ich777/open-gpu-kernel-modules
1
0
Fork 0
mirror of https://github.com/NVIDIA/open-gpu-kernel-modules.git synced 2025-07-22 22:53:01 +02:00
Code Issues Projects Releases Wiki Activity

570.133.20

Browse Source
This commit is contained in:
Bernhard Stoeckner
2025-04-17 17:56:49 +02:00
parent c5e439fea4
commit e8113f665d
15 changed files with 47 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
README.md
View File

@@ -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 570.133.07.
version 570.133.20.
## 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
570.133.07 driver release. This can be achieved by installing
570.133.20 driver release. This can be achieved by installing
the NVIDIA GPU driver from the .run file using the `--no-kernel-modules`
option. E.g.,
@@ -185,7 +185,7 @@ table below).
For details on feature support and limitations, see the NVIDIA GPU driver
end user README here:
https://us.download.nvidia.com/XFree86/Linux-x86_64/570.133.07/README/kernel_open.html
https://us.download.nvidia.com/XFree86/Linux-x86_64/570.133.20/README/kernel_open.html
For vGPU support, please refer to the README.vgpu packaged in the vGPU Host
Package for more details.
@@ -943,11 +943,11 @@ Subsystem Device ID.
| NVIDIA B200 | 2901 10DE 1999 |
| NVIDIA B200 | 2901 10DE 199B |
| NVIDIA B200 | 2901 10DE 20DA |
| NVIDIA HGX GB200 | 2941 10DE 2046 |
| NVIDIA HGX GB200 | 2941 10DE 20CA |
| NVIDIA HGX GB200 | 2941 10DE 20D5 |
| NVIDIA HGX GB200 | 2941 10DE 21C9 |
| NVIDIA HGX GB200 | 2941 10DE 21CA |
| NVIDIA GB200 | 2941 10DE 2046 |
| NVIDIA GB200 | 2941 10DE 20CA |
| NVIDIA GB200 | 2941 10DE 20D5 |
| NVIDIA GB200 | 2941 10DE 21C9 |
| NVIDIA GB200 | 2941 10DE 21CA |
| NVIDIA GeForce RTX 5090 | 2B85 |
| NVIDIA GeForce RTX 5090 D | 2B87 |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 2BB1 1028 204B |

2
kernel-open/Kbuild
View File

@@ -86,7 +86,7 @@ EXTRA_CFLAGS += -I$(src)/common/inc
EXTRA_CFLAGS += -I$(src)
EXTRA_CFLAGS += -Wall $(DEFINES) $(INCLUDES) -Wno-cast-qual -Wno-format-extra-args
EXTRA_CFLAGS += -D__KERNEL__ -DMODULE -DNVRM
EXTRA_CFLAGS += -DNV_VERSION_STRING=\"570.133.07\"
EXTRA_CFLAGS += -DNV_VERSION_STRING=\"570.133.20\"
ifneq ($(SYSSRCHOST1X),)
EXTRA_CFLAGS += -I$(SYSSRCHOST1X)

6
kernel-open/common/inc/nv-linux.h
View File

@@ -345,8 +345,6 @@ extern int nv_pat_mode;
#define NV_PAGE_COUNT(page) \
((unsigned int)page_count(page))
#define NV_GET_PAGE_COUNT(page_ptr) \
(NV_PAGE_COUNT(NV_GET_PAGE_STRUCT(page_ptr->phys_addr)))
#define NV_GET_PAGE_FLAGS(page_ptr) \
(NV_GET_PAGE_STRUCT(page_ptr->phys_addr)->flags)
@@ -1161,10 +1159,6 @@ typedef struct nvidia_pte_s {
NvU64 phys_addr;
unsigned long virt_addr;
NvU64 dma_addr;
#ifdef CONFIG_XEN
unsigned int guest_pfn;
#endif
unsigned int page_count;
} nvidia_pte_t;
#if defined(CONFIG_DMA_SHARED_BUFFER)

