550.107.02

This commit is contained in:
Bernhard Stoeckner 2024-07-29 10:22:58 +02:00
parent caa2dd11a0
commit 2cca8b3fd5
No known key found for this signature in database
GPG Key ID: 7D23DC2750FAC2E1
66 changed files with 1016 additions and 447 deletions

View File

@ -1,222 +0,0 @@
# Changelog
## Release 550 Entries
### [550.100] 2024-07-09
### [550.90.07] 2024-06-04
### [550.78] 2024-04-25
### [550.76] 2024-04-17
### [550.67] 2024-03-19
### [550.54.15] 2024-03-18
### [550.54.14] 2024-02-23
#### Added
- Added vGPU Host and vGPU Guest support. For vGPU Host, please refer to the README.vgpu packaged in the vGPU Host Package for more details.
### [550.40.07] 2024-01-24
#### Fixed
- Set INSTALL_MOD_DIR only if it's not defined, [#570](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/570) by @keelung-yang
## Release 545 Entries
### [545.29.06] 2023-11-22
#### Fixed
- The brightness control of NVIDIA seems to be broken, [#573](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/573)
### [545.29.02] 2023-10-31
### [545.23.06] 2023-10-17
#### Fixed
- Fix always-false conditional, [#493](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/493) by @meme8383
#### Added
- Added beta-quality support for GeForce and Workstation GPUs. Please see the "Open Linux Kernel Modules" chapter in the NVIDIA GPU driver end user README for details.
## Release 535 Entries
### [535.129.03] 2023-10-31
### [535.113.01] 2023-09-21
#### Fixed
- Fixed building main against current centos stream 8 fails, [#550](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/550) by @airlied
### [535.104.05] 2023-08-22
### [535.98] 2023-08-08
### [535.86.10] 2023-07-31
### [535.86.05] 2023-07-18
### [535.54.03] 2023-06-14
### [535.43.02] 2023-05-30
#### Fixed
- Fixed console restore with traditional VGA consoles.
#### Added
- Added support for Run Time D3 (RTD3) on Ampere and later GPUs.
- Added support for G-Sync on desktop GPUs.
## Release 530 Entries
### [530.41.03] 2023-03-23
### [530.30.02] 2023-02-28
#### Changed
- GSP firmware is now distributed as `gsp_tu10x.bin` and `gsp_ga10x.bin` to better reflect the GPU architectures supported by each firmware file in this release.
- The .run installer will continue to install firmware to /lib/firmware/nvidia/<version> and the nvidia.ko kernel module will load the appropriate firmware for each GPU at runtime.
#### Fixed
- Add support for resizable BAR on Linux when NVreg_EnableResizableBar=1 module param is set. [#3](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/3) by @sjkelly
#### Added
- Support for power management features like Suspend, Hibernate and Resume.
## Release 525 Entries
### [525.147.05] 2023-10-31
#### Fixed
- Fix nvidia_p2p_get_pages(): Fix double-free in register-callback error path, [#557](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/557) by @BrendanCunningham
### [525.125.06] 2023-06-26
### [525.116.04] 2023-05-09
### [525.116.03] 2023-04-25
### [525.105.17] 2023-03-30
### [525.89.02] 2023-02-08
### [525.85.12] 2023-01-30
### [525.85.05] 2023-01-19
#### Fixed
- Fix build problems with Clang 15.0, [#377](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/377) by @ptr1337
### [525.78.01] 2023-01-05
### [525.60.13] 2022-12-05
### [525.60.11] 2022-11-28
#### Fixed
- Fixed nvenc compatibility with usermode clients [#104](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/104)
### [525.53] 2022-11-10
#### Changed
- GSP firmware is now distributed as multiple firmware files: this release has `gsp_tu10x.bin` and `gsp_ad10x.bin` replacing `gsp.bin` from previous releases.
- Each file is named after a GPU architecture and supports GPUs from one or more architectures. This allows GSP firmware to better leverage each architecture's capabilities.
- The .run installer will continue to install firmware to `/lib/firmware/nvidia/<version>` and the `nvidia.ko` kernel module will load the appropriate firmware for each GPU at runtime.
#### Fixed
- Add support for IBT (indirect branch tracking) on supported platforms, [#256](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/256) by @rnd-ash
- Return EINVAL when [failing to] allocating memory, [#280](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/280) by @YusufKhan-gamedev
- Fix various typos in nvidia/src/kernel, [#16](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/16) by @alexisgeoffrey
- Added support for rotation in X11, Quadro Sync, Stereo, and YUV 4:2:0 on Turing.
## Release 520 Entries
### [520.61.07] 2022-10-20
### [520.56.06] 2022-10-12
#### Added
- Introduce support for GeForce RTX 4090 GPUs.
### [520.61.05] 2022-10-10
#### Added
- Introduce support for NVIDIA H100 GPUs.
#### Fixed
- Fix/Improve Makefile, [#308](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/308/) by @izenynn
- Make nvLogBase2 more efficient, [#177](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/177/) by @DMaroo
- nv-pci: fixed always true expression, [#195](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/195/) by @ValZapod
## Release 515 Entries
### [515.76] 2022-09-20
#### Fixed
- Improved compatibility with new Linux kernel releases
- Fixed possible excessive GPU power draw on an idle X11 or Wayland desktop when driving high resolutions or refresh rates
### [515.65.07] 2022-10-19
### [515.65.01] 2022-08-02
#### Fixed
- Collection of minor fixes to issues, [#6](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/61) by @Joshua-Ashton
- Remove unnecessary use of acpi_bus_get_device().
### [515.57] 2022-06-28
#### Fixed
- Backtick is deprecated, [#273](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/273) by @arch-user-france1
### [515.48.07] 2022-05-31
#### Added
- List of compatible GPUs in README.md.
#### Fixed
- Fix various README capitalizations, [#8](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/8) by @27lx
- Automatically tag bug report issues, [#15](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/15) by @thebeanogamer
- Improve conftest.sh Script, [#37](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/37) by @Nitepone
- Update HTTP link to HTTPS, [#101](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/101) by @alcaparra
- moved array sanity check to before the array access, [#117](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/117) by @RealAstolfo
- Fixed some typos, [#122](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/122) by @FEDOyt
- Fixed capitalization, [#123](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/123) by @keroeslux
- Fix typos in NVDEC Engine Descriptor, [#126](https://github.com/NVIDIA/open-gpu-kernel-modules/pull/126) from @TrickyDmitriy
- Extranous apostrohpes in a makefile script [sic], [#14](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/14) by @kiroma
- HDMI no audio @ 4K above 60Hz, [#75](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/75) by @adolfotregosa
- dp_configcaps.cpp:405: array index sanity check in wrong place?, [#110](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/110) by @dcb314
- NVRM kgspInitRm_IMPL: missing NVDEC0 engine, cannot initialize GSP-RM, [#116](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/116) by @kfazz
- ERROR: modpost: "backlight_device_register" [...nvidia-modeset.ko] undefined, [#135](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/135) by @sndirsch
- aarch64 build fails, [#151](https://github.com/NVIDIA/open-gpu-kernel-modules/issues/151) by @frezbo
### [515.43.04] 2022-05-11
- Initial release.

