nvidia 580.76.05 kernel-7 patch + docs

README.md

@@ -0,0 +1,67 @@
+# nvidia-kernel-patches
+
+Source patches that make NVIDIA's driver build against kernels newer than the driver
+knows about. Currently: **580.76.05 (kernel-open) on 7.0.14-11-pve** (Proxmox VE 9.2.10,
+host `spicy`). New driver/kernel combos get their own patch + section as they happen.
+
+Referenced from the house infra repo (trikilli → `hosts/spicy-rtx3090-bringup.md`),
+which holds the full bring-up narrative (LXC passthrough, DKMS setup, device majors).
+This repo is just the patch and how to use it.
+
+## nvidia-580.76.05-kernel7.patch
+
+Applies to the `kernel-open/` tree extracted from `NVIDIA-Linux-x86_64-580.76.05.run`.
+Deployed on spicy 2026-08-10; DKMS source lives at `/usr/src/nvidia-580.76.05` on-host.
+
+### Why the stock driver fails on kernel 7.x
+
+Three independent causes:
+
+1. **GCC 14 breaks NVIDIA's conftest.** The "is function X present" probes are
+   *inverted*: they call X with no arguments and treat a successful compile (old
+   compilers: implicit-declaration warning) as "absent". GCC 14 makes implicit
+   declarations hard errors → every probe fails to compile → everything detected
+   "present" → driver calls APIs that don't exist (and builds Tegra-only code on x86).
+2. **Kernel 7.x headers need `-std=gnu11 -fms-extensions`.** kbuild passes these;
+   conftest didn't. Without `-fms-extensions`, anonymous tagged-struct members
+   (e.g. `struct filename { struct __filename_head; ... }`) silently drop and a
+   `static_assert` on the struct size kills every probe including fs/device/pci headers.
+3. **Real API changes** — table below.
+
+### What the patch does
+
+| File | Change |
+|---|---|
+| `conftest.sh` | Add `-std=gnu11 -fms-extensions -Wno-error=implicit-function-declaration -Wno-error=int-conversion -Wno-error=incompatible-pointer-types` to CFLAGS |
+| `common/inc/nv-time.h`, `nvidia/os-interface.c` | `in_irq()` → `in_hardirq()` (removed) |
+| `common/inc/nv-mm.h` | `vm_flags` writes → `vm_flags_set/clear()` / `vma_flags_set_word/clear_word()` (vm_flags is a const union member now) |
+| `nvidia/nv-dma.c` | `dma_map_ops.map_resource` → `map_phys` (phys-addr DMA rework) |
+| `nvidia/nv-pci.c` | `pci_resize_resource()` gained 4th arg `exclude_bars` → pass `0` |
+| `nvidia-uvm/uvm_pmm_gpu.c` | `dev_pagemap_ops.page_free(page)` → `.folio_free(folio)` — thin wrappers added |
+| `nvidia-uvm/uvm_hmm.c` | `zone_device_page_init(page)` → `(page, pgmap, order)` signature |
+
+Use **kernel-open**, not the closed blob: GA102+ is fully supported, and the open
+modules are `Dual MIT/GPL` so they can call the kernel's GPL-only VMA-lock API —
+the closed blob reimplements those internals and chases a moving target.
+Exclude `nvidia-drm` (DRM API churn, headless boxes don't need it) and `nvidia-peermem`.
+
+### Apply + install
+
+```sh
+sh NVIDIA-Linux-x86_64-580.76.05.run --extract-only --target nv580
+cd nv580 && patch -p1 < nvidia-580.76.05-kernel7.patch   # paths are kernel-open/...
+cp -r kernel-open /usr/src/nvidia-580.76.05              # + dkms.conf: set version,
+                                                          #   exclude nvidia-drm/peermem
+dkms add nvidia/580.76.05 && dkms build nvidia/580.76.05 && dkms install nvidia/580.76.05
+sh NVIDIA-Linux-x86_64-580.76.05.run --no-kernel-modules --silent --no-x-check  # userspace
+```
+
+Container userspace must match the host module version **exactly**; install inside
+each LXC with `--no-kernel-modules`.
+
+### Caveats
+
+- Every kernel update re-runs the DKMS build against new headers — further API churn
+  means further patching. Verify `nvidia-smi` (host, then containers) after any reboot.
