535.161.07

Author: niv

CHANGELOG.md

@@ -2,6 +2,8 @@
 
 ## Release 535 Entries
 
+### [535.161.07] 2024-02-22
+
 ### [535.154.05] 2024-01-16
 
 ### [535.146.02] 2023-12-07

README.md

@@ -1,7 +1,7 @@
 # NVIDIA Linux Open GPU Kernel Module Source
 
 This is the source release of the NVIDIA Linux open GPU kernel modules,
-version 535.154.05.
+version 535.161.07.
 
 
 ## How to Build
@@ -17,7 +17,7 @@ as root:
 
 Note that the kernel modules built here must be used with GSP
 firmware and user-space NVIDIA GPU driver components from a corresponding
-535.154.05 driver release.  This can be achieved by installing
+535.161.07 driver release.  This can be achieved by installing
 the NVIDIA GPU driver from the .run file using the `--no-kernel-modules`
 option.  E.g.,
 
@@ -180,15 +180,15 @@ software applications.
 ## Compatible GPUs
 
 The open-gpu-kernel-modules can be used on any Turing or later GPU
-(see the table below). However, in the 535.154.05 release,
+(see the table below). However, in the 535.161.07 release,
 GeForce and Workstation support is still considered alpha-quality.
 
 To enable use of the open kernel modules on GeForce and Workstation GPUs,
 set the "NVreg_OpenRmEnableUnsupportedGpus" nvidia.ko kernel module
 parameter to 1. For more details, see the NVIDIA GPU driver end user
 README here:
 
-https://us.download.nvidia.com/XFree86/Linux-x86_64/535.154.05/README/kernel_open.html
+https://us.download.nvidia.com/XFree86/Linux-x86_64/535.161.07/README/kernel_open.html
 
 In the below table, if three IDs are listed, the first is the PCI Device 
 ID, the second is the PCI Subsystem Vendor ID, and the third is the PCI
@@ -746,12 +746,15 @@ Subsystem Device ID.
 | NVIDIA H800 PCIe                                | 2322 10DE 17A4 |
 | NVIDIA H800                                     | 2324 10DE 17A6 |
 | NVIDIA H800                                     | 2324 10DE 17A8 |
+| NVIDIA H20                                      | 2329 10DE 198B |
+| NVIDIA H20                                      | 2329 10DE 198C |
 | NVIDIA H100 80GB HBM3                           | 2330 10DE 16C0 |
 | NVIDIA H100 80GB HBM3                           | 2330 10DE 16C1 |
 | NVIDIA H100 PCIe                                | 2331 10DE 1626 |
 | NVIDIA H100                                     | 2339 10DE 17FC |
 | NVIDIA H800 NVL                                 | 233A 10DE 183A |
 | NVIDIA GH200 120GB                              | 2342 10DE 16EB |
+| NVIDIA GH200 120GB                              | 2342 10DE 1805 |
 | NVIDIA GH200 480GB                              | 2342 10DE 1809 |
 | NVIDIA GeForce RTX 3060 Ti                      | 2414           |
 | NVIDIA GeForce RTX 3080 Ti Laptop GPU           | 2420           |
@@ -805,6 +808,7 @@ Subsystem Device ID.
 | NVIDIA RTX A2000 12GB                           | 2571 10DE 1611 |
 | NVIDIA RTX A2000 12GB                           | 2571 17AA 1611 |
 | NVIDIA GeForce RTX 3050                         | 2582           |
+| NVIDIA GeForce RTX 3050                         | 2584           |
 | NVIDIA GeForce RTX 3050 Ti Laptop GPU           | 25A0           |
 | NVIDIA GeForce RTX 3050Ti Laptop GPU            | 25A0 103C 8928 |
 | NVIDIA GeForce RTX 3050Ti Laptop GPU            | 25A0 103C 89F9 |
@@ -837,7 +841,6 @@ Subsystem Device ID.
 | NVIDIA RTX A2000 Embedded GPU                   | 25FA           |
 | NVIDIA RTX A500 Embedded GPU                    | 25FB           |
 | NVIDIA GeForce RTX 4090                         | 2684           |
-| NVIDIA GeForce RTX 4090 D                       | 2685           |
 | NVIDIA RTX 6000 Ada Generation                  | 26B1 1028 16A1 |
 | NVIDIA RTX 6000 Ada Generation                  | 26B1 103C 16A1 |
 | NVIDIA RTX 6000 Ada Generation                  | 26B1 10DE 16A1 |
@@ -846,6 +849,7 @@ Subsystem Device ID.
 | NVIDIA RTX 5000 Ada Generation                  | 26B2 103C 17FA |
 | NVIDIA RTX 5000 Ada Generation                  | 26B2 10DE 17FA |
 | NVIDIA RTX 5000 Ada Generation                  | 26B2 17AA 17FA |
+| NVIDIA RTX 5880 Ada Generation                  | 26B3 103C 1934 |
 | NVIDIA RTX 5880 Ada Generation                  | 26B3 10DE 1934 |
 | NVIDIA L40                                      | 26B5 10DE 169D |
 | NVIDIA L40                                      | 26B5 10DE 17DA |