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 550.100.
version 550.107.02.
## 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
550.100 driver release. This can be achieved by installing
550.107.02 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_64/550.100/README/kernel_open.html
https://us.download.nvidia.com/XFree86/Linux-x86_64/550.107.02/README/kernel_open.html
For vGPU support, please refer to the README.vgpu packaged in the vGPU Host
Package for more details.
@ -834,10 +834,12 @@ Subsystem Device ID.
| NVIDIA GeForce RTX 2050 | 25AD |
| NVIDIA RTX A1000 | 25B0 1028 1878 |
| NVIDIA RTX A1000 | 25B0 103C 1878 |
| NVIDIA RTX A1000 | 25B0 103C 8D96 |
| NVIDIA RTX A1000 | 25B0 10DE 1878 |
| NVIDIA RTX A1000 | 25B0 17AA 1878 |
| NVIDIA RTX A400 | 25B2 1028 1879 |
| NVIDIA RTX A400 | 25B2 103C 1879 |
| NVIDIA RTX A400 | 25B2 103C 8D95 |
| NVIDIA RTX A400 | 25B2 10DE 1879 |
| NVIDIA RTX A400 | 25B2 17AA 1879 |
| NVIDIA A16 | 25B6 10DE 14A9 |
@ -912,6 +914,7 @@ Subsystem Device ID.
| NVIDIA GeForce RTX 4060 Ti | 2805 |
| NVIDIA GeForce RTX 4060 | 2808 |
| NVIDIA GeForce RTX 4070 Laptop GPU | 2820 |
| NVIDIA GeForce RTX 3050 A Laptop GPU | 2822 |
| NVIDIA RTX 3000 Ada Generation Laptop GPU | 2838 |
| NVIDIA GeForce RTX 4070 Laptop GPU | 2860 |
| NVIDIA GeForce RTX 4060 | 2882 |

