535.104.05

Author: niv

CHANGELOG.md

@@ -2,6 +2,8 @@
 
 ## Release 535 Entries
 
+### [535.104.05] 2023-08-22
+
 ### [535.98] 2023-08-08
 
 ### [535.86.10] 2023-07-31

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 535.98.
+version 535.104.05.
 
 
 ## 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
-535.98 driver release.  This can be achieved by installing
+535.104.05 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 535.98 release,
+(see the table below). However, in the 535.104.05 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_64/535.98/README/kernel_open.html
+https://us.download.nvidia.com/XFree86/Linux-x86_64/535.104.05/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
@@ -665,6 +665,7 @@ Subsystem Device ID.
 | NVIDIA PG506-232                                | 20B6 10DE 1492 |
 | NVIDIA A30                                      | 20B7 10DE 1532 |
 | NVIDIA A30                                      | 20B7 10DE 1804 |
+| NVIDIA A800-SXM4-40GB                           | 20BD 10DE 17F4 |
 | NVIDIA A100-PCIE-40GB                           | 20F1 10DE 145F |
 | NVIDIA A800-SXM4-80GB                           | 20F3 10DE 179B |
 | NVIDIA A800-SXM4-80GB                           | 20F3 10DE 179C |
@@ -676,6 +677,10 @@ Subsystem Device ID.
 | NVIDIA A800-SXM4-80GB                           | 20F3 10DE 17A2 |
 | NVIDIA A800 80GB PCIe                           | 20F5 10DE 1799 |
 | NVIDIA A800 80GB PCIe LC                        | 20F5 10DE 179A |
+| NVIDIA A800 40GB Active                         | 20F6 1028 180A |
+| NVIDIA A800 40GB Active                         | 20F6 103C 180A |
+| NVIDIA A800 40GB Active                         | 20F6 10DE 180A |
+| NVIDIA A800 40GB Active                         | 20F6 17AA 180A |
 | NVIDIA GeForce GTX 1660 Ti                      | 2182           |
 | NVIDIA GeForce GTX 1660                         | 2184           |
 | NVIDIA GeForce GTX 1650 SUPER                   | 2187           |
@@ -734,13 +739,15 @@ Subsystem Device ID.
 | NVIDIA A10                                      | 2236 10DE 1482 |
 | NVIDIA A10G                                     | 2237 10DE 152F |
 | NVIDIA A10M                                     | 2238 10DE 1677 |
+| NVIDIA H100 NVL                                 | 2321 10DE 1839 |
 | NVIDIA H800 PCIe                                | 2322 10DE 17A4 |
 | NVIDIA H800                                     | 2324 10DE 17A6 |
 | NVIDIA H800                                     | 2324 10DE 17A8 |
 | 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 GeForce RTX 3060 Ti                      | 2414           |
 | NVIDIA GeForce RTX 3080 Ti Laptop GPU           | 2420           |
 | NVIDIA RTX A5500 Laptop GPU                     | 2438           |
@@ -835,10 +842,13 @@ Subsystem Device ID.
 | NVIDIA RTX 5000 Ada Generation                  | 26B2 17AA 17FA |
 | NVIDIA L40                                      | 26B5 10DE 169D |
 | NVIDIA L40                                      | 26B5 10DE 17DA |
+| NVIDIA L40S                                     | 26B9 10DE 1851 |
+| NVIDIA L40S                                     | 26B9 10DE 18CF |
 | NVIDIA GeForce RTX 4080                         | 2704           |
 | 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                         | 2786           |
 | NVIDIA GeForce RTX 4080 Laptop GPU              | 27A0           |
@@ -855,6 +865,7 @@ Subsystem Device ID.
 | NVIDIA RTX 4000 Ada Generation Laptop GPU       | 27BA           |
 | NVIDIA RTX 3500 Ada Generation Laptop GPU       | 27BB           |
 | NVIDIA GeForce RTX 4080 Laptop GPU              | 27E0           |
+| NVIDIA RTX 3500 Ada Generation Embedded GPU     | 27FB           |
 | NVIDIA GeForce RTX 4060 Ti                      | 2803           |
 | NVIDIA GeForce RTX 4060 Ti                      | 2805           |
 | NVIDIA GeForce RTX 4070 Laptop GPU              | 2820           |
@@ -866,3 +877,4 @@ Subsystem Device ID.
 | NVIDIA RTX 2000 Ada Generation Laptop GPU       | 28B8           |
 | NVIDIA GeForce RTX 4060 Laptop GPU              | 28E0           |
 | NVIDIA GeForce RTX 4050 Laptop GPU              | 28E1           |
+| NVIDIA RTX 2000 Ada Generation Embedded GPU     | 28F8           |