kernel-open/Kbuild

@@ -72,7 +72,7 @@ EXTRA_CFLAGS += -I$(src)/common/inc
 EXTRA_CFLAGS += -I$(src)
 EXTRA_CFLAGS += -Wall $(DEFINES) $(INCLUDES) -Wno-cast-qual -Wno-error -Wno-format-extra-args
 EXTRA_CFLAGS += -D__KERNEL__ -DMODULE -DNVRM
-EXTRA_CFLAGS += -DNV_VERSION_STRING=\"535.154.05\"
+EXTRA_CFLAGS += -DNV_VERSION_STRING=\"535.161.07\"
 
 ifneq ($(SYSSRCHOST1X),)
  EXTRA_CFLAGS += -I$(SYSSRCHOST1X)

kernel-open/common/inc/nv.h

@@ -615,6 +615,14 @@ typedef enum
 #define NV_IS_DEVICE_IN_SURPRISE_REMOVAL(nv)    \
         (((nv)->flags & NV_FLAG_IN_SURPRISE_REMOVAL) != 0)
 
+/*
+ * For console setup by EFI GOP, the base address is BAR1.
+ * For console setup by VBIOS, the base address is BAR2 + 16MB.
+ */
+#define NV_IS_CONSOLE_MAPPED(nv, addr)  \
+        (((addr) == (nv)->bars[NV_GPU_BAR_INDEX_FB].cpu_address) || \
+         ((addr) == ((nv)->bars[NV_GPU_BAR_INDEX_IMEM].cpu_address + 0x1000000)))
+
 #define NV_SOC_IS_ISO_IOMMU_PRESENT(nv)     \
         ((nv)->iso_iommu_present)
 
@@ -874,6 +882,8 @@ NvBool    NV_API_CALL nv_match_gpu_os_info(nv_state_t *, void *);
 NvU32     NV_API_CALL nv_get_os_type(void);
 
 void      NV_API_CALL nv_get_updated_emu_seg(NvU32 *start, NvU32 *end);
+void      NV_API_CALL nv_get_screen_info(nv_state_t *, NvU64 *, NvU16 *, NvU16 *, NvU16 *, NvU16 *, NvU64 *);
+
 struct dma_buf;
 typedef struct nv_dma_buf nv_dma_buf_t;
 struct drm_gem_object;

kernel-open/common/inc/os-interface.h

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 1999-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 1999-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: MIT
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -162,10 +162,9 @@ NvBool      NV_API_CALL  os_is_vgx_hyper             (void);
 NV_STATUS   NV_API_CALL  os_inject_vgx_msi           (NvU16, NvU64, NvU32);
 NvBool      NV_API_CALL  os_is_grid_supported        (void);
 NvU32       NV_API_CALL  os_get_grid_csp_support     (void);
-void        NV_API_CALL  os_get_screen_info          (NvU64 *, NvU16 *, NvU16 *, NvU16 *, NvU16 *, NvU64, NvU64);
 void        NV_API_CALL  os_bug_check                (NvU32, const char *);
 NV_STATUS   NV_API_CALL  os_lock_user_pages          (void *, NvU64, void **, NvU32);
-NV_STATUS   NV_API_CALL  os_lookup_user_io_memory    (void *, NvU64, NvU64 **, void**);
+NV_STATUS   NV_API_CALL  os_lookup_user_io_memory    (void *, NvU64, NvU64 **);
 NV_STATUS   NV_API_CALL  os_unlock_user_pages        (NvU64, void *);
 NV_STATUS   NV_API_CALL  os_match_mmap_offset        (void *, NvU64, NvU64 *);
 NV_STATUS   NV_API_CALL  os_get_euid                 (NvU32 *);

