugger then. (The executable X. Not startx or similar scripts, that would be an unnecessary complication.)
(In reply to Alex S from comment #15) Hmm... X (symlinked to /usr/local/bin/Xorg.wrap) hesitates to run inside lldb, claiming that need to run by console user, lldb is run on console, though. So I started X from console and it didn't crash, sanely closed by ^C on console. This strongly hints at this could be a Mate DE problem. Temporarily switching to xfce, and it started fine. Thibault, what desktop environment are you using? If it's Mate, it would be almost clear that it's problem of Mate. I'll try rebuilding *mate* ports later, after rebuilding nvidia-driver 510.47.03 and report back what happened. (Now I'm on 470.)
(In reply to Tomoaki AOKI from comment #16) I am using LightDM for login manager, so it crash right when LightDM start. For the window manager I am using awesome-wm, and even with startx it crash. And with Wayland I am able to use optimus for certain program
(In reply to Thibault Payet from comment #17) So it could be depending on DE or WM, not only Mate DE. Rebuilding *mate* didn't help, and rebuilding almost all (maybe not loaded on startup and too large ones, webkit* and llvm* are omitted, and gcr and jasper failed) components also didn't help. Switching back to nvidia-driver 470.86 after above-mentioned rebuild made Mate DE start fine again \. As xfce4 is almost unusable for my daily use (too small rendering for 15.6 inch 4k monitor), I'm looking for another DM to better proceedeng tests. Maybe need fair amount of time, though.
FWIW, none of the mentioned programs crash for me. (In reply to Tomoaki AOKI from comment #18) You might want to try applying https://gitlab.freedesktop.org/xorg/xserver/-/commit/22d5818851967408bb7c903cb345b7ca8766094c.
(In reply to Alex S from comment #19) Wow! Thanks. That did it! Adding the diff you pointed to x11-servers/xorg-server/files allowed Mate DE to start. *Unfortunately, I couldn't find the way to download diff. So I hand-edited copy&pasted text to proper patch to add to ports files/ directory. I'll upload it here later for use until x11-servers/xorg-server is updated to contain this fix. Before reading your reply, I've tried x11/lumina and it started fine. But it was currently unusable for me. I need to find somehow scaling Gtk based apps to try further. Anyway, thanks for your advice! I can sanely run Mate DE on nvidia-driver 510.47.03 now.
Created attachment 231781 [details] File to add to x11-servers/xorg-server/files/ A patch file to place into x11-servers/xorg-server/files directory. Hot fix until x11-servers/xorg-server is updated to contain this fix. As file name can be different, anyone use this would need deleting it before git up to any point after and including next update of x11-servers/xorg-server.
(In reply to Tomoaki AOKI from comment #20) > *Unfortunately, I couldn't find the way to download diff. For that you need to append .patch (or .diff) to the url: https://gitlab.freedesktop.org/xorg/xserver/-/commit/22d5818851967408bb7c903cb345b7ca8766094c.patch. I actually wanted to link the corresponding issue instead, but all that UI noise makes the commit somewhat difficult to spot.
I did a quick test, manually patching the diff, and seems to be working great. Loaded nvidia-modeset, started Wayfire (wayland) and I can properly offload vkcube-xlib to Nvidia (nvidia GPU is selected and nvidia-smi reports a working process), vkcube-wayland displays nvidia as the chosen GPU but no cube is shown. This might just be a minor bug.
I tested with xorg 1.20.14 (which should include the patch mentioned in this thread), but I still have the same issue. And of course when I try to run X in a gdb it simply lock (I cannot vt switch back). My test are done by just launching X so that there is no question of which desktop environment does not work. Would it be feasible to launch X inside of valgrind ? Or maybe trying to build xorg with asan ? Not sure what to do for further debugging. With Wayland I have no issue with optimus, so I guess that there is some issue with GLX somewhere.