2
kernel-open/nvidia-uvm/uvm_linux.h
View File

@@ -183,7 +183,7 @@ static inline const struct cpumask *uvm_cpumask_of_node(int node)
printk(fmt, ##__VA_ARGS__); \
} while (0)
#define NV_UVM_GFP_FLAGS (GFP_KERNEL)
#define NV_UVM_GFP_FLAGS (GFP_KERNEL | __GFP_NOMEMALLOC)
#if defined(NVCPU_X86)
/* Some old IA32 kernels don't have 64/64 division routines,

2
kernel-open/nvidia-uvm/uvm_va_block.c
View File

@@ -1569,7 +1569,7 @@ static NV_STATUS block_alloc_cpu_chunk(uvm_va_block_t *block,
if (status == NV_OK)
break;
if (flags & UVM_CPU_CHUNK_ALLOC_FLAGS_STRICT) {
if ((flags & UVM_CPU_CHUNK_ALLOC_FLAGS_STRICT) && (num_possible_nodes() > 1)) {
flags &= ~UVM_CPU_CHUNK_ALLOC_FLAGS_STRICT;
numa_fallback = true;
status = uvm_cpu_chunk_alloc(alloc_size, flags, NUMA_NO_NODE, chunk);

25
kernel-open/nvidia/nv-vm.c
View File

@@ -431,7 +431,6 @@ NV_STATUS nv_alloc_contig_pages(
page_ptr = at->page_table[i];
page_ptr->phys_addr = phys_addr;
page_ptr->page_count = NV_GET_PAGE_COUNT(page_ptr);
page_ptr->virt_addr = virt_addr;
page_ptr->dma_addr = nv_phys_to_dma(dev, page_ptr->phys_addr);
@@ -486,17 +485,6 @@ void nv_free_contig_pages(
{
page_ptr = at->page_table[i];
if (NV_GET_PAGE_COUNT(page_ptr) != page_ptr->page_count)
{
static int count = 0;
if (count++ < NV_MAX_RECURRING_WARNING_MESSAGES)
{
nv_printf(NV_DBG_ERRORS,
"NVRM: VM: %s: page count != initial page count (%u,%u)\n",
__FUNCTION__, NV_GET_PAGE_COUNT(page_ptr),
page_ptr->page_count);
}
}
NV_MAYBE_UNRESERVE_PAGE(page_ptr);
}
@@ -600,7 +588,6 @@ NV_STATUS nv_alloc_system_pages(
page_ptr = at->page_table[(i * os_pages_in_page) + sub_page_idx];
page_ptr->phys_addr = phys_addr;
page_ptr->page_count = NV_GET_PAGE_COUNT(page_ptr);
page_ptr->virt_addr = sub_page_virt_addr;
//
@@ -667,18 +654,6 @@ void nv_free_system_pages(
{
page_ptr = at->page_table[i];
if (NV_GET_PAGE_COUNT(page_ptr) != page_ptr->page_count)
{
static int count = 0;
if (count++ < NV_MAX_RECURRING_WARNING_MESSAGES)
{
nv_printf(NV_DBG_ERRORS,
"NVRM: VM: %s: page count != initial page count (%u,%u)\n",
__FUNCTION__, NV_GET_PAGE_COUNT(page_ptr),
page_ptr->page_count);
}
}
NV_MAYBE_UNRESERVE_PAGE(page_ptr);
}

