535.183.06

NVIDIA · Jul 9, 2024 · c588c38 · c588c38
1 parent 4459285
commit c588c38
Show file tree

Hide file tree

Showing 18 changed files with 292 additions and 82 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,6 +2,8 @@
 
 ## Release 535 Entries
 
+### [535.183.06] 2024-07-09
+
 ### [535.183.01] 2024-06-04
 
 ### [535.179] 2024-05-09

diff --git a/README.md b/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 535.183.01.
+version 535.183.06.
 
 
 ## 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
-535.183.01 driver release.  This can be achieved by installing
+535.183.06 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 535.183.01 release,
+(see the table below). However, in the 535.183.06 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_64/535.183.01/README/kernel_open.html
+https://us.download.nvidia.com/XFree86/Linux-x86_64/535.183.06/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

diff --git a/kernel-open/Kbuild b/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=\"535.183.01\"
+EXTRA_CFLAGS += -DNV_VERSION_STRING=\"535.183.06\"
 
 ifneq ($(SYSSRCHOST1X),)
  EXTRA_CFLAGS += -I$(SYSSRCHOST1X)

diff --git a/kernel-open/conftest.sh b/kernel-open/conftest.sh
@@ -1208,6 +1208,23 @@ compile_test() {
             compile_check_conftest "$CODE" "NV_VFIO_DEVICE_OPS_HAS_BIND_IOMMUFD" "" "types"
         ;;
 
+        vfio_device_ops_has_detach_ioas)
+            #
+            # Determine if 'vfio_device_ops' struct has 'detach_ioas' field.
+            #
+            # Added by commit 9048c7341c4df9cae04c154a8b0f556dbe913358 ("vfio-iommufd: Add detach_ioas
+            # support for physical VFIO devices
+            #
+            CODE="
+            #include <linux/pci.h>
+            #include <linux/vfio.h>
+            int conftest_vfio_device_ops_has_detach_ioas(void) {
+                return offsetof(struct vfio_device_ops, detach_ioas);
+            }"
+
+            compile_check_conftest "$CODE" "NV_VFIO_DEVICE_OPS_HAS_DETACH_IOAS" "" "types"
+        ;;
+
         pci_irq_vector_helpers)
             #
             # Determine if pci_alloc_irq_vectors(), pci_free_irq_vectors()
@@ -6772,10 +6789,12 @@ case "$5" in
         #
         VERBOSE=$6
         iommu=CONFIG_VFIO_IOMMU_TYPE1
+        iommufd_vfio_container=CONFIG_IOMMUFD_VFIO_CONTAINER
         mdev=CONFIG_VFIO_MDEV
         kvm=CONFIG_KVM_VFIO
         vfio_pci_core=CONFIG_VFIO_PCI_CORE
         VFIO_IOMMU_PRESENT=0
+        VFIO_IOMMUFD_VFIO_CONTAINER_PRESENT=0
         VFIO_MDEV_PRESENT=0
         KVM_PRESENT=0
         VFIO_PCI_CORE_PRESENT=0
@@ -6785,6 +6804,10 @@ case "$5" in
                 VFIO_IOMMU_PRESENT=1
             fi
 
+            if (test_configuration_option ${iommufd_vfio_container} || test_configuration_option ${iommufd_vfio_container}_MODULE); then
+                VFIO_IOMMUFD_VFIO_CONTAINER_PRESENT=1
+            fi
+
             if (test_configuration_option ${mdev} || test_configuration_option ${mdev}_MODULE); then
                 VFIO_MDEV_PRESENT=1
             fi
@@ -6797,7 +6820,7 @@ case "$5" in
                 VFIO_PCI_CORE_PRESENT=1
             fi
 