(In reply to Thibault Payet from comment #24) > xorg 1.20.14 (which should include the patch mentioned in this thread) It shouldn't and it doesn't. That said, I would be surprised if that patch works for you. > And of course when I try to run X in a gdb it simply lock (I cannot vt switch back). Oh well. Comment out https://github.com/freedesktop/xorg-xserver/blob/aeed57d722f2eb978c17fd7e859334d34af38d05/os/osinit.c#L173-L201 and run X normally. Should be enough to get an actual coredump.
(In reply to Alex S from comment #25) Sadly it it just lock up and I cannot vt switch. After a reboot my last two line of Xorg.0.log is [ 58.988] (II) NVIDIA GLX Module 510.47.03 Mon Jan 24 22:35:33 UTC 2022 [ 58.989] (II) NVIDIA: The X server supports PRIME Render Offload.
(In reply to Thibault Payet from comment #26) Is there a core file or not?
(In reply to Alex S from comment #27) Indeed there is one, not sure if it appear after playing with valgrind with a version without the removal Here a backtrace #0 thr_kill () at thr_kill.S:4 #1 0x00000008009ef064 in __raise (s=s@entry=6) at /usr/src/lib/libc/gen/raise.c:52 #2 0x0000000800aa3f29 in abort () at /usr/src/lib/libc/stdlib/abort.c:67 #3 0x0000000000420aea in OsAbort () #4 0x0000000000427bac in AbortServer () #5 0x00000000004262ba in FatalError () #6 0x000000000041e325 in OsSigHandler () #7 0x000000080092ce0e in handle_signal (actp=actp@entry=0x7fffffffdf00, sig=sig@entry=11, info=info@entry=0x7fffffffe2f0, ucp=ucp@entry=0x7fffffffdf80) at /usr/src/lib/libthr/thread/thr_sig.c:301 #8 0x000000080092c3cf in thr_sighandler (sig=11, info=0x7fffffffe2f0, _ucp=0x7fffffffdf80) at /usr/src/lib/libthr/thread/thr_sig.c:246 #9 <signal handler called> #10 0x0000000000000000 in ?? () #11 0x0000000801ca750f in ?? () from /usr/local/lib/xorg/modules/drivers/nvidia_drv.so #12 0x0000000801344100 in ?? () #13 0x0000000801349c28 in ?? () #14 0x0000000000000000 in ?? () Since I need a gui for inspecting the core file, my lib for nvidia changed, so I don't have .dynamic section that matches (here the warning) warning: .dynamic section for "/usr/local/lib/xorg/modules/drivers/nvidia_drv.so" is not at the expected address (wrong library or version mismatch?) warning: .dynamic section for "/usr/local/lib/libEGL_nvidia.so.0" is not at the expected address (wrong library or version mismatch?) warning: .dynamic section for "/usr/local/lib/libnvidia-glsi.so.1" is not at the expected address (wrong library or version mismatch?) warning: .dynamic section for "/usr/local/lib/libnvidia-eglcore.so.1" is not at the expected address (wrong library or version mismatch?) warning: .dynamic section for "/usr/local/lib/xorg/modules/extensions/libglxserver_nvidia.so" is not at the expected address (wrong library or version mismatch?) warning: .dynamic section for "/usr/local/lib/libnvidia-glcore.so.1" is not at the expected address (wrong library or version mismatch?) Core was generated by `/usr/local/bin/Xorg :0 -auth /root/.serverauth.42519'. Program terminated with signal SIGABRT, Aborted.
Hum, this seems just to be the abort when I press the power button..