16
src/common/inc/nvBldVer.h
View File

@@ -43,18 +43,18 @@
#endif
#if defined(NV_LINUX) || defined(NV_BSD) || defined(NV_SUNOS)
#define NV_BUILD_BRANCH_VERSION "rel/gpu_drv/r570/r572_77-376"
#define NV_BUILD_CHANGELIST_NUM (35688848)
#define NV_BUILD_BRANCH_VERSION "rel/gpu_drv/r570/r572_77-389"
#define NV_BUILD_CHANGELIST_NUM (35826456)
#define NV_BUILD_TYPE "Official"
#define NV_BUILD_NAME "rel/gpu_drv/r570/r572_77-376"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (35688848)
#define NV_BUILD_NAME "rel/gpu_drv/r570/r572_77-389"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (35826456)
#else /* Windows builds */
#define NV_BUILD_BRANCH_VERSION "r572_77-2"
#define NV_BUILD_CHANGELIST_NUM (35681611)
#define NV_BUILD_BRANCH_VERSION "r572_77-4"
#define NV_BUILD_CHANGELIST_NUM (35784913)
#define NV_BUILD_TYPE "Official"
#define NV_BUILD_NAME "572.80"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (35681611)
#define NV_BUILD_NAME "572.95"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (35784913)
#define NV_BUILD_BRANCH_BASE_VERSION R570
#endif
// End buildmeister python edited section

2
src/common/inc/nvUnixVersion.h
View File

@@ -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 "570.133.07"
#define NV_VERSION_STRING "570.133.20"
#else

3
src/nvidia/generated/g_chipset_nvoc.h
View File

@@ -157,6 +157,9 @@ struct PCIECONFIGSPACEBASE
#define CL_AER_ERROR_SOURCE (CL_AER_BEGIN + 0x34)
#define CL_AER_END (CL_AER_BEGIN + 0x34)
// Advanced Error Reporting Root Error Status ERR_COR Subclass Capable Mask
#define CL_AER_ROOT_ERROR_STATUS_ERR_COR_SUBCLASS_MASK (NVBIT32(7) | NVBIT32(8))
// PCI Express Device Capabilities 2
#define CL_PCIE_DEV_CAP_2_ATOMICS_SUPPORTED_BIT NVBIT(6)
#define CL_PCIE_DEV_CAP_2_ATOMIC_32BIT NVBIT(7)

8
src/nvidia/generated/g_gpu_acct_nvoc.h
View File

@@ -49,7 +49,6 @@ extern "C" {
#include "ctrl/ctrl0000/ctrl0000gpuacct.h"
#include "ctrl/ctrl0000/ctrl0000gpu.h" // NV0000_CTRL_GPU_MAX_ATTACHED_GPUS
#include "ctrl/ctrl2080/ctrl2080perf.h" // NV2080_CTRL_PERF_GET_GPUMON_PERFMON_UTIL_SAMPLES_V2_PARAMS
#include "rmapi/client.h"
typedef struct TMR_EVENT TMR_EVENT;
@@ -75,7 +74,6 @@ typedef struct
{
NvU32 procId; // Pid of the process.
NvU32 procType; // Type of the process.
struct RmClient *pClient; // Process' RmClient.
NvU32 gpuUtil; // Process's average GR engine utilization.
NvU64 sumUtil; // Running sum of process's GR engine utilization.
NvU32 fbUtil; // Process's average FB bandwidth utilization.
@@ -242,15 +240,15 @@ static inline NV_STATUS gpuacctClearAccountingData(struct GpuAccounting *arg1, N
#define gpuacctClearAccountingData(arg1, arg2, arg3) gpuacctClearAccountingData_IMPL(arg1, arg2, arg3)
#endif //__nvoc_gpu_acct_h_disabled
NV_STATUS gpuacctStartGpuAccounting_IMPL(struct GpuAccounting *arg1, NvU32 arg2, NvU32 arg3, NvU32 arg4, struct RmClient *arg5);
NV_STATUS gpuacctStartGpuAccounting_IMPL(struct GpuAccounting *arg1, NvU32 arg2, NvU32 arg3, NvU32 arg4);
#ifdef __nvoc_gpu_acct_h_disabled
static inline NV_STATUS gpuacctStartGpuAccounting(struct GpuAccounting *arg1, NvU32 arg2, NvU32 arg3, NvU32 arg4, struct RmClient *arg5) {
static inline NV_STATUS gpuacctStartGpuAccounting(struct GpuAccounting *arg1, NvU32 arg2, NvU32 arg3, NvU32 arg4) {
NV_ASSERT_FAILED_PRECOMP("GpuAccounting was disabled!");
return NV_ERR_NOT_SUPPORTED;
}
#else //__nvoc_gpu_acct_h_disabled
#define gpuacctStartGpuAccounting(arg1, arg2, arg3, arg4, arg5) gpuacctStartGpuAccounting_IMPL(arg1, arg2, arg3, arg4, arg5)
#define gpuacctStartGpuAccounting(arg1, arg2, arg3, arg4) gpuacctStartGpuAccounting_IMPL(arg1, arg2, arg3, arg4)
#endif //__nvoc_gpu_acct_h_disabled
NV_STATUS gpuacctStopGpuAccounting_IMPL(struct GpuAccounting *arg1, NvU32 arg2, NvU32 arg3, NvU32 arg4);

