550.67

Author: niv

CHANGELOG.md

@@ -2,6 +2,8 @@
 
 ## Release 550 Entries
 
+### [550.67] 2024-03-19
+
 ### [550.54.15] 2024-03-18
 
 ### [550.54.14] 2024-02-23

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 550.54.15.
+version 550.67.
 
 
 ## 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
-550.54.15 driver release.  This can be achieved by installing
+550.67 driver release.  This can be achieved by installing
 the NVIDIA GPU driver from the .run file using the `--no-kernel-modules`
 option.  E.g.,
 
@@ -188,7 +188,7 @@ encountered specific to them.
 For details on feature support and limitations, see the NVIDIA GPU driver
 end user README here:
 
-https://us.download.nvidia.com/XFree86/Linux-x86_64/550.54.15/README/kernel_open.html
+https://us.download.nvidia.com/XFree86/Linux-x86_64/550.67/README/kernel_open.html
 
 For vGPU support, please refer to the README.vgpu packaged in the vGPU Host
 Package for more details.
@@ -867,13 +867,15 @@ Subsystem Device ID.
 | NVIDIA GeForce RTX 4080 SUPER                   | 2702           |
 | NVIDIA GeForce RTX 4080                         | 2704           |
 | NVIDIA GeForce RTX 4070 Ti SUPER                | 2705           |
+| NVIDIA GeForce RTX 4070                         | 2709           |
 | NVIDIA GeForce RTX 4090 Laptop GPU              | 2717           |
 | NVIDIA RTX 5000 Ada Generation Laptop GPU       | 2730           |
 | NVIDIA GeForce RTX 4090 Laptop GPU              | 2757           |
 | NVIDIA RTX 5000 Ada Generation Embedded GPU     | 2770           |
 | NVIDIA GeForce RTX 4070 Ti                      | 2782           |
 | NVIDIA GeForce RTX 4070 SUPER                   | 2783           |
 | NVIDIA GeForce RTX 4070                         | 2786           |
+| NVIDIA GeForce RTX 4060 Ti                      | 2788           |
 | NVIDIA GeForce RTX 4080 Laptop GPU              | 27A0           |
 | NVIDIA RTX 4000 SFF Ada Generation              | 27B0 1028 16FA |
 | NVIDIA RTX 4000 SFF Ada Generation              | 27B0 103C 16FA |
@@ -896,6 +898,7 @@ Subsystem Device ID.
 | NVIDIA RTX 3500 Ada Generation Embedded GPU     | 27FB           |
 | NVIDIA GeForce RTX 4060 Ti                      | 2803           |
 | NVIDIA GeForce RTX 4060 Ti                      | 2805           |
+| NVIDIA GeForce RTX 4060                         | 2808           |
 | NVIDIA GeForce RTX 4070 Laptop GPU              | 2820           |
 | NVIDIA RTX 3000 Ada Generation Laptop GPU       | 2838           |
 | NVIDIA GeForce RTX 4070 Laptop GPU              | 2860           |

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-format-extra-args
 EXTRA_CFLAGS += -D__KERNEL__ -DMODULE -DNVRM
-EXTRA_CFLAGS += -DNV_VERSION_STRING=\"550.54.15\"
+EXTRA_CFLAGS += -DNV_VERSION_STRING=\"550.67\"
 
 ifneq ($(SYSSRCHOST1X),)
  EXTRA_CFLAGS += -I$(SYSSRCHOST1X)
@@ -170,6 +170,8 @@ NV_CONFTEST_CMD := /bin/sh $(NV_CONFTEST_SCRIPT) \
 NV_CFLAGS_FROM_CONFTEST := $(shell $(NV_CONFTEST_CMD) build_cflags)
 
 NV_CONFTEST_CFLAGS = $(NV_CFLAGS_FROM_CONFTEST) $(EXTRA_CFLAGS) -fno-pie
+NV_CONFTEST_CFLAGS += $(call cc-disable-warning,pointer-sign)
+NV_CONFTEST_CFLAGS += $(call cc-option,-fshort-wchar,)
 NV_CONFTEST_CFLAGS += -Wno-error
 
 NV_CONFTEST_COMPILE_TEST_HEADERS := $(obj)/conftest/macros.h

kernel-open/common/inc/nv-linux.h

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2001-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2001-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: MIT
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -1989,31 +1989,6 @@ static inline NvBool nv_platform_use_auto_online(nv_linux_state_t *nvl)
     return nvl->numa_info.use_auto_online;
 }
 