-            if [ "$VFIO_IOMMU_PRESENT" != "0" ] && [ "$KVM_PRESENT" != "0" ] ; then
+            if ([ "$VFIO_IOMMU_PRESENT" != "0" ] || [ "$VFIO_IOMMUFD_VFIO_CONTAINER_PRESENT" != "0" ])&& [ "$KVM_PRESENT" != "0" ] ; then
                 # vGPU requires either MDEV or vfio-pci-core framework to be present.
                 if [ "$VFIO_MDEV_PRESENT" != "0" ] || [ "$VFIO_PCI_CORE_PRESENT" != "0" ]; then
                     exit 0
@@ -6806,8 +6829,8 @@ case "$5" in
 
             echo "Below CONFIG options are missing on the kernel for installing";
             echo "NVIDIA vGPU driver on KVM host";
-            if [ "$VFIO_IOMMU_PRESENT" = "0" ]; then
-                echo "CONFIG_VFIO_IOMMU_TYPE1";
+            if [ "$VFIO_IOMMU_PRESENT" = "0" ] && [ "$VFIO_IOMMUFD_VFIO_CONTAINER_PRESENT" = "0" ]; then
+                echo "either CONFIG_VFIO_IOMMU_TYPE1 or CONFIG_IOMMUFD_VFIO_CONTAINER";
             fi
 
             if [ "$VFIO_MDEV_PRESENT" = "0" ] && [ "$VFIO_PCI_CORE_PRESENT" = "0" ]; then

diff --git a/src/common/inc/nvBldVer.h b/src/common/inc/nvBldVer.h
@@ -43,18 +43,18 @@
 #endif
 
 #if defined(NV_LINUX) || defined(NV_BSD) || defined(NV_SUNOS)
-#define NV_BUILD_BRANCH_VERSION         "rel/gpu_drv/r535/r538_67-552"
-#define NV_BUILD_CHANGELIST_NUM         (34280977)
+#define NV_BUILD_BRANCH_VERSION         "rel/gpu_drv/r535/r538_67-557"
+#define NV_BUILD_CHANGELIST_NUM         (34462766)
 #define NV_BUILD_TYPE                   "Official"
-#define NV_BUILD_NAME                   "rel/gpu_drv/r535/r538_67-552"
-#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34280977)
+#define NV_BUILD_NAME                   "rel/gpu_drv/r535/r538_67-557"
+#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34462766)
 
 #else     /* Windows builds */
-#define NV_BUILD_BRANCH_VERSION         "r538_67-1"
-#define NV_BUILD_CHANGELIST_NUM         (34280977)
+#define NV_BUILD_BRANCH_VERSION         "r538_67-2"
+#define NV_BUILD_CHANGELIST_NUM         (34397468)
 #define NV_BUILD_TYPE                   "Official"
-#define NV_BUILD_NAME                   "538.69"
-#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34280977)
+#define NV_BUILD_NAME                   "538.78"
+#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34397468)
 #define NV_BUILD_BRANCH_BASE_VERSION    R535
 #endif
 // End buildmeister python edited section

diff --git a/src/common/inc/nvUnixVersion.h b/src/common/inc/nvUnixVersion.h
@@ -4,7 +4,7 @@
 #if defined(NV_LINUX) || defined(NV_BSD) || defined(NV_SUNOS) || defined(NV_VMWARE) || defined(NV_QNX) || defined(NV_INTEGRITY) || \
     (defined(RMCFG_FEATURE_PLATFORM_GSP) && RMCFG_FEATURE_PLATFORM_GSP == 1)
 
-#define NV_VERSION_STRING               "535.183.01"
+#define NV_VERSION_STRING               "535.183.06"
 
 #else
 

diff --git a/src/common/nvlink/interface/nvlink.h b/src/common/nvlink/interface/nvlink.h
@@ -500,7 +500,8 @@ NvlStatus nvlink_lib_get_remote_conn_info(nvlink_link *link, nvlink_conn_info *c
  */
 NvlStatus nvlink_lib_discover_and_get_remote_conn_info(nvlink_link      *end,
                                                        nvlink_conn_info *conn_info,
-                                                       NvU32             flags);
+                                                       NvU32             flags,
+                                                       NvBool            bForceDiscovery);
 