10
src/nvidia/generated/g_nv_name_released.h
View File

@@ -5409,11 +5409,11 @@ static const CHIPS_RELEASED sChipsReleased[] = {
{ 0x2901, 0x1999, 0x10de, "NVIDIA B200" },
{ 0x2901, 0x199b, 0x10de, "NVIDIA B200" },
{ 0x2901, 0x20da, 0x10de, "NVIDIA B200" },
{ 0x2941, 0x2046, 0x10de, "NVIDIA HGX GB200" },
{ 0x2941, 0x20ca, 0x10de, "NVIDIA HGX GB200" },
{ 0x2941, 0x20d5, 0x10de, "NVIDIA HGX GB200" },
{ 0x2941, 0x21c9, 0x10de, "NVIDIA HGX GB200" },
{ 0x2941, 0x21ca, 0x10de, "NVIDIA HGX GB200" },
{ 0x2941, 0x2046, 0x10de, "NVIDIA GB200" },
{ 0x2941, 0x20ca, 0x10de, "NVIDIA GB200" },
{ 0x2941, 0x20d5, 0x10de, "NVIDIA GB200" },
{ 0x2941, 0x21c9, 0x10de, "NVIDIA GB200" },
{ 0x2941, 0x21ca, 0x10de, "NVIDIA GB200" },
{ 0x2B85, 0x0000, 0x0000, "NVIDIA GeForce RTX 5090" },
{ 0x2B87, 0x0000, 0x0000, "NVIDIA GeForce RTX 5090 D" },
{ 0x2BB1, 0x204b, 0x1028, "NVIDIA RTX PRO 6000 Blackwell Workstation Edition" },