If I update just the nvidia driver to have the correct mapping the backtrace is as follow #0 thr_kill () at thr_kill.S:4 #1 0x00000008009ef064 in __raise (s=s@entry=6) at /usr/src/lib/libc/gen/raise.c:52 #2 0x0000000800aa3f29 in abort () at /usr/src/lib/libc/stdlib/abort.c:67 #3 0x0000000000420aea in OsAbort () #4 0x0000000000427bac in AbortServer () #5 0x00000000004262ba in FatalError () #6 0x000000000041e325 in OsSigHandler () #7 0x000000080092ce0e in handle_signal (actp=actp@entry=0x7fffffffdf00, sig=sig@entry=11, info=info@entry=0x7fffffffe2f0, ucp=ucp@entry=0x7fffffffdf80) at /usr/src/lib/libthr/thread/thr_sig.c:301 #8 0x000000080092c3cf in thr_sighandler (sig=11, info=0x7fffffffe2f0, _ucp=0x7fffffffdf80) at /usr/src/lib/libthr/thread/thr_sig.c:246 #9 <signal handler called> #10 0x0000000000000000 in ?? () #11 0x0000000801ca750f in ?? () from /usr/local/lib/xorg/modules/drivers/nvidia_drv.so #12 0x0000000801c508c3 in ?? () from /usr/local/lib/xorg/modules/drivers/nvidia_drv.so #13 0x0000000801c8811d in ?? () from /usr/local/lib/xorg/modules/drivers/nvidia_drv.so #14 0x0000000801c8c912 in ?? () from /usr/local/lib/xorg/modules/drivers/nvidia_drv.so #15 0x00000008020461c3 in ?? () from /usr/local/lib/xorg/modules/drivers/nvidia_drv.so #16 0x0000000000000001 in ?? () #17 0x00007fffffffeaf0 in ?? () #18 0x0000000000476248 in ?? () #19 0x0000000800b3b8b0 in ?? () from /lib/libc.so.7 #20 0x0000000000476264 in xf86NumInputDrivers () #21 0x00000000002e60d4 in InitOutput () Backtrace stopped: previous frame inner to this frame (corrupt stack?) It seems stuck during InitOutput by looking at the backtrace.
(In reply to Thibault Payet from comment #30) Did you explicitly specify BusIDs in your config?
(In reply to Alex S from comment #31) Yes for both intel gpu (modesetting) and nvidia gpu
(In reply to Thibault Payet from comment #32) Can you change 0x02 to 0x01 at offset 0x0A58B1 in nvidia_drv.so? Does this have any effect?
(In reply to Alex S from comment #33) I'd be interested in `Option "UseDisplayDevice" "None"` config option as well.
FYI, stable driver was updated to 510.54. https://www.nvidia.com/Download/driverResults.aspx/187011/en-us
(In reply to Jung-uk Kim from comment #35) > FYI No further updates unless requested by the port maintainer. (Or there is an obvious error in the patch.) If I wanted to maintain an alternative port tree I would put this straight into a GitHub repo.
(In reply to Alex S from comment #34) Thank you it was this missing option in the Screen section (for dGPU) that was missing. I don't understand why it was working with 495.46 but not for 510.X I tested with 510.54, but I think that with 510.47.03 it should be the same.
After some days of usage, everything is fine, X Wayland hardware acceleration works. But for some reason, any native app that uses Nvidia as the renderer such as vkcube-wayland or alacritty does not display any window, even though the process is running (I can see nvidia-smi having a memory usage and tracking a process PID which is in this case, alacritty fore example). I don't know if it is a Vulkan issue or some missing package/option.