+- `-Wno-error=incompatible-pointer-types` in conftest is scoped to probe compiles only;
+  the real module build keeps default warnings.

nvidia-580.76.05-kernel7.patch

@@ -0,0 +1,147 @@
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/common/inc/nv-mm.h ./kernel-open/common/inc/nv-mm.h
+--- /tmp/nv-pristine/kernel-open/common/inc/nv-mm.h	2025-08-07 16:13:32.000000000 -0400
++++ ./kernel-open/common/inc/nv-mm.h	2026-08-10 18:38:01.233430548 -0400
+@@ -202,11 +202,11 @@
+ {
+ #if !NV_CAN_CALL_VMA_START_WRITE
+     nv_vma_start_write(vma);
+-    ACCESS_PRIVATE(vma, __vm_flags) |= flags;
++    vma_flags_set_word(&vma->flags, flags);
+ #elif defined(NV_VM_AREA_STRUCT_HAS_CONST_VM_FLAGS)
+     vm_flags_set(vma, flags);
+ #else
+-    vma->vm_flags |= flags;
++    vm_flags_set(vma, flags);
+ #endif
+ }
+ 
+@@ -214,11 +214,11 @@
+ {
+ #if !NV_CAN_CALL_VMA_START_WRITE
+     nv_vma_start_write(vma);
+-    ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
++    vma_flags_clear_word(&vma->flags, flags);
+ #elif defined(NV_VM_AREA_STRUCT_HAS_CONST_VM_FLAGS)
+     vm_flags_clear(vma, flags);
+ #else
+-    vma->vm_flags &= ~flags;
++    vm_flags_clear(vma, flags);
+ #endif
+ }
+ 
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/common/inc/nv-time.h ./kernel-open/common/inc/nv-time.h
+--- /tmp/nv-pristine/kernel-open/common/inc/nv-time.h	2025-08-07 16:13:29.000000000 -0400
++++ ./kernel-open/common/inc/nv-time.h	2026-08-10 18:36:21.200197921 -0400
+@@ -82,7 +82,7 @@
+     ktime_get_raw_ts64(&tm1);
+ #endif
+ 
+-    if (in_irq() && (us > NV_MAX_ISR_DELAY_US))
++    if (in_hardirq() && (us > NV_MAX_ISR_DELAY_US))
+         return NV_ERR_GENERIC;
+ 
+     mdelay_safe_msec = us / 1000;
+@@ -127,7 +127,7 @@
+     tm_start = tm_aux;
+ #endif
+ 
+-    if (in_irq() && (ms > NV_MAX_ISR_DELAY_MS))
++    if (in_hardirq() && (ms > NV_MAX_ISR_DELAY_MS))
+     {
+         return NV_ERR_GENERIC;
+     }
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/conftest.sh ./kernel-open/conftest.sh
+--- /tmp/nv-pristine/kernel-open/conftest.sh	2025-08-07 15:51:13.000000000 -0400
++++ ./kernel-open/conftest.sh	2026-08-10 18:38:01.205430483 -0400
+@@ -164,7 +164,7 @@
+     MACH_CFLAGS="$MACH_CFLAGS -I$SOURCE_ARCH_HEADERS/asm/mach-default"
+ 
+     CFLAGS="$BASE_CFLAGS $MACH_CFLAGS $OUTPUT_CFLAGS -include $AUTOCONF_FILE"
+-    CFLAGS="$CFLAGS -I$SOURCE_HEADERS"
++    CFLAGS="$CFLAGS -I$SOURCE_HEADERS -std=gnu11 -fms-extensions -Wno-error=implicit-function-declaration -Wno-error=int-conversion -Wno-error=incompatible-pointer-types"
+     CFLAGS="$CFLAGS -I$SOURCE_HEADERS/uapi"
+     CFLAGS="$CFLAGS -I$SOURCE_HEADERS/xen"
+     CFLAGS="$CFLAGS -I$OUTPUT_HEADERS/generated/uapi"
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/nvidia/nv-dma.c ./kernel-open/nvidia/nv-dma.c
+--- /tmp/nv-pristine/kernel-open/nvidia/nv-dma.c	2025-08-07 16:13:19.000000000 -0400
++++ ./kernel-open/nvidia/nv-dma.c	2026-08-10 18:36:21.214197952 -0400