View File

@ -72,7 +72,7 @@ EXTRA_CFLAGS += -I$(src)/common/inc
EXTRA_CFLAGS += -I$(src)
EXTRA_CFLAGS += -Wall $(DEFINES) $(INCLUDES) -Wno-cast-qual -Wno-format-extra-args
EXTRA_CFLAGS += -D__KERNEL__ -DMODULE -DNVRM
EXTRA_CFLAGS += -DNV_VERSION_STRING=\"550.100\"
EXTRA_CFLAGS += -DNV_VERSION_STRING=\"550.107.02\"
ifneq ($(SYSSRCHOST1X),)
EXTRA_CFLAGS += -I$(SYSSRCHOST1X)

View File

@ -1045,7 +1045,7 @@ NV_STATUS NV_API_CALL nv_vgpu_get_bar_info(nvidia_stack_t *, nv_state_t *, con
NvU64 *, NvU64 *, NvU32 *, NvBool *, NvU8 *);
NV_STATUS NV_API_CALL nv_vgpu_get_hbm_info(nvidia_stack_t *, nv_state_t *, const NvU8 *, NvU64 *, NvU64 *);
NV_STATUS NV_API_CALL nv_vgpu_process_vf_info(nvidia_stack_t *, nv_state_t *, NvU8, NvU32, NvU8, NvU8, NvU8, NvBool, void *);
NV_STATUS NV_API_CALL nv_gpu_bind_event(nvidia_stack_t *);
NV_STATUS NV_API_CALL nv_gpu_bind_event(nvidia_stack_t *, NvU32, NvBool *);
NV_STATUS NV_API_CALL nv_gpu_unbind_event(nvidia_stack_t *, NvU32, NvBool *);
NV_STATUS NV_API_CALL nv_get_usermap_access_params(nv_state_t*, nv_usermap_access_params_t*);

View File

@ -218,6 +218,8 @@ extern NvU32 os_page_size;
extern NvU64 os_page_mask;
extern NvU8 os_page_shift;
extern NvBool os_cc_enabled;
extern NvBool os_cc_sev_snp_enabled;
extern NvBool os_cc_snp_vtom_enabled;
extern NvBool os_cc_tdx_enabled;
extern NvBool os_dma_buf_enabled;
extern NvBool os_imex_channel_is_supported;

View File

@ -5102,6 +5102,42 @@ compile_test() {
compile_check_conftest "$CODE" "NV_CC_PLATFORM_PRESENT" "" "functions"
;;
cc_attr_guest_sev_snp)
#
# Determine if 'CC_ATTR_GUEST_SEV_SNP' is present.
#
# Added by commit aa5a461171f9 ("x86/mm: Extend cc_attr to
# include AMD SEV-SNP") in v5.19.
#
CODE="
#if defined(NV_LINUX_CC_PLATFORM_H_PRESENT)
#include <linux/cc_platform.h>
#endif
enum cc_attr cc_attributes = CC_ATTR_GUEST_SEV_SNP;
"
compile_check_conftest "$CODE" "NV_CC_ATTR_SEV_SNP" "" "types"
;;
hv_get_isolation_type)
#
# Determine if 'hv_get_isolation_type()' is present.
# Added by commit faff44069ff5 ("x86/hyperv: Add Write/Read MSR
# registers via ghcb page") in v5.16.
#
CODE="
#if defined(NV_ASM_MSHYPERV_H_PRESENT)
#include <asm/mshyperv.h>
#endif
void conftest_hv_get_isolation_type(void) {
int i;
hv_get_isolation_type(i);
}"
compile_check_conftest "$CODE" "NV_HV_GET_ISOLATION_TYPE" "" "functions"
;;
drm_prime_pages_to_sg_has_drm_device_arg)
#
# Determine if drm_prime_pages_to_sg() has 'dev' argument.