kernel-open/conftest.sh

@@ -5213,10 +5213,16 @@ compile_test() {
             # Added by commit 7b7b27214bba ("mm/memory_hotplug: introduce
             # add_memory_driver_managed()") in v5.8-rc1 (2020-06-05)
             #
+            # Before commit 3a0aaefe4134 ("mm/memory_hotplug: guard more
+            # declarations by CONFIG_MEMORY_HOTPLUG") in v5.10, the
+            # add_memory_driver_managed() was not guarded.
+            #
             CODE="
             #include <linux/memory_hotplug.h>
             void conftest_add_memory_driver_managed() {
+            #if defined(CONFIG_MEMORY_HOTPLUG)
                 add_memory_driver_managed();
+            #endif
             }"
 
             compile_check_conftest "$CODE" "NV_ADD_MEMORY_DRIVER_MANAGED_PRESENT" "" "functions"
@@ -6396,10 +6402,16 @@ compile_test() {
             # DRM_UNLOCKED was removed by commit 2798ffcc1d6a ("drm: Remove
             # locking for legacy ioctls and DRM_UNLOCKED") in Linux
             # next-20231208.
+            #
+            # DRM_UNLOCKED definition was moved from drmP.h to drm_ioctl.h by
+            # commit 2640981f3600 ("drm: document drm_ioctl.[hc]") in v4.12.
             CODE="
             #if defined(NV_DRM_DRM_IOCTL_H_PRESENT)
             #include <drm/drm_ioctl.h>
             #endif
+            #if defined(NV_DRM_DRMP_H_PRESENT)
+            #include <drm/drmP.h>
+            #endif
             int flags = DRM_UNLOCKED;"
 
             compile_check_conftest "$CODE" "NV_DRM_UNLOCKED_IOCTL_FLAG_PRESENT" "" "types"

kernel-open/nvidia/nv-mmap.c

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 1999-2021 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 1999-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: MIT
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -577,12 +577,9 @@ int nvidia_mmap_helper(
             //
             // This path is similar to the sysmem mapping code.
             // TODO: Refactor is needed as part of bug#2001704.
-            // Use pfn_valid to determine whether the physical address has
-            // backing struct page. This is used to isolate P8 from P9.
             //
             if ((nv_get_numa_status(nvl) == NV_NUMA_STATUS_ONLINE) &&
-                !IS_REG_OFFSET(nv, access_start, access_len) &&
-                (pfn_valid(PFN_DOWN(mmap_start))))
+                !IS_REG_OFFSET(nv, access_start, access_len))
             {
                 ret = nvidia_mmap_numa(vma, mmap_context);
                 if (ret)