(In reply to Thibault Payet from comment #37) I don't think the option should be required, so there is some kind of a bug there. No sure what would be the trigger there. I'll leave this in case we need to redo the stack trace divination later: (lldb) mem read 0x0000000801ca750f 0x801ca750f: 48 85 c0 0f 84 5a 03 00 00 48 8b 70 10 48 8d 3d H....Z...H.p.H.= 0x801ca751f: 34 53 42 00 b9 06 00 00 00 f3 a6 48 8b 33 48 89 4SB........H.3H. (lldb) mem read 0x0000000801c508c3 0x801c508c3: 83 f8 01 74 1b 85 c0 0f 84 7a ff ff ff 48 89 ee ...t.....z...H.. 0x801c508d3: 48 89 df e8 b5 78 05 00 85 c0 0f 84 67 ff ff ff H....x......g... (lldb) mem read 0x0000000801c8811d 0x801c8811d: 84 c0 0f 84 2b f6 ff ff 4c 89 e7 e8 93 e7 fc ff ....+...L....... 0x801c8812d: b8 01 00 00 00 e9 09 fa ff ff 66 0f 1f 84 00 00 ..........f..... (lldb) mem read 0x0000000801c8c912 0x801c8c912: 84 c0 0f 84 44 ff ff ff 48 89 df e8 7e e9 ff ff ....D...H...~... 0x801c8c922: 84 c0 0f 84 34 ff ff ff 48 8b 5d 20 80 bb 1d 01 ....4...H.] .... (lldb) mem read 0x00000008020461c3 0x8020461c3: 84 c0 0f 84 56 ff ff ff 48 89 df e8 5d a6 00 00 ....V...H...]... 0x8020461d3: 48 89 ef e8 55 48 c2 ff 48 89 ef e8 2d 9e c2 ff H...UH..H...-... (There is no ASLR on 13, so, yes, addresses are exactly the same for me.)
(In reply to Daniel Pérez from comment #38) There an unofficial nvidia-drm port you might try for comparison: https://github.com/amshafer/nvidia-driver. Might a bit better for Wayland.
(In reply to Alex S from comment #40) * fare I'd stick to Nvidia on Xorg for a few more driver releases, though.
It's no problem, Nvidia is not my main card as I have Optimus. It's nice at least to have hardware accel for Xwayland tasks (and somehow also reduces power usage compared to having no driver attached). I'll wait until a new release comes, I'm fine as is.
*** Bug 262230 has been marked as a duplicate of this bug. ***
Alright, let's do this! Leeroy Jenkins! __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo | grep -i nvid server glx vendor string: NVIDIA Corporation client glx vendor string: NVIDIA Corporation, NVIDIA Corporation OpenGL vendor string: NVIDIA Corporation OpenGL renderer string: NVIDIA GeForce GTX 1050/PCIe/SSE2 OpenGL core profile version string: 4.6.0 NVIDIA 510.54 OpenGL core profile shading language version string: 4.60 NVIDIA OpenGL version string: 4.6.0 NVIDIA 510.54 OpenGL shading language version string: 4.60 NVIDIA OpenGL ES profile version string: OpenGL ES 3.2 NVIDIA 510.54 X Error of failed request: BadMatch (invalid parameter attributes) Major opcode of failed request: 78 (X_CreateColormap) Serial number of failed request: 129 Current serial number in output stream: 136 I still get this, but in practice I've played a lot of Steam games on Nvidia where Vulkan was at least working fine.
(In reply to Alex S from comment #40) fwiw that repo still doesn't have the latest bits required for true wayland support (still waiting on getting internal changes through but it's in progress :)). I don't recommend using it yet. Somewhat related, you won't be able to get optimus laptops working with prime since nvidia-drm isn't stable yet. You can try bumblebee, but I haven't done that personally so YMMV.
(In reply to tod.jackson from comment #44) on the other side, running __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears -info works on Wayland, via Xwayland, and Nvidia is correctly chosen as the renderer. I still have the problem that native apps that are offloaded don't render at all :(
Yeah, works fine is probably an overstatement. >.> I mean, this upgrade still has a lot of fixes for the traditional user.
(In reply to Tomoaki AOKI from comment #21) This patch does not prevent Xorg 1.20.13 from crashing while working with nvidia 510.54 driver. At least on my Lenovo IP 3 laptop with AMD Cezzane (Vega GPU) and NVIDIA GeForce RTX 3050 graphics. Nvidia driver 470.86 does not work either, the X gets stuck with black screen while mode switch.
There are 2 completely different crashes there, if that's not obvious enough.