View File

@ -97,5 +97,6 @@ NV_HEADER_PRESENCE_TESTS = \
linux/sync_file.h \
linux/cc_platform.h \
asm/cpufeature.h \
linux/mpi.h
linux/mpi.h \
asm/mshyperv.h

View File

@ -1689,7 +1689,7 @@ int nv_drm_get_crtc_crc32_v2_ioctl(struct drm_device *dev,
struct NvKmsKapiCrcs crc32;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -ENOENT;
return -EOPNOTSUPP;
}
crtc = nv_drm_crtc_find(dev, filep, params->crtc_id);
@ -1717,7 +1717,7 @@ int nv_drm_get_crtc_crc32_ioctl(struct drm_device *dev,
struct NvKmsKapiCrcs crc32;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -ENOENT;
return -EOPNOTSUPP;
}
crtc = nv_drm_crtc_find(dev, filep, params->crtc_id);

View File

@ -480,6 +480,22 @@ static int nv_drm_load(struct drm_device *dev, unsigned long flags)
return -ENODEV;
}
#if defined(NV_DRM_FBDEV_GENERIC_AVAILABLE)
/*
* If fbdev is enabled, take modeset ownership now before other DRM clients
* can take master (and thus NVKMS ownership).
*/
if (nv_drm_fbdev_module_param) {
if (!nvKms->grabOwnership(pDevice)) {
nvKms->freeDevice(pDevice);
NV_DRM_DEV_LOG_ERR(nv_dev, "Failed to grab NVKMS modeset ownership");
return -EBUSY;
}
nv_dev->hasFramebufferConsole = NV_TRUE;
}
#endif
mutex_lock(&nv_dev->lock);
/* Set NvKmsKapiDevice */
@ -590,6 +606,15 @@ static void __nv_drm_unload(struct drm_device *dev)
return;
}
/* Release modeset ownership if fbdev is enabled */
#if defined(NV_DRM_FBDEV_GENERIC_AVAILABLE)
if (nv_dev->hasFramebufferConsole) {
drm_atomic_helper_shutdown(dev);
nvKms->releaseOwnership(nv_dev->pDevice);
}
#endif
cancel_delayed_work_sync(&nv_dev->hotplug_event_work);
mutex_lock(&nv_dev->lock);
@ -834,13 +859,18 @@ static int nv_drm_get_dpy_id_for_connector_id_ioctl(struct drm_device *dev,
struct drm_file *filep)
{
struct drm_nvidia_get_dpy_id_for_connector_id_params *params = data;
struct drm_connector *connector;
struct nv_drm_connector *nv_connector;
int ret = 0;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -EOPNOTSUPP;
}
// Importantly, drm_connector_lookup (with filep) will only return the
// connector if we are master, a lessee with the connector, or not master at
// all. It will return NULL if we are a lessee with other connectors.
struct drm_connector *connector =
nv_drm_connector_lookup(dev, filep, params->connectorId);
struct nv_drm_connector *nv_connector;
int ret = 0;
connector = nv_drm_connector_lookup(dev, filep, params->connectorId);
if (!connector) {
return -EINVAL;
@ -873,6 +903,11 @@ static int nv_drm_get_connector_id_for_dpy_id_ioctl(struct drm_device *dev,
int ret = -EINVAL;
#if defined(NV_DRM_CONNECTOR_LIST_ITER_PRESENT)
struct drm_connector_list_iter conn_iter;
#endif
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -EOPNOTSUPP;
}
#if defined(NV_DRM_CONNECTOR_LIST_ITER_PRESENT)
nv_drm_connector_list_iter_begin(dev, &conn_iter);
#endif
@ -1085,6 +1120,10 @@ static int nv_drm_grant_permission_ioctl(struct drm_device *dev, void *data,
{
struct drm_nvidia_grant_permissions_params *params = data;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -EOPNOTSUPP;
}
if (params->type == NV_DRM_PERMISSIONS_TYPE_MODESET) {
return nv_drm_grant_modeset_permission(dev, params, filep);
} else if (params->type == NV_DRM_PERMISSIONS_TYPE_SUB_OWNER) {
@ -1250,6 +1289,10 @@ static int nv_drm_revoke_permission_ioctl(struct drm_device *dev, void *data,
{
struct drm_nvidia_revoke_permissions_params *params = data;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -EOPNOTSUPP;
}
if (params->type == NV_DRM_PERMISSIONS_TYPE_MODESET) {
if (!params->dpyId) {
return -EINVAL;
@ -1771,11 +1814,6 @@ void nv_drm_register_drm_device(const nv_gpu_info_t *gpu_info)
if (nv_drm_fbdev_module_param &&
drm_core_check_feature(dev, DRIVER_MODESET)) {
if (!nvKms->grabOwnership(nv_dev->pDevice)) {
NV_DRM_DEV_LOG_ERR(nv_dev, "Failed to grab NVKMS modeset ownership");
goto failed_grab_ownership;
}
if (bus_is_pci) {
struct pci_dev *pdev = to_pci_dev(device);
@ -1786,8 +1824,6 @@ void nv_drm_register_drm_device(const nv_gpu_info_t *gpu_info)
#endif
}
drm_fbdev_generic_setup(dev, 32);
nv_dev->hasFramebufferConsole = NV_TRUE;
}
#endif /* defined(NV_DRM_FBDEV_GENERIC_AVAILABLE) */
@ -1798,12 +1834,6 @@ void nv_drm_register_drm_device(const nv_gpu_info_t *gpu_info)
return; /* Success */
#if defined(NV_DRM_FBDEV_GENERIC_AVAILABLE)
failed_grab_ownership:
drm_dev_unregister(dev);
#endif
failed_drm_register:
nv_drm_dev_free(dev);
@ -1870,12 +1900,6 @@ void nv_drm_remove_devices(void)
struct nv_drm_device *next = dev_list->next;
struct drm_device *dev = dev_list->dev;
#if defined(NV_DRM_FBDEV_GENERIC_AVAILABLE)
if (dev_list->hasFramebufferConsole) {
drm_atomic_helper_shutdown(dev);
nvKms->releaseOwnership(dev_list->pDevice);
}
#endif
drm_dev_unregister(dev);
nv_drm_dev_free(dev);