kernel-open/nvidia/nv.c

@@ -5640,3 +5640,128 @@ NV_STATUS NV_API_CALL nv_get_egm_info(
     NV_DEV_PRINTF(NV_DBG_INFO, nv, "Cannot get EGM info\n");
     return NV_ERR_NOT_SUPPORTED;
 }
+
+void NV_API_CALL nv_get_screen_info(
+    nv_state_t  *nv,
+    NvU64       *pPhysicalAddress,
+    NvU16       *pFbWidth,
+    NvU16       *pFbHeight,
+    NvU16       *pFbDepth,
+    NvU16       *pFbPitch,
+    NvU64       *pFbSize
+)
+{
+    *pPhysicalAddress = 0;
+    *pFbWidth = *pFbHeight = *pFbDepth = *pFbPitch = *pFbSize = 0;
+
+#if defined(CONFIG_FB) && defined(NV_NUM_REGISTERED_FB_PRESENT)
+    if (num_registered_fb > 0)
+    {
+        int i;
+
+        for (i = 0; i < num_registered_fb; i++)
+        {
+            if (!registered_fb[i])
+                continue;
+
+            /* Make sure base address is mapped to GPU BAR */
+            if (NV_IS_CONSOLE_MAPPED(nv, registered_fb[i]->fix.smem_start))
+            {
+                *pPhysicalAddress = registered_fb[i]->fix.smem_start;
+                *pFbWidth = registered_fb[i]->var.xres;
+                *pFbHeight = registered_fb[i]->var.yres;
+                *pFbDepth = registered_fb[i]->var.bits_per_pixel;
+                *pFbPitch = registered_fb[i]->fix.line_length;
+                *pFbSize = (NvU64)(*pFbHeight) * (NvU64)(*pFbPitch);
+                return;
+            }
+        }
+    }
+#endif
+
+    /*
+     * If the screen info is not found in the registered FBs then fallback
+     * to the screen_info structure.
+     *
+     * The SYSFB_SIMPLEFB option, if enabled, marks VGA/VBE/EFI framebuffers as
+     * generic framebuffers so the new generic system-framebuffer drivers can
+     * be used instead. DRM_SIMPLEDRM drives the generic system-framebuffers
+     * device created by SYSFB_SIMPLEFB.
+     *
+     * SYSFB_SIMPLEFB registers a dummy framebuffer which does not contain the
+     * information required by nv_get_screen_info(), therefore you need to
+     * fall back onto the screen_info structure.
+     *
+     * After commit b8466fe82b79 ("efi: move screen_info into efi init code")
+     * in v6.7, 'screen_info' is exported as GPL licensed symbol for ARM64.
+     */
+
+#if NV_CHECK_EXPORT_SYMBOL(screen_info)
+    /*
+     * If there is not a framebuffer console, return 0 size.
+     *
+     * orig_video_isVGA is set to 1 during early Linux kernel
+     * initialization, and then will be set to a value, such as
+     * VIDEO_TYPE_VLFB or VIDEO_TYPE_EFI if an fbdev console is used.
+     */
+    if (screen_info.orig_video_isVGA > 1)
+    {
+        NvU64 physAddr = screen_info.lfb_base;
+#if defined(VIDEO_CAPABILITY_64BIT_BASE)
+        physAddr |= (NvU64)screen_info.ext_lfb_base << 32;
+#endif
+
+        /* Make sure base address is mapped to GPU BAR */
+        if (NV_IS_CONSOLE_MAPPED(nv, physAddr))
+        {
+            *pPhysicalAddress = physAddr;
+            *pFbWidth = screen_info.lfb_width;
+            *pFbHeight = screen_info.lfb_height;
+            *pFbDepth = screen_info.lfb_depth;
+            *pFbPitch = screen_info.lfb_linelength;
+            *pFbSize = (NvU64)(*pFbHeight) * (NvU64)(*pFbPitch);
+        }
+    }
+#else
+    {
+        nv_linux_state_t *nvl = NV_GET_NVL_FROM_NV_STATE(nv);
+        struct pci_dev *pci_dev = nvl->pci_dev;
+        int i;
+
+        if (pci_dev == NULL)
+            return;
+
+        BUILD_BUG_ON(NV_GPU_BAR_INDEX_IMEM != NV_GPU_BAR_INDEX_FB + 1);
+        for (i = NV_GPU_BAR_INDEX_FB; i <= NV_GPU_BAR_INDEX_IMEM; i++)
+        {
+            int bar_index = nv_bar_index_to_os_bar_index(pci_dev, i);
+            struct resource *gpu_bar_res = &pci_dev->resource[bar_index];
+            struct resource *res = gpu_bar_res->child;
+
+            /*
+             * Console resource will become child resource of pci-dev resource.
+             * Check if child resource start address matches with expected
+             * console start address.
+             */
+            if ((res != NULL) &&
+                NV_IS_CONSOLE_MAPPED(nv, res->start))
+            {
+                NvU32 res_name_len = strlen(res->name);
+
+                /*
+                 * The resource name ends with 'fb' (efifb, vesafb, etc.).
+                 * For simple-framebuffer, the resource name is 'BOOTFB'.
+                 * Confirm if the resources name either ends with 'fb' or 'FB'.
+                 */
+                if ((res_name_len > 2) &&
+                    !strcasecmp((res->name + res_name_len - 2), "fb"))
+                {
+                    *pPhysicalAddress = res->start;
+                    *pFbSize = resource_size(res);
+                    return;
+                }
+            }
+        }
+    }
+#endif
+}