Arch Linux User Repository

@changhuazhiye you can manually change the source url in the PKGBUILD.

can't download https://download.nvidia.com/XFree86/NVIDIA-kernel-module-source/NVIDIA-kernel-module-source-590.48.01.run in china,but can download https://cn.download.nvidia.cn/XFree86/Linux-x86_64/590.48.01/NVIDIA-Linux-x86_64-590.48.01.run . Are there any other ways to fix this？

I think we'll need something like this until the source is fixed for the 6.17+ kernels because of the drm_helper_mode_fill_fb_struct signature change here:

--- a/kernel-open/nvidia-drm/nvidia-drm-fb.h    2025-08-11 15:31:13.457595856 -0700
+++ b/kernel-open/nvidia-drm/nvidia-drm-fb.h    2025-08-11 15:38:25.764583014 -0700
@@ -23,6 +23,7 @@
 #ifndef __NVIDIA_DRM_FB_H__
 #define __NVIDIA_DRM_FB_H__

+#include <linux/version.h>
 #include "nvidia-drm-conftest.h"

 #if defined(NV_DRM_AVAILABLE)
@@ -50,10 +51,18 @@
     return container_of(fb, struct nv_drm_framebuffer, base);
 }

+#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 17, 0)
 struct drm_framebuffer *nv_drm_framebuffer_create(
     struct drm_device *dev,
     struct drm_file *file,
     const struct drm_mode_fb_cmd2 *cmd);
+#else
+struct drm_framebuffer *nv_drm_framebuffer_create(
+    struct drm_device *dev,
+    struct drm_file *file,
+    const struct drm_format_info *info,
+    const struct drm_mode_fb_cmd2 *cmd);
+#endif

 #endif /* NV_DRM_AVAILABLE */

--- a/kernel-open/nvidia-drm/nvidia-drm-fb.c    2025-08-11 15:31:13.457595856 -0700
+++ b/kernel-open/nvidia-drm/nvidia-drm-fb.c    2025-08-11 15:39:02.195358911 -0700
@@ -24,6 +24,8 @@

 #if defined(NV_DRM_AVAILABLE)

+#include <linux/version.h>
+
 #include "nvidia-drm-priv.h"
 #include "nvidia-drm-ioctl.h"
 #include "nvidia-drm-fb.h"
@@ -33,6 +35,9 @@
 #include "nvidia-drm-format.h"

 #include <drm/drm_crtc_helper.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 17, 0)
+#include <drm/drm_fourcc.h>
+#endif

 static void __nv_drm_framebuffer_free(struct nv_drm_framebuffer *nv_fb)
 {
@@ -239,10 +244,18 @@
     return -EINVAL;
 }

+#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 17, 0)
+struct drm_framebuffer *nv_drm_framebuffer_create(
+    struct drm_device *dev,
+    struct drm_file *file,
+    const struct drm_mode_fb_cmd2 *cmd)
+#else
 struct drm_framebuffer *nv_drm_framebuffer_create(
     struct drm_device *dev,
     struct drm_file *file,
+    const struct drm_format_info *info,
     const struct drm_mode_fb_cmd2 *cmd)
+#endif
 {
     struct nv_drm_device *nv_dev = to_nv_device(dev);
     struct nv_drm_framebuffer *nv_fb;
@@ -286,10 +299,23 @@

     /* Fill out framebuffer metadata from the userspace fb creation request */

+#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 17, 0)
+    drm_helper_mode_fill_fb_struct(
+        dev,
+        &nv_fb->base,
+        cmd);
+#else
+    if (!info) {
+        u64 modifier = (cmd->flags & DRM_MODE_FB_MODIFIERS) ? 
+                       cmd->modifier[0] : DRM_FORMAT_MOD_INVALID;
+        info = drm_get_format_info(dev, cmd->pixel_format, modifier);
+    }
     drm_helper_mode_fill_fb_struct(
         dev,
         &nv_fb->base,
+        info,
         cmd);
+#endif

     /*
      * Finish up FB initialization by creating the backing NVKMS surface and

@dbermond my kernel build does create a directory in /usr/lib/modules. That is not the issue. The issue is that when it's symlinked back to the source, there is no version file where this build script is looking with the following snippet:

local _kernver
    if [ -d "/usr/lib/modules/$(uname -r)" ]
    then
        _kernver="$(<"/usr/lib/modules/$(uname -r)/build/version")"
    else
        _kernver="$(find /usr/lib/modules -mindepth 1 -maxdepth 1 -type d | head -n1)"
        _kernver="$(<"/usr/lib/modules/${_kernver##*/}/build/version")"
fi

if i'm not mistaken, the version file at that location has been deprecated and is now at include/generated/uapi/linux/version.h. what I provided below would change the PKGBUILD to only look at module folders found in /usr/lib/modules without the need for a specific version file. With the latest update and a locally built 6.15.2, this was still necessary. Perhaps there's a better way to write it, but it's functional at this point with what's below.