-typedef struct {
-    NvU64 base;
-    NvU64 size;
-    NvU32 nodeId;
-    int ret;
-} remove_numa_memory_info_t;
-
-static void offline_numa_memory_callback
-(
-    void *args
-)
-{
-#ifdef NV_OFFLINE_AND_REMOVE_MEMORY_PRESENT
-    remove_numa_memory_info_t *pNumaInfo = (remove_numa_memory_info_t *)args;
-#ifdef NV_REMOVE_MEMORY_HAS_NID_ARG
-    pNumaInfo->ret = offline_and_remove_memory(pNumaInfo->nodeId,
-                                               pNumaInfo->base,
-                                               pNumaInfo->size);
-#else
-    pNumaInfo->ret = offline_and_remove_memory(pNumaInfo->base,
-                                               pNumaInfo->size);
-#endif
-#endif
-}
-
 typedef enum
 {
     NV_NUMA_STATUS_DISABLED             = 0,

kernel-open/conftest.sh

@@ -3096,6 +3096,22 @@ compile_test() {
 
         ;;
 
+        foll_longterm_present)
+            #
+            # Determine if FOLL_LONGTERM enum is present or not
+            #
+            # Added by commit 932f4a630a69 ("mm/gup: replace
+            # get_user_pages_longterm() with FOLL_LONGTERM") in
+            # v5.2
+            #
+            CODE="
+            #include <linux/mm.h>
+            int foll_longterm = FOLL_LONGTERM;
+            "
+
+            compile_check_conftest "$CODE" "NV_FOLL_LONGTERM_PRESENT" "" "types"
+        ;;
+
         vfio_pin_pages_has_vfio_device_arg)
             #
             # Determine if vfio_pin_pages() kABI accepts "struct vfio_device *"
@@ -5152,11 +5168,15 @@ compile_test() {
             # commit 49a3f51dfeee ("drm/gem: Use struct dma_buf_map in GEM
             # vmap ops and convert GEM backends") in v5.11.
             #
+            # Note that the 'map' argument type is changed from 'struct dma_buf_map'
+            # to 'struct iosys_map' by commit 7938f4218168 ("dma-buf-map: Rename
+            # to iosys-map) in v5.18.
+            #
             CODE="
             #include <drm/drm_gem.h>
             int conftest_drm_gem_object_vmap_has_map_arg(
-                    struct drm_gem_object *obj, struct dma_buf_map *map) {
-                return obj->funcs->vmap(obj, map);
+                    struct drm_gem_object *obj) {
+                return obj->funcs->vmap(obj, NULL);
             }"
 
             compile_check_conftest "$CODE" "NV_DRM_GEM_OBJECT_VMAP_HAS_MAP_ARG" "" "types"

kernel-open/nvidia-drm/nvidia-drm-drv.c

@@ -1903,8 +1903,33 @@ void nv_drm_remove_devices(void)
  */
 void nv_drm_suspend_resume(NvBool suspend)
 {
+    static DEFINE_MUTEX(nv_drm_suspend_mutex);
+    static NvU32 nv_drm_suspend_count = 0;
+    struct nv_drm_device *nv_dev;
+
+    mutex_lock(&nv_drm_suspend_mutex);
+
+    /*
+     * Count the number of times the driver is asked to suspend. Suspend all DRM
+     * devices on the first suspend call and resume them on the last resume
+     * call.  This is necessary because the kernel may call nvkms_suspend()
+     * simultaneously for each GPU, but NVKMS itself also suspends all GPUs on
+     * the first call.
+     */
+    if (suspend) {
+        if (nv_drm_suspend_count++ > 0) {
+            goto done;
+        }
+    } else {
+        BUG_ON(nv_drm_suspend_count == 0);
+
+        if (--nv_drm_suspend_count > 0) {
+            goto done;
+        }
+    }
+
 #if defined(NV_DRM_ATOMIC_MODESET_AVAILABLE)
-    struct nv_drm_device *nv_dev = dev_list;
+    nv_dev = dev_list;
 