View File

@ -465,10 +465,15 @@ int nv_drm_prime_fence_context_create_ioctl(struct drm_device *dev,
{
struct nv_drm_device *nv_dev = to_nv_device(dev);
struct drm_nvidia_prime_fence_context_create_params *p = data;
struct nv_drm_prime_fence_context *nv_prime_fence_context =
__nv_drm_prime_fence_context_new(nv_dev, p);
struct nv_drm_prime_fence_context *nv_prime_fence_context;
int err;
if (nv_dev->pDevice == NULL) {
return -EOPNOTSUPP;
}
nv_prime_fence_context = __nv_drm_prime_fence_context_new(nv_dev, p);
if (!nv_prime_fence_context) {
goto done;
}
@ -523,6 +528,11 @@ int nv_drm_gem_prime_fence_attach_ioctl(struct drm_device *dev,
struct nv_drm_fence_context *nv_fence_context;
nv_dma_fence_t *fence;
if (nv_dev->pDevice == NULL) {
ret = -EOPNOTSUPP;
goto done;
}
if (p->__pad != 0) {
NV_DRM_DEV_LOG_ERR(nv_dev, "Padding fields must be zeroed");
goto done;
@ -1312,6 +1322,10 @@ int nv_drm_semsurf_fence_ctx_create_ioctl(struct drm_device *dev,
struct nv_drm_semsurf_fence_ctx *ctx;
int err;
if (nv_dev->pDevice == NULL) {
return -EOPNOTSUPP;
}
if (p->__pad != 0) {
NV_DRM_DEV_LOG_ERR(nv_dev, "Padding fields must be zeroed");
return -EINVAL;
@ -1473,6 +1487,11 @@ int nv_drm_semsurf_fence_create_ioctl(struct drm_device *dev,
int ret = -EINVAL;
int fd;
if (nv_dev->pDevice == NULL) {
ret = -EOPNOTSUPP;
goto done;
}
if (p->__pad != 0) {
NV_DRM_DEV_LOG_ERR(nv_dev, "Padding fields must be zeroed");
goto done;
@ -1635,6 +1654,10 @@ int nv_drm_semsurf_fence_wait_ioctl(struct drm_device *dev,
unsigned long flags;
int ret = -EINVAL;
if (nv_dev->pDevice == NULL) {
return -EOPNOTSUPP;
}
if (p->pre_wait_value >= p->post_wait_value) {
NV_DRM_DEV_LOG_ERR(
nv_dev,
@ -1743,6 +1766,11 @@ int nv_drm_semsurf_fence_attach_ioctl(struct drm_device *dev,
nv_dma_fence_t *fence;
int ret = -EINVAL;
if (nv_dev->pDevice == NULL) {
ret = -EOPNOTSUPP;
goto done;
}
nv_gem = nv_drm_gem_object_lookup(nv_dev->dev, filep, p->handle);
if (!nv_gem) {

