| Summary: | nvidia-driver: NVIDIA MEM resource alloc failed | ||
|---|---|---|---|
| Product: | Ports & Packages | Reporter: | christian <contato> |
| Component: | Individual Port(s) | Assignee: | freebsd-ports-bugs (Nobody) <ports-bugs> |
| Status: | New --- | ||
| Severity: | Affects Only Me | CC: | ashafer, junchoon |
| Priority: | --- | ||
| Version: | Latest | ||
| Hardware: | amd64 | ||
| OS: | Any | ||
How do you load nvidia driver? Via /boot/loader.conf[.local]? Or just missed but actually /etc/rc.conf[.local]? If the former, your driver could be half-read and not working normally. (Memory area that loader allocates for kernel and kernel modules before starting kernel is quite limited.) If the latter, but actually missing nvidia-modeset in kld_list, try adding it first. If the latter and actually specifying nvidia-modeset to be added in kld_list somewhere after kld_list="i915kms" line, trying latest Production branch of driver, 550.127.05, or latest BETA branch of the driver, 565.57.01 by overriding driver version and disable checksum could help. There is my PR, Bug 282312 – x11/nvidia-driver: Update to 550.127.05 with x11/linux-nvidia-libs and related DRM ports [1]. This is needed to try 565.57.01 if you also want x11/linux-nvidia-libs, graphics/nvidia-drm-[510|515|61]-kmod ports. [1] https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=282312 I'm sorry, in rc.conf kld_list="nvidia-modeset", but trying nvidia-drm, only nvidia and no works. Black screen freeze. (In reply to Tomoaki AOKI from comment #1) I used this wiki: https://wiki.freebsd.org/Graphics https://badland.io/prime-configuration.md I tried all options. but I don't think it's the driver. Exists something tunable in kernel to alloc mem card device? thank you for helping me. (In reply to Tomoaki AOKI from comment #1) I tried in terminal, but this error output command, so i don'n know if option works perfectly: # sysctl hw.pci.allow_unsupported_io_range=1 sysctl: unknown oid 'hw.pci.allow_unsupported_io_range' I compiled nvidia-driver 565.57.01, bus not worked. nvidia0: <NVIDIA GeForce RTX 3050 6GB Laptop GPU> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: 0x1000000 bytes of rid 0x10 res 3 failed (0, 0xffffffffffffffff). nvidia0: NVRM: NVIDIA MEM resource alloc failed, BAR0 @ 0x10. nvidia0: NVRM: NVIDIA hardware alloc failed. device_attach: nvidia0 attach returned 6 nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 565.57.01 Thu Oct 10 suggestions? thanks! my dmesg output 1st os.rwlock_sx @ nvidia_os.c:782 2nd os.rwlock_sx @ nvidia_os.c:782 (In reply to christian from comment #4) The tunable is already gone long before [2]. [2] https://cgit.freebsd.org/src/commit/?id=e4b59fc50065e183c020898b461f49b7ba1483bd author Warner Losh <imp@FreeBSD.org> 2004-01-11 06:52:31 +0000 committer Warner Losh <imp@FreeBSD.org> 2004-01-11 06:52:31 +0000 commit e4b59fc50065e183c020898b461f49b7ba1483bd (patch) Add support for subtractive decoding bridges. These bridges pass all signals to addresses to the child busses. Typically, ProgIf of 1 means a subtractive bridge. However, Intel has a whole lot of ones with a ProgIf of 80 that are also subtractive. We cope with these bridges too. This eliminates hw.pci.allow_unsupported_io_range because that had almost the same effect as these patches (almost means 'buggy'). Remove the bogus checks for ISA bus locations: these cycles aren't special and are only passed by transparent bridges. We allow any range to succeed. If the range is a superset of the range that's decoded, trim the resource to that range. Otherwise, pass the range unchanged. This will change the location that PC Card and CardBus cards are attached. This might bogusly cause some overlapping allocation that wasn't present before, but the overlapping fixes need to be in the pci level. There's also a few formatting changes here. Notes Notes: svn path=/head/; revision=124365 (In reply to christian from comment #5) My quite wild prediction was that GSP firmware, which were introduced starting from 560 series of drivers, could manage resources used for GPUs, and your GPU is new enough to have GSP in it. GSP firmwares are contained in x11/nvidia-driver and should be automatically loaded for supported GPUs. (My Pascal generation of GPU is too old to even test it, as it does not contain GSP in it.) Other possibilities I can think of for now: *If possible, disable iGPU (in CPU) via UEFI firmware (or legacy BIOS if applicable). *Disable devices you don't actually need via UEFI firmware / legacy BIOS. *Change VRAM size configured in UEFI firmware / legacy BIOS. And if you're using graphics/nvidia-drm-[510|515|61]-kmod in conjunction with x11/nvidia-driver, legacy way for overriding driver version is insufficient. Both Austin Shafer (the developer of the nvidia DRM ports) and me couldn't