(In reply to Alex S from comment #49) It was not, thanks for clarifying. BTW, 510.54 driver does not detect NVIDIA card name, here's what I get in dmesg: nvidia0: <Unknown> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: child nvidia0 requested pci_enable_io nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 510.54 Tue Feb 8 04:25:23 UTC 2022 While, 470.86 gives: nvidia0: <NVIDIA GeForce RTX 3050 Laptop GPU> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: child nvidia0 requested pci_enable_io nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 470.86 Tue Oct 26 21:43:42 UTC 2021 I wonder if this is the same for other models, it can be the reason why Xorg crashes.
Just upgraded to X.Org X Server 1.20.14 from ports, same crash persists both with nvidia and amdgpu drivers. X.Org X Server 1.20.14 X Protocol Version 11, Revision 0 Build Operating System: FreeBSD 13.1-BETA1 amd64 Current Operating System: FreeBSD butterfly 13.1-BETA1 FreeBSD 13.1-BETA1 #3: Sun Mar 13 06:28:35 +05 2022 rz@butterfly:/usr/obj/usr/src/amd64.amd64/sys/GENERIC amd64 Build Date: 15 March 2022 03:38:05AM Current version of pixman: 0.40.0 Before reporting problems, check http://wiki.x.org to make sure that you have the latest version. Markers: (--) probed, (**) from config file, (==) default setting, (++) from command line, (!!) notice, (II) informational, (WW) warning, (EE) error, (NI) not implemented, (??) unknown. (==) Log file: "/var/log/Xorg.0.log", Time: Tue Mar 15 03:43:11 2022 (==) Using config file: "/etc/X11/xorg.conf" (==) Using system config directory "/usr/local/share/X11/xorg.conf.d" (II) AMDGPU(0): [KMS] Kernel modesetting enabled. amdgpu: os_same_file_description couldn't determine if two DRM fds reference the same file description. If they do, bad things may happen! Assertion failed: (key->initialized), function dixGetPrivateAddr, file /usr/local/include/xorg/privates.h, line 121. (EE) (EE) Backtrace: (EE) 0: /usr/local/bin/Xorg (OsInit+0x38a) [0x41c96a] (EE) unw_get_proc_name failed: no unwind info found [-10] (EE) 1: /lib/libthr.so.3 (?+0x0) [0x80093158e] (EE) unw_get_proc_name failed: no unwind info found [-10] (EE) 2: /lib/libthr.so.3 (?+0x0) [0x800930b3f] (EE) 3: ? (?+0x0) [0x7ffffffff8a3] (EE) unw_get_proc_name failed: no unwind info found [-10] (EE) 4: /lib/libc.so.7 (?+0x0) [0x800a7c2ca] (EE) unw_get_proc_name failed: no unwind info found [-10] (EE) 5: /lib/libthr.so.3 (?+0x0) [0x800930a00] (EE) 6: ? (?+0x0) [0x0] (EE) unw_get_proc_name failed: no unwind info found [-10] (EE) 7: /lib/libc.so.7 (?+0x0) [0x8009f4c74] (EE) unw_step failed: unspecified (general) error [-1] (EE) (EE) Fatal server error: (EE) Caught signal 6 (Abort trap). Server aborting
(In reply to Ruslan Zalata from comment #51) > same crash persists both with nvidia and amdgpu drivers. Nope.