 
 /************************************************************************************************/

diff --git a/src/common/nvlink/kernel/nvlink/core/nvlink_discovery.c b/src/common/nvlink/kernel/nvlink/core/nvlink_discovery.c
@@ -47,7 +47,8 @@ nvlink_core_discover_and_get_remote_end
 (
     nvlink_link  *end,
     nvlink_link **remote_end,
-    NvU32         flags
+    NvU32         flags,
+    NvBool        bForceDiscovery
 )
 {
     nvlink_intranode_conn *conn      = NULL;
@@ -67,7 +68,7 @@ nvlink_core_discover_and_get_remote_end
         return;
     }
 
-    if (nvlinkLibCtx.bNewEndpoints)
+    if (nvlinkLibCtx.bNewEndpoints || bForceDiscovery)
     {
         if (!_nvlink_core_all_links_initialized())
         {

diff --git a/src/common/nvlink/kernel/nvlink/interface/nvlink_ioctl_entry.c b/src/common/nvlink/kernel/nvlink/interface/nvlink_ioctl_entry.c
@@ -1481,7 +1481,7 @@ _nvlink_lib_ctrl_device_discover_peer_link
         (linkMode == NVLINK_LINKSTATE_SLEEP))
     {
         nvlink_link   *remoteLink = NULL;
-        nvlink_core_discover_and_get_remote_end(link, &remoteLink, 0);
+        nvlink_core_discover_and_get_remote_end(link, &remoteLink, 0, NV_FALSE);
         if (remoteLink == NULL)
         {
             NVLINK_PRINT((DBG_MODULE_NVLINK_CORE, NVLINK_DBG_LEVEL_INFO,

diff --git a/src/common/nvlink/kernel/nvlink/interface/nvlink_kern_discovery_entry.c b/src/common/nvlink/kernel/nvlink/interface/nvlink_kern_discovery_entry.c
@@ -182,7 +182,8 @@ nvlink_lib_discover_and_get_remote_conn_info
 (
     nvlink_link      *end,
     nvlink_conn_info *conn_info,
-    NvU32             flags
+    NvU32             flags,
+    NvBool            bForceDiscovery
 )
 {
     NvlStatus      status                            = NVL_SUCCESS;
@@ -257,7 +258,8 @@ nvlink_lib_discover_and_get_remote_conn_info
     conn_info->bConnected = NV_FALSE;
 
     // Get the remote_end of the link
-    nvlink_core_discover_and_get_remote_end(end, &remote_end, flags);
+    nvlink_core_discover_and_get_remote_end(end, &remote_end, flags,
+                                            bForceDiscovery);
 
     if (remote_end)
     {

diff --git a/src/common/nvlink/kernel/nvlink/nvlink_helper.h b/src/common/nvlink/kernel/nvlink/nvlink_helper.h
@@ -174,7 +174,8 @@ void nvlink_core_correlate_conn_by_token(nvlink_link *srcLink, NvU64 writeToken,
  */
 void nvlink_core_discover_and_get_remote_end(nvlink_link  *end,
                                              nvlink_link **remote_end,
-                                             NvU32         flags);
+                                             NvU32         flags,
+                                             NvBool        bForceDiscovery);
 
 
 /************************************************************************************************/

diff --git a/src/common/nvswitch/kernel/lr10/lr10.c b/src/common/nvswitch/kernel/lr10/lr10.c
@@ -4071,7 +4071,9 @@ nvswitch_ctrl_get_nvlink_status_lr10
         }
         else
         {
-            nvlink_lib_discover_and_get_remote_conn_info(link, &conn_info, NVLINK_STATE_CHANGE_SYNC);
+            nvlink_lib_discover_and_get_remote_conn_info(link, &conn_info,
+                                                         NVLINK_STATE_CHANGE_SYNC,
+                                                         NV_FALSE);
         }
 