View File

@ -380,7 +380,7 @@ int nv_drm_gem_import_nvkms_memory_ioctl(struct drm_device *dev,
int ret;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
ret = -EINVAL;
ret = -EOPNOTSUPP;
goto failed;
}
@ -430,7 +430,7 @@ int nv_drm_gem_export_nvkms_memory_ioctl(struct drm_device *dev,
int ret = 0;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
ret = -EINVAL;
ret = -EOPNOTSUPP;
goto done;
}
@ -483,7 +483,7 @@ int nv_drm_gem_alloc_nvkms_memory_ioctl(struct drm_device *dev,
int ret = 0;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
ret = -EINVAL;
ret = -EOPNOTSUPP;
goto failed;
}

View File

@ -319,7 +319,7 @@ int nv_drm_gem_identify_object_ioctl(struct drm_device *dev,
struct nv_drm_gem_object *nv_gem = NULL;
if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
return -EINVAL;
return -EOPNOTSUPP;
}
nv_dma_buf = nv_drm_gem_object_dma_buf_lookup(dev, filep, p->handle);

View File

@ -1,5 +1,5 @@
/*******************************************************************************
Copyright (c) 2013-2023 NVIDIA Corporation
Copyright (c) 2013-2024 NVIDIA Corporation
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to
@ -423,7 +423,9 @@ static void uvm_get_unaddressable_range(NvU32 num_va_bits, NvU64 *first, NvU64 *
UVM_ASSERT(first);
UVM_ASSERT(outer);
if (uvm_platform_uses_canonical_form_address()) {
// Maxwell GPUs (num_va_bits == 40b) do not support canonical form address
// even when plugged into platforms using it.
if (uvm_platform_uses_canonical_form_address() && num_va_bits > 40) {
*first = 1ULL << (num_va_bits - 1);
*outer = (NvU64)((NvS64)(1ULL << 63) >> (64 - num_va_bits));
}

View File

@ -96,6 +96,10 @@
#include <linux/cc_platform.h>
#endif
#if defined(NV_ASM_MSHYPERV_H_PRESENT) && defined(NVCPU_X86_64)
#include <asm/mshyperv.h>
#endif
#if defined(NV_ASM_CPUFEATURE_H_PRESENT)
#include <asm/cpufeature.h>
#endif
@ -285,6 +289,17 @@ void nv_detect_conf_compute_platform(
#if defined(NV_CC_PLATFORM_PRESENT)
os_cc_enabled = cc_platform_has(CC_ATTR_GUEST_MEM_ENCRYPT);
#if defined(NV_CC_ATTR_SEV_SNP)
os_cc_sev_snp_enabled = cc_platform_has(CC_ATTR_GUEST_SEV_SNP);
#endif
#if defined(NV_HV_GET_ISOLATION_TYPE) && IS_ENABLED(CONFIG_HYPERV) && defined(NVCPU_X86_64)
if (hv_get_isolation_type() == HV_ISOLATION_TYPE_SNP)
{
os_cc_snp_vtom_enabled = NV_TRUE;
}
#endif
#if defined(X86_FEATURE_TDX_GUEST)
if (cpu_feature_enabled(X86_FEATURE_TDX_GUEST))
{
@ -293,8 +308,10 @@ void nv_detect_conf