determine why, but graphics/nvidia-drm-[510|515|61]-kmod ports requires x11/nvidia-driver/Makefile.version to be edited for the version you want (here, 565.57.01). Without it, /dev/dri/* would not appear on boot and DRM driver shouldn't work. (In reply to Tomoaki AOKI from comment #8) thanks, i did. add in distinfo: SHA256 (NVIDIA-FreeBSD-x86_64-565.57.01.tar.xz) = 808227e4250d019501834f536157e2a3271cc16d58862a73d20d315862c2425e SIZE (NVIDIA-FreeBSD-x86_64-565.57.01.tar.xz) = 21610082 Makefile.version # NVIDIA Distversion # # This will be included from x11/nvidia-driver and the nvidia-drm po NVIDIA_DISTVERSION = 565.57.01 make install works. But not success driver. nvidia0: <NVIDIA GeForce RTX 3050 6GB Laptop GPU> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: 0x1000000 bytes of rid 0x10 res 3 failed (0, 0xffffffffffffffff). nvidia0: NVRM: NVIDIA MEM resource alloc failed, BAR0 @ 0x10. nvidia0: NVRM: NVIDIA hardware alloc failed. device_attach: nvidia0 attach returned 6 nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 565.57.01 Thu Oct 10 (In reply to Tomoaki AOKI from comment #8) I go to try. thanks My laptop is Prime. No disable iGPU. Legion 5. rtx 3050 6GB VRAM (In reply to Tomoaki AOKI from comment #8) driver 470 ( 470.161.03_1 ) not works. :( nvidia0: <Unknown> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io nvidia0: NVRM: NVIDIA MEM resource alloc failed, BAR0 @ 0x10. nvidia0: NVRM: NVIDIA hardware alloc failed. device_attach: nvidia0 attach returned 6 nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 470.161.03 Wed Oct 19 00:01:15 UTC 2022 with nvidia-drm-kmod acquiring duplicate lock of same type: "os.rwlock_sx" 1st os.rwlock_sx @ nvidia_os.c:782 2nd os.rwlock_sx @ nvidia_os.c:782 stack backtrace: #0 0xffffffff80bc761c at witness_debugger+0x6c #1 0xffffffff80b5c7cd at _sx_xlock+0x5d #2 0xffffffff84bef012 at os_acquire_rwlock_write+0x32 #3 0xffffffff8490a5f0 at _nv044172rm+0x10 nvidia0: <NVIDIA GeForce RTX 3050 6GB Laptop GPU> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: 0x1000000 bytes of rid 0x10 res 3 failed (0, 0xffffffffffffffff). nvidia0: NVRM: NVIDIA MEM resource alloc failed, BAR0 @ 0x10. nvidia0: NVRM: NVIDIA hardware alloc failed. device_attach: nvidia0 attach returned 6 nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 550.120 Fri Sep 13 09:32:47 UTC 2024 sry I'm trying, but nothing... I have not seen this before. The odd thing is I do actually have a legion 5 with freebsd on it but have never seen an issue like this. Only difference from yours is I have an AMD CPU, so maybe that's part of the reason.
If it happens on older drivers like 470 that should rule out GSP.
> vgapci0: 0x1000000 bytes of rid 0x10 res 3 failed (0, 0xffffffffffffffff).
This line seems to indicate something in the vgapci driver for the kernel is going wrong. I would guess something is up with the way it is accessing the PCI bus. I'm not sure what that would be though. You might want to test with a linux liveusb to quickly verify that it's not a problem with the hardware.
Hello, and apologies for the delay in my response. I tested the GPU on both Linux and Windows, and it works as expected on both platforms. This confirms that the hardware is functioning correctly, so it seems the issue is specific to FreeBSD. Thank you for your patience and support. (In reply to Austin Shafer from comment #13) In Brazil, not find amd cpu only intel. :( o FreeBSD 15 have support the card graphics integrated. I'm using it. But tested 13 and 14 and both with error. |
I'm sorry, but i'm trying all not sucess. Edited /boot/loader.conf in FreeBSD 14 and now 15-current (current worked my integrated graphics) it's not worked my card nvidia 3050 6gb: dmesg output nvidia0: <NVIDIA GeForce RTX 3050 6GB Laptop GPU> on vgapci0 vgapci0: child nvidia0 requested pci_enable_io vgapci0: 0x1000000 bytes of rid 0x10 res 3 failed (0, 0xffffffffffffffff). nvidia0: NVRM: NVIDIA MEM resource alloc failed, BAR0 @ 0x10. nvidia0: NVRM: NVIDIA hardware alloc failed. device_attach: nvidia0 attach returned 6 my /etc/rc.conf # cat /etc/rc.conf hostname="legion5" keymap="br.kbd" wlans_iwlwifi0="wlan0" ifconfig_wlan0="WPA DHCP" ifconfig_wlan0_ipv6="inet6 accept_rtadv" sshd_enable="YES" moused_enable="YES" powerd_enable="YES" # Set dumpdev to "AUTO" to enable crash dumps, "NO" to disable dumpdev="AUTO" kld_list="i915kms" linux_enable="YES" dbus_enable="YES" utility_enable="YES" /boot/loader.conf hw.pci.allow_unsupported_io_range=1 my output pciconf -lv | grep -B3 VGA vendor = 'Intel Corporation' device = 'Raptor Lake-P [UHD Graphics]' class = display subclass = VGA -- vendor = 'NVIDIA Corporation' device = 'GN20-P0-R-K2 [GeForce RTX 3050 6GB Laptop GPU]' class = display subclass = VGA using nvidia-driver-550.120 trying file package or compiled ports no sucess.Sorry my bad english.