39
src/nvidia/src/kernel/diagnostics/gpu_acct.c
View File

@@ -38,7 +38,7 @@ static NvU64 gpuacctGetCurrTime(void);
static NV_STATUS gpuacctAddProcEntry(GPU_ACCT_PROC_DATA_STORE *, GPUACCT_PROC_ENTRY *, NvBool);
static NV_STATUS gpuacctRemoveProcEntry(GPU_ACCT_PROC_DATA_STORE *, GPUACCT_PROC_ENTRY *);
static NV_STATUS gpuacctLookupProcEntry(GPU_ACCT_PROC_DATA_STORE *, NvU32, GPUACCT_PROC_ENTRY **);
static NV_STATUS gpuacctAllocProcEntry(GPU_ACCT_PROC_DATA_STORE *, NvU32, NvU32, RmClient *, GPUACCT_PROC_ENTRY **);
static NV_STATUS gpuacctAllocProcEntry(GPU_ACCT_PROC_DATA_STORE *, NvU32, NvU32, GPUACCT_PROC_ENTRY **);
static NV_STATUS gpuacctFreeProcEntry(GPU_ACCT_PROC_DATA_STORE *, GPUACCT_PROC_ENTRY *);
static NV_STATUS gpuacctCleanupDataStore(GPU_ACCT_PROC_DATA_STORE *);
static NV_STATUS gpuacctDestroyDataStore(GPU_ACCT_PROC_DATA_STORE *);
@@ -218,7 +218,6 @@ void gpuacctDestruct_IMPL
* @param[in] pDS Pointer to data store where process entry is to be added.
* @param[in] pid PID of the process.
* @param[in] procType Type of the process.
* @param[in] pClient Process RmClient
* @param[out] ppEntry Pointer to process entry.
*
* @return NV_OK
@@ -233,7 +232,6 @@ gpuacctAllocProcEntry
GPU_ACCT_PROC_DATA_STORE *pDS,
NvU32 pid,
NvU32 procType,
RmClient *pClient,
GPUACCT_PROC_ENTRY **ppEntry
)
{
@@ -252,7 +250,6 @@ gpuacctAllocProcEntry
pEntry->procId = pid;
pEntry->procType = procType;
pEntry->pClient = pClient;
status = gpuacctAddProcEntry(pDS, pEntry, NV_TRUE);
if (status != NV_OK)
@@ -653,8 +650,7 @@ gpuacctStartGpuAccounting_IMPL
GpuAccounting *pGpuAcct,
NvU32 gpuInstance,
NvU32 pid,
NvU32 subPid,
RmClient *pClient
NvU32 subPid
)
{
OBJGPU *pGpu;
@@ -713,7 +709,7 @@ gpuacctStartGpuAccounting_IMPL
// Create entry for the incoming pid.
status = gpuacctAllocProcEntry(pDS, searchPid,
NV_GPUACCT_PROC_TYPE_CPU, pClient, &pEntry);
NV_GPUACCT_PROC_TYPE_CPU, &pEntry);
NV_ASSERT_OR_RETURN(status == NV_OK, status);
NV_ASSERT_OR_RETURN(pEntry != NULL, NV_ERR_NO_MEMORY);
@@ -866,7 +862,6 @@ gpuacctStopGpuAccounting_IMPL
}
// Move the entry to dead procs data store
pEntry->pClient = NULL;
status = gpuacctRemoveProcEntry(pLiveDS, pEntry);
if (status != NV_OK)
{
@@ -1201,25 +1196,6 @@ gpuacctGetProcAcctInfo_IMPL
return NV_OK;
}
static NV_STATUS _gpuAcctGetPidValue(GPUACCT_PROC_ENTRY *pEntry, NvU32 vmIndex, NvU32 *nsPid)
{
if ((vmIndex != NV_INVALID_VM_INDEX) ||
(pEntry->pClient == NULL) ||
(pEntry->pClient->pOsPidInfo == NULL))
{
*nsPid = pEntry->procId;
}
else
{
if (osFindNsPid(pEntry->pClient->pOsPidInfo, nsPid) != NV_OK)
{
return NV_ERR_OBJECT_NOT_FOUND;
}
}
return NV_OK;
}
/*!
* Gets all the pids for which accounting data is available.
*
@@ -1239,7 +1215,6 @@ gpuacctGetAcctPids_IMPL
{
GPUACCT_PROC_ENTRY *pEntry;
GPU_ACCT_PROC_LIST *pList;
NV_STATUS status;
OBJGPU *pGpu;
NvU32 count;
NvU32 vmPid;
@@ -1292,7 +1267,7 @@ gpuacctGetAcctPids_IMPL
if (vmIndex == NV_INVALID_VM_INDEX)
{
// Skip dead info if the request is for baremetal and from container.
// Skip dead info if the requester is not from root namespace.
if (osIsInitNs() != NV_TRUE)
{
goto addLiveProc;
@@ -1333,11 +1308,7 @@ addLiveProc:
pEntry = iter.pValue;
if (pEntry && pEntry->procType == NV_GPUACCT_PROC_TYPE_GPU)
{
status = _gpuAcctGetPidValue(pEntry, vmIndex, &pParams->pidTbl[count]);
if (status == NV_OK)
{
count++;
}
pParams->pidTbl[count++] = pEntry->procId;
}
}

2
src/nvidia/src/kernel/gpu/device.c
View File

@@ -431,7 +431,7 @@ _deviceInit
// If gpuacctStartGpuAccounting() fails, just assert and print error.
// gpuacctStartGpuAccounting() is not a major failure, we will continue with deviceInit() as normal.
if ((pRsClient->type == CLIENT_TYPE_USER) && (gpuacctStartGpuAccounting(pGpuAcct,
pGpu->gpuInstance, pClient->ProcID, pClient->SubProcessID, pClient) != NV_OK))
pGpu->gpuInstance, pClient->ProcID, pClient->SubProcessID) != NV_OK))
{
NV_ASSERT(0);
NV_PRINTF(LEVEL_ERROR,