         // Set NVLINK per-link caps

diff --git a/src/common/sdk/nvidia/inc/ctrl/ctrl2080/ctrl2080nvlink.h b/src/common/sdk/nvidia/inc/ctrl/ctrl2080/ctrl2080nvlink.h
@@ -2971,8 +2971,91 @@ typedef struct NV2080_CTRL_NVLINK_POST_FAULT_UP_PARAMS {
     NvU32 linkId;
 } NV2080_CTRL_NVLINK_POST_FAULT_UP_PARAMS;
 
-#define NV2080_CTRL_CMD_NVLINK_POST_FAULT_UP (0x20803043U) /* finn: Evaluated from "(FINN_NV20_SUBDEVICE_0_NVLINK_INTERFACE_ID << 8) | NV2080_CTRL_NVLINK_POST_FAULT_UP_PARAMS_MESSAGE_ID" */
+#define NV2080_CTRL_CMD_NVLINK_POST_FAULT_UP     (0x20803043U) /* finn: Evaluated from "(FINN_NV20_SUBDEVICE_0_NVLINK_INTERFACE_ID << 8) | NV2080_CTRL_NVLINK_POST_FAULT_UP_PARAMS_MESSAGE_ID" */
+
+#define NV2080_CTRL_NVLINK_PORT_EVENT_COUNT_SIZE 64U
+
+/*
+* Structure to store port event information
+*
+*   portEventType
+*       Type of port even that occurred: NVLINK_PORT_EVENT_TYPE*
+*
+*   gpuId
+*       Gpu that port event occurred on
+*
+*   linkId
+*       Link id that port event occurred on
+*
+*   time
+*       Platform time (nsec) when event occurred
+*/
+typedef struct NV2080_CTRL_NVLINK_PORT_EVENT {
+    NvU32 portEventType;
+    NvU32 gpuId;
+    NvU32 linkId;
+    NV_DECLARE_ALIGNED(NvU64 time, 8);
+} NV2080_CTRL_NVLINK_PORT_EVENT;
+
+/*
+* NV2080_CTRL_CMD_NVLINK_GET_PORT_EVENTS
+*
+* This command returns the port up and port down events that have occurred
+*
+* Parameters:
+*
+*   portEventIndex [IN/OUT]
+*      On input: The index of the first port event at which to start reading out of the driver.
+*
+*      On output: The index of the first port event that wasn't reported through the 'port event' array
+*                 in this call to NV2080_CTRL_CMD_NVLINK_GET_PORT_EVENTS. 
+*
+*   nextPortEventIndex[OUT]
+*      The index that will be assigned to the next port event that occurs.
+*      Users of the GET_PORT_EVENTS control call may set 'portEventIndex' to this field on initialization
+*      to bypass port events that have already occurred without making multiple control calls.
+*
+*   portEventCount [OUT]
+*      Number of port events returned by the call. Currently, portEventCount is limited
+*      by NV2080_CTRL_NVLINK_PORT_EVENT_COUNT_SIZE. In order to query all the port events, a
+*      client needs to keep calling the control till portEventCount is zero.
+*
+*   bOverflow [OUT]
+*       True when the port event log is overflowed and no longer contains all the port
+*       events that have occurred, false otherwise.
+*
+*   portEvent [OUT]
+*      The port event entires.
+*/
+#define NV2080_CTRL_NVLINK_GET_PORT_EVENTS_PARAMS_MESSAGE_ID (0x44U)
+
+typedef struct NV2080_CTRL_NVLINK_GET_PORT_EVENTS_PARAMS {
+    NV_DECLARE_ALIGNED(NvU64 portEventIndex, 8);
+    NV_DECLARE_ALIGNED(NvU64 nextPortEventIndex, 8);
+    NvU32  portEventCount;
+    NvBool bOverflow;
+    NV_DECLARE_ALIGNED(NV2080_CTRL_NVLINK_PORT_EVENT portEvent[NV2080_CTRL_NVLINK_PORT_EVENT_COUNT_SIZE], 8);
+} NV2080_CTRL_NVLINK_GET_PORT_EVENTS_PARAMS;
+
+#define NV2080_CTRL_CMD_NVLINK_GET_PORT_EVENTS (0x20803044U) /* finn: Evaluated from "(FINN_NV20_SUBDEVICE_0_NVLINK_INTERFACE_ID << 8) | NV2080_CTRL_NVLINK_GET_PORT_EVENTS_PARAMS_MESSAGE_ID" */
+
+/*
+* NV2080_CTRL_CMD_NVLINK_CYCLE_LINK
+*
+* This command cycles a link by faulting it and then retraining the link
+*
+* Parameters:
+*
+*   linkId [IN]
+*      The link id of the link to be cycled
+*/
+#define NV2080_CTRL_NVLINK_CYCLE_LINK_PARAMS_MESSAGE_ID (0x45U)
+
+typedef struct NV2080_CTRL_NVLINK_CYCLE_LINK_PARAMS {
+    NvU32 linkId;
+} NV2080_CTRL_NVLINK_CYCLE_LINK_PARAMS;
 