     /*
      * NVKMS shuts down all heads on suspend. Update DRM state accordingly.
@@ -1930,6 +1955,9 @@ void nv_drm_suspend_resume(NvBool suspend)
         }
     }
 #endif /* NV_DRM_ATOMIC_MODESET_AVAILABLE */
+
+done:
+    mutex_unlock(&nv_drm_suspend_mutex);
 }
 
 #endif /* NV_DRM_AVAILABLE */

kernel-open/nvidia-modeset/nvidia-modeset-linux.c

@@ -56,7 +56,11 @@
 #include "nv-lock.h"
 #include "nv-chardev-numbers.h"
 
-#if !defined(CONFIG_RETPOLINE)
+/*
+ * Commit aefb2f2e619b ("x86/bugs: Rename CONFIG_RETPOLINE =>
+ * CONFIG_MITIGATION_RETPOLINE) in v6.8 renamed CONFIG_RETPOLINE.
+ */
+#if !defined(CONFIG_RETPOLINE) && !defined(CONFIG_MITIGATION_RETPOLINE)
 #include "nv-retpoline.h"
 #endif
 
@@ -499,8 +503,9 @@ nvkms_event_queue_changed(nvkms_per_open_handle_t *pOpenKernel,
 
 static void nvkms_suspend(NvU32 gpuId)
 {
+    nvKmsKapiSuspendResume(NV_TRUE /* suspend */);
+
     if (gpuId == 0) {
-        nvKmsKapiSuspendResume(NV_TRUE /* suspend */);
         nvkms_write_lock_pm_lock();
     }
 
@@ -517,8 +522,9 @@ static void nvkms_resume(NvU32 gpuId)
 
     if (gpuId == 0) {
         nvkms_write_unlock_pm_lock();
-        nvKmsKapiSuspendResume(NV_FALSE /* suspend */);
     }
+
+    nvKmsKapiSuspendResume(NV_FALSE /* suspend */);
 }
 
 

kernel-open/nvidia-uvm/uvm_channel_test.c

@@ -691,12 +691,16 @@ static NV_STATUS stress_test_all_gpus_in_va(uvm_va_space_t *va_space,
             if (uvm_test_rng_range_32(&rng, 0, 1) == 0) {
                 NvU32 random_stream_index = uvm_test_rng_range_32(&rng, 0, num_streams - 1);
                 uvm_test_stream_t *random_stream = &streams[random_stream_index];
-                uvm_push_acquire_tracker(&stream->push, &random_stream->tracker);
-                snapshot_counter(&stream->push,
-                                 random_stream->counter_mem,
-                                 stream->other_stream_counter_snapshots_mem,
-                                 i,
-                                 random_stream->queued_counter_repeat);
+
+                if ((random_stream->push.gpu == gpu) || uvm_push_allow_dependencies_across_gpus()) {
+                    uvm_push_acquire_tracker(&stream->push, &random_stream->tracker);
+
+                    snapshot_counter(&stream->push,
+                                     random_stream->counter_mem,
+                                     stream->other_stream_counter_snapshots_mem,
+                                     i,
+                                     random_stream->queued_counter_repeat);
+                }
             }
 
             uvm_push_end(&stream->push);

kernel-open/nvidia-uvm/uvm_fault_buffer_flush_test.c

@@ -51,8 +51,10 @@ NV_STATUS uvm_test_fault_buffer_flush(UVM_TEST_FAULT_BUFFER_FLUSH_PARAMS *params
 
     uvm_va_space_up_read(va_space);
 
-    if (uvm_processor_mask_empty(retained_gpus))
-        return NV_ERR_INVALID_DEVICE;
+    if (uvm_processor_mask_empty(retained_gpus)) {
+        status = NV_ERR_INVALID_DEVICE;
+        goto out;
+    }
 
     for (i = 0; i < params->iterations; i++) {
         if (fatal_signal_pending(current)) {

kernel-open/nvidia-uvm/uvm_global.h

@@ -409,4 +409,10 @@ NV_STATUS uvm_service_block_context_init(void);
 // Release fault service contexts if any exist.
 void uvm_service_block_context_exit(void);
 
+// Allocate a service block context
+uvm_service_block_context_t *uvm_service_block_context_alloc(struct mm_struct *mm);
+
+// Free a servic block context
+void uvm_service_block_context_free(uvm_service_block_context_t *service_context);
+
 #endif // __UVM_GLOBAL_H__