17
src/nvidia/src/kernel/gpu/rc/kernel_rc.c
View File

@@ -448,8 +448,8 @@ krcCheckBusError_KERNEL
&clDevCtrlStatus) == NV_OK &&
clDevCtrlStatusFlags != 0)
{
NV_PRINTF(LEVEL_ERROR,
"PCI-E corelogic status has pending errors (CL_PCIE_DEV_CTRL_STATUS = %08X):\n",
NV_PRINTF(LEVEL_INFO,
"PCI-E corelogic: Pending errors in DEV_CTRL_STATUS = %08X\n",
clDevCtrlStatus);
clDevCtrlStatusFlags_Org = clDevCtrlStatusFlags;
@@ -457,7 +457,7 @@ krcCheckBusError_KERNEL
if (clDevCtrlStatusFlags &
NV2080_CTRL_BUS_INFO_PCIE_LINK_ERRORS_CORR_ERROR)
{
NV_PRINTF(LEVEL_ERROR, " _CORR_ERROR_DETECTED\n");
NV_PRINTF(LEVEL_INFO, "PCI-E corelogic: CORR_ERROR_DETECTED\n");
// not much interested in this one
clDevCtrlStatusFlags &=
~NV2080_CTRL_BUS_INFO_PCIE_LINK_ERRORS_CORR_ERROR;
@@ -465,26 +465,27 @@ krcCheckBusError_KERNEL
if (clDevCtrlStatusFlags &
NV2080_CTRL_BUS_INFO_PCIE_LINK_ERRORS_NON_FATAL_ERROR)
{
NV_PRINTF(LEVEL_ERROR, " _NON_FATAL_ERROR_DETECTED\n");
NV_PRINTF(LEVEL_INFO, "PCI-E corelogic: NON_FATAL_ERROR_DETECTED\n");
}
if (clDevCtrlStatusFlags &
NV2080_CTRL_BUS_INFO_PCIE_LINK_ERRORS_FATAL_ERROR)
{
NV_PRINTF(LEVEL_ERROR, " _FATAL_ERROR_DETECTED\n");
NV_PRINTF(LEVEL_ERROR, "PCI-E corelogic: FATAL_ERROR_DETECTED\n");
}
if (clDevCtrlStatusFlags &
NV2080_CTRL_BUS_INFO_PCIE_LINK_ERRORS_UNSUPP_REQUEST)
{
NV_PRINTF(LEVEL_ERROR, " _UNSUPP_REQUEST_DETECTED\n");
NV_PRINTF(LEVEL_INFO, "PCI-E corelogic: UNSUPP_REQUEST_DETECTED\n");
}
}
// Corelogic AER
if (pCl != NULL && clPcieReadAerCapability(pGpu, pCl, &clAer) == NV_OK &&
(clAer.UncorrErrStatusReg != 0 || clAer.RooErrStatus != 0))
(clAer.UncorrErrStatusReg != 0 ||
(clAer.RooErrStatus & ~CL_AER_ROOT_ERROR_STATUS_ERR_COR_SUBCLASS_MASK) != 0))
{
NV_PRINTF(LEVEL_ERROR,
"PCE-I Advanced Error Reporting Corelogic Info:\n");
"PCI-E Advanced Error Reporting Corelogic Info:\n");
NV_PRINTF(LEVEL_ERROR,
" Uncorr Error Status Register : %08X\n",
clAer.UncorrErrStatusReg);

2
version.mk
View File

@@ -1,4 +1,4 @@
NVIDIA_VERSION = 570.133.07
NVIDIA_VERSION = 570.133.20
# This file.
VERSION_MK_FILE := $(lastword $(MAKEFILE_LIST))