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=\"535.98\"
+EXTRA_CFLAGS += -DNV_VERSION_STRING=\"535.104.05\"
 
 ifneq ($(SYSSRCHOST1X),)
  EXTRA_CFLAGS += -I$(SYSSRCHOST1X)

kernel-open/common/inc/nv_uvm_types.h

@@ -566,8 +566,11 @@ typedef struct UvmPlatformInfo_tag
     // Out: ATS (Address Translation Services) is supported
     NvBool atsSupported;
 
-    // Out: AMD SEV (Secure Encrypted Virtualization) is enabled
-    NvBool sevEnabled;
+    // Out: True if HW trusted execution, such as AMD's SEV-SNP or Intel's TDX,
+    // is enabled in the VM, indicating that Confidential Computing must be
+    // also enabled in the GPU(s); these two security features are either both
+    // enabled, or both disabled.
+    NvBool confComputingEnabled;
 } UvmPlatformInfo;
 
 typedef struct UvmGpuClientInfo_tag

kernel-open/conftest.sh

@@ -6341,6 +6341,21 @@ compile_test() {
             compile_check_conftest "$CODE" "NV_MEMPOLICY_HAS_HOME_NODE" "" "types"
         ;;
 
+        mmu_interval_notifier)
+            #
+            # Determine if mmu_interval_notifier struct is present or not
+            #
+            # Added by commit 99cb252f5 ("mm/mmu_notifier: add an interval tree
+            # notifier") in v5.10 (2019-11-12).
+            #
+            CODE="
+            #include <linux/mmu_notifier.h>
+            struct mmu_interval_notifier interval_notifier;
+            "
+
+            compile_check_conftest "$CODE" "NV_MMU_INTERVAL_NOTIFIER" "" "types"
+        ;;
+
         # When adding a new conftest entry, please use the correct format for
         # specifying the relevant upstream Linux kernel commit.
         #

kernel-open/nvidia-uvm/nvidia-uvm.Kbuild

@@ -110,5 +110,6 @@ NV_CONFTEST_TYPE_COMPILE_TESTS += handle_mm_fault_has_mm_arg
 NV_CONFTEST_TYPE_COMPILE_TESTS += handle_mm_fault_has_pt_regs_arg
 NV_CONFTEST_TYPE_COMPILE_TESTS += mempolicy_has_unified_nodes
 NV_CONFTEST_TYPE_COMPILE_TESTS += mempolicy_has_home_node
+NV_CONFTEST_TYPE_COMPILE_TESTS += mmu_interval_notifier
 
 NV_CONFTEST_SYMBOL_COMPILE_TESTS += is_export_symbol_present_int_active_memcg

kernel-open/nvidia-uvm/uvm_ats.c

@@ -44,6 +44,8 @@ void uvm_ats_init(const UvmPlatformInfo *platform_info)
 
 void uvm_ats_init_va_space(uvm_va_space_t *va_space)
 {
+    uvm_init_rwsem(&va_space->ats.lock, UVM_LOCK_ORDER_LEAF);
+
     if (UVM_ATS_IBM_SUPPORTED())
         uvm_ats_ibm_init_va_space(va_space);
 }

kernel-open/nvidia-uvm/uvm_ats.h

@@ -28,17 +28,32 @@
 #include "uvm_forward_decl.h"
 #include "uvm_ats_ibm.h"
 #include "nv_uvm_types.h"
+#include "uvm_lock.h"
 
     #include "uvm_ats_sva.h"
 
     #define UVM_ATS_SUPPORTED() (UVM_ATS_IBM_SUPPORTED() || UVM_ATS_SVA_SUPPORTED())
 
+// ATS prefetcher uses hmm_range_fault() to query residency information.
+// hmm_range_fault() needs CONFIG_HMM_MIRROR. To detect racing CPU invalidates
+// of memory regions while hmm_range_fault() is being called, MMU interval
+// notifiers are needed.
+    #if defined(CONFIG_HMM_MIRROR) && defined(NV_MMU_INTERVAL_NOTIFIER)
+        #define UVM_ATS_PREFETCH_SUPPORTED() 1
+    #else
+        #define UVM_ATS_PREFETCH_SUPPORTED() 0
+    #endif
+
 typedef struct
 {
     // Mask of gpu_va_spaces which are registered for ATS access. The mask is
     // indexed by gpu->id. This mask is protected by the VA space lock.
     uvm_processor_mask_t registered_gpu_va_spaces;
 
+    // Protects racing invalidates in the VA space while hmm_range_fault() is
+    // being called in ats_compute_residency_mask().
+    uvm_rw_semaphore_t lock;
+
     union
     {
         uvm_ibm_va_space_t ibm;