+@@ -718,7 +718,7 @@
+ #endif
+     }
+ 
+-    return (ops->map_resource != NULL);
++    return (ops->map_phys != NULL);
+ }
+ 
+ /* DMA-map a peer device's C2C aperture for peer access. */
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/nvidia/nv-pci.c ./kernel-open/nvidia/nv-pci.c
+--- /tmp/nv-pristine/kernel-open/nvidia/nv-pci.c	2025-08-07 16:13:18.000000000 -0400
++++ ./kernel-open/nvidia/nv-pci.c	2026-08-10 18:37:07.883305841 -0400
+@@ -229,7 +229,7 @@
+ 
+ resize:
+     /* Attempt to resize BAR1 to the largest supported size */
+-    r = pci_resize_resource(pci_dev, NV_GPU_BAR1, requested_size);
++    r = pci_resize_resource(pci_dev, NV_GPU_BAR1, requested_size, 0);
+ 
+     if (r) {
+         if (r == -ENOSPC)
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/nvidia/os-interface.c ./kernel-open/nvidia/os-interface.c
+--- /tmp/nv-pristine/kernel-open/nvidia/os-interface.c	2025-08-07 16:13:20.000000000 -0400
++++ ./kernel-open/nvidia/os-interface.c	2026-08-10 18:36:21.201197924 -0400
+@@ -371,7 +371,7 @@
+ 
+ NvBool NV_API_CALL os_is_isr(void)
+ {
+-    return (in_irq());
++    return (in_hardirq());
+ }
+ 
+ // return TRUE if the caller is the super-user
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/nvidia-uvm/uvm_hmm.c ./kernel-open/nvidia-uvm/uvm_hmm.c
+--- /tmp/nv-pristine/kernel-open/nvidia-uvm/uvm_hmm.c	2025-08-07 16:13:56.000000000 -0400
++++ ./kernel-open/nvidia-uvm/uvm_hmm.c	2026-08-10 18:37:52.543410234 -0400
+@@ -2004,7 +2004,7 @@
+ 
+     hmm_mark_gpu_chunk_referenced(va_block, gpu, gpu_chunk);
+     UVM_ASSERT(!page_count(dpage));
+-    zone_device_page_init(dpage);
++    zone_device_page_init(dpage, page_pgmap(dpage), 0);
+     dpage->zone_device_data = gpu_chunk;
+     atomic64_inc(&va_block->hmm.va_space->hmm.allocated_page_count);
+ 
+diff -ruN '--exclude=*.log' '--exclude=*.orig' '--exclude=conftest*.c' /tmp/nv-pristine/kernel-open/nvidia-uvm/uvm_pmm_gpu.c ./kernel-open/nvidia-uvm/uvm_pmm_gpu.c
+--- /tmp/nv-pristine/kernel-open/nvidia-uvm/uvm_pmm_gpu.c	2025-08-07 16:13:54.000000000 -0400
++++ ./kernel-open/nvidia-uvm/uvm_pmm_gpu.c	2026-08-10 18:37:52.540410227 -0400
+@@ -3168,9 +3168,14 @@
+     UVM_ENTRY_RET(devmem_fault(vmf));
+ }
+ 
++static void devmem_folio_free(struct folio *folio)
++{
++    devmem_page_free(&folio->page);
++}
++
+ static const struct dev_pagemap_ops uvm_pmm_devmem_ops =
+ {
+-    .page_free = devmem_page_free,
++    .folio_free = devmem_folio_free,
+     .migrate_to_ram = devmem_fault_entry,
+ };
+ 
+@@ -3325,9 +3330,14 @@
+     nv_kref_put(&p2p_mem->refcount, device_p2p_page_free_wake);
+ }
+ 
++static void device_p2p_folio_free(struct folio *folio)
++{
++    device_p2p_page_free(&folio->page);
++}
++
+ static const struct dev_pagemap_ops uvm_device_p2p_pgmap_ops =
+ {
+-    .page_free = device_p2p_page_free,
++    .folio_free = device_p2p_folio_free,
+ };
+ 
+ void uvm_pmm_gpu_device_p2p_init(uvm_gpu_t *gpu)