+#define NV2080_CTRL_CMD_NVLINK_CYCLE_LINK (0x20803045U) /* finn: Evaluated from "(FINN_NV20_SUBDEVICE_0_NVLINK_INTERFACE_ID << 8) | NV2080_CTRL_NVLINK_CYCLE_LINK_PARAMS_MESSAGE_ID" */
 
 /*
  * NV2080_CTRL_CMD_NVLINK_IS_REDUCED_CONFIG
@@ -2982,13 +3065,13 @@ typedef struct NV2080_CTRL_NVLINK_POST_FAULT_UP_PARAMS {
  * [out] bReducedNvlinkConfig
  *     Link number which the sequence should be triggered
  */
-#define NV2080_CTRL_NVLINK_IS_REDUCED_CONFIG_PARAMS_MESSAGE_ID (0x44U)
+#define NV2080_CTRL_NVLINK_IS_REDUCED_CONFIG_PARAMS_MESSAGE_ID (0x46U)
 
 typedef struct NV2080_CTRL_NVLINK_IS_REDUCED_CONFIG_PARAMS {
     NvBool bReducedNvlinkConfig;
 } NV2080_CTRL_NVLINK_IS_REDUCED_CONFIG_PARAMS;
 
-#define NV2080_CTRL_CMD_NVLINK_IS_REDUCED_CONFIG (0x20803044U) /* finn: Evaluated from "(FINN_NV20_SUBDEVICE_0_NVLINK_INTERFACE_ID << 8) | NV2080_CTRL_NVLINK_IS_REDUCED_CONFIG_PARAMS_MESSAGE_ID" */
+#define NV2080_CTRL_CMD_NVLINK_IS_REDUCED_CONFIG (0x20803046U) /* finn: Evaluated from "(FINN_NV20_SUBDEVICE_0_NVLINK_INTERFACE_ID << 8) | NV2080_CTRL_NVLINK_IS_REDUCED_CONFIG_PARAMS_MESSAGE_ID" */
 
 
 /* _ctrl2080nvlink_h_ */

diff --git a/src/nvidia/generated/g_kernel_nvlink_nvoc.h b/src/nvidia/generated/g_kernel_nvlink_nvoc.h
@@ -316,6 +316,7 @@ struct KernelNvlink {
     NvBool PRIVATE_FIELD(bEnableSafeModeAtLoad);
     NvBool PRIVATE_FIELD(bEnableAli);
     NvBool PRIVATE_FIELD(bFloorSwept);
+    NvU32 PRIVATE_FIELD(numPortEvents);
     NvBool PRIVATE_FIELD(bLinkTrainingDebugSpew);
     NvBool PRIVATE_FIELD(bDisableL2Mode);
     NvU32 PRIVATE_FIELD(nvlinkLinkSpeed);