Just upgraded to 510.60 Nvidia driver, it detects card correctly now: nvidia0: <NVIDIA GeForce RTX 3050 Laptop GPU> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: child nvidia0 requested pci_enable_io nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 510.60.02 Wed Mar 16 11:03:12 UTC 2022 pid 8474 (Xorg), jid 0, uid 0: exited on signal 6 (core dumped) Yet Xorg 1.20.14 still crashes at same place, can use it in modesetting mode only. rz@butterfly:~ % uname -a FreeBSD butterfly 13.1-BETA2 FreeBSD 13.1-BETA2 #0 releng/13.1-n249982-88358b6dcbe: Thu Mar 17 02:23:35 UTC 2022 root@releng3.nyi.freebsd.org:/usr/obj/usr/src/amd64.amd64/sys/GENERIC amd64
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=73c40502a978f4f8daafa02896d7cf21e352e130 commit 73c40502a978f4f8daafa02896d7cf21e352e130 Author: Kevin Bowling <kbowling@FreeBSD.org> AuthorDate: 2022-04-04 20:46:41 +0000 Commit: Kevin Bowling <kbowling@FreeBSD.org> CommitDate: 2022-04-04 20:48:53 +0000 x11/nvidia-driver-390, x11/linux-nvidia-libs-390: update to 390.147 PR: 261666 Approved by: maintainer timeout Sponsored by: BBOX.io x11/linux-nvidia-libs-390/Makefile | 2 +- x11/linux-nvidia-libs/distinfo | 6 +++--- x11/nvidia-driver-390/Makefile | 2 +- x11/nvidia-driver/distinfo | 6 +++--- 4 files changed, 8 insertions(+), 8 deletions(-)
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=cbf529cb2c6d08b370f13706a8400bbc0b6d193c commit cbf529cb2c6d08b370f13706a8400bbc0b6d193c Author: Kevin Bowling <kbowling@FreeBSD.org> AuthorDate: 2022-04-04 20:38:49 +0000 Commit: Kevin Bowling <kbowling@FreeBSD.org> CommitDate: 2022-04-04 20:48:53 +0000 x11/nvidia-driver, x11/linux-nvidia-libs: update to 510.60.02 PR: 261666 Approved by: maintainer timeout Sponsored by: BBOX.io x11/linux-nvidia-libs/Makefile | 2 +- x11/linux-nvidia-libs/distinfo | 6 +++--- x11/nvidia-driver/Makefile | 2 +- x11/nvidia-driver/distinfo | 6 +++--- 4 files changed, 8 insertions(+), 8 deletions(-)
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=10ed81c24e86e84660bd61c8ab31d66f21e14e2f commit 10ed81c24e86e84660bd61c8ab31d66f21e14e2f Author: Alex S <iwtcex@gmail.com> AuthorDate: 2022-04-04 20:18:23 +0000 Commit: Kevin Bowling <kbowling@FreeBSD.org> CommitDate: 2022-04-04 20:48:52 +0000 x11/nvidia-driver, x11/linux-nvidia-libs: update to 510.47.03 PR: 261666 Approved by: maintainer timeout Sponsored by: BBOX.io x11/linux-nvidia-libs/Makefile | 45 ++++++++++++++++++++++++++---------------- x11/linux-nvidia-libs/distinfo | 6 +++--- x11/nvidia-driver/Makefile | 11 +++++++---- x11/nvidia-driver/distinfo | 6 +++--- 4 files changed, 41 insertions(+), 27 deletions(-)
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/ports/commit/?id=9e7f5392edf0e9a9bb78f7ecfe433b0e7dbd624b commit 9e7f5392edf0e9a9bb78f7ecfe433b0e7dbd624b Author: Alex S <iwtcex@gmail.com> AuthorDate: 2022-04-04 20:22:10 +0000 Commit: Kevin Bowling <kbowling@FreeBSD.org> CommitDate: 2022-04-04 20:48:53 +0000 x11/nvidia-driver, x11/linux-nvidia-libs: Add 470.103.01 This legacy branch driver is to maintain support for Keplar GPUs. PR: 261666 Approved by: maintainer timeout Sponsored by: BBOX.io x11/linux-nvidia-libs-470/Makefile (new) | 8 ++++++++ x11/linux-nvidia-libs/distinfo | 4 +++- x11/nvidia-driver-470/Makefile (new) | 8 ++++++++ x11/nvidia-driver/Makefile.common | 12 ++++++++---- x11/nvidia-driver/distinfo | 4 +++- 5 files changed, 30 insertions(+), 6 deletions(-)