@sonofevropa firstly, your build issue should be fixed by your custom kernel, and not in this package, as it should correctly provide a versioned directory for kernel modules. Secondly, your suggestion does not make sense, as the '_kernver' variable is not used in the build() function.

Every time I build this, it requires intervention to build against my locally built kernel, which with traditional compilation does not produce a version file. I'd like to propose a change to the package build:

local _kernver
    if [ -d "/usr/lib/modules/$(uname -r)" ]
    then
        _kernver="$(uname -r)"
    else
        _kernver="$(find /usr/lib/modules -mindepth 1 -maxdepth 1 -type d | head -n1)"
        _kernver=_kernver="${_kernver##*/}"
    fi

@nikdog I have just checked after deleting my local cached upstream sources, and the package is building perfectly fine. I cannot reproduce your issue.

Not sure why I'm getting this error. Would seem to be bad source files, but then everyone would have this problem.

 [ nvidia            ]  CC           _out/Linux_x86_64/g_nvid_string.c
_out/Linux_x86_64/g_nvid_string.c:1:25: warning: missing terminating " character
    1 | const char  NVRM_ID[] = "nvidia id: NVIDIA UNIX Open Kernel Module for x86_64  575.51.02  Release Build  (archlinux_builder@archlinux)  May 18 07:20:21.275 [notice] Tor 0.4.8.16 running on Linux with Libevent 2.1.12-stable, OpenSSL 3.5.0, Zlib 1.3.1, Liblzma 5.8.1, Libzstd 1.5.7 and Glibc 2.41 as libc.
      |                         ^
_out/Linux_x86_64/g_nvid_string.c:1:25: error: missing terminating " character
    1 | const char  NVRM_ID[] = "nvidia id: NVIDIA UNIX Open Kernel Module for x86_64  575.51.02  Release Build  (archlinux_builder@archlinux)  May 18 07:20:21.275 [notice] Tor 0.4.8.16 running on Linux with Libevent 2.1.12-stable, OpenSSL 3.5.0, Zlib 1.3.1, Liblzma 5.8.1, Libzstd 1.5.7 and Glibc 2.41 as libc.
      |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_out/Linux_x86_64/g_nvid_string.c:2:1: error: ‘May’ undeclared here (not in a function)
    2 | May 18 07:20:21.275 [notice] Tor can't help you if you use it wrong! Learn how to be safe at https://support.torproject.org/faq/staying-anonymous/
      | ^~~
_out/Linux_x86_64/g_nvid_string.c:2:5: error: expected ‘,’ or ‘;’ before numeric constant
    2 | May 18 07:20:21.275 [notice] Tor can't help you if you use it wrong! Learn how to be safe at https://support.torproject.org/faq/staying-anonymous/
      |     ^~
_out/Linux_x86_64/g_nvid_string.c:2:37: warning: missing terminating ' character
    2 | May 18 07:20:21.275 [notice] Tor can't help you if you use it wrong! Learn how to be safe at https://support.torproject.org/faq/staying-anonymous/
      |                                     ^
_out/Linux_x86_64/g_nvid_string.c:2:37: error: missing terminating ' character
    2 | May 18 07:20:21.275 [notice] Tor can't help you if you use it wrong! Learn how to be safe at https://support.torproject.org/faq/staying-anonymous/
      |                                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_out/Linux_x86_64/g_nvid_string.c:3:48: warning: multi-character character constant [-Wmultichar]
    3 | May 18 07:20:21.275 [warn] Command-line option 'PDT' with no value. Failing.
      |                                                ^~~~~
_out/Linux_x86_64/g_nvid_string.c:4:69: warning: missing terminating " character
    4 | May 18 07:20:21.275 [err] Reading config failed--see warnings above.";
      |                                                                     ^
_out/Linux_x86_64/g_nvid_string.c:4:69: error: missing terminating " character
    4 | May 18 07:20:21.275 [err] Reading config failed--see warnings above.";
      |                                                                     ^~
make[1]: *** [Makefile:203: _out/Linux_x86_64/g_nvid_string.o] Error 1
make[1]: Leaving directory '/home/nikdog/.cache/yay/nvidia-open-beta/src/open-gpu-kernel-modules-575.51.02/src/nvidia'
make: *** [Makefile:34: src/nvidia/_out/Linux_x86_64/nv-kernel.o] Error 2
make: Leaving directory '/home/nikdog/.cache/yay/nvidia-open-beta/src/open-gpu-kernel-modules-575.51.02'
==> ERROR: A failure occurred in build().
    Aborting...
 -> error making: nvidia-open-beta-exit status 4

@@brianc Try to build the package in a clean chroot. Alternatively, use nvidia-open-beta-dkms.

@brianc Are you using the nvidia-open-beta or nvidia-open-beta-dkms package? the dkms package should automatically rebuild and install itself properly after a kernel update, so if you are not using that one, give it a try. Also, do you have the headers package for your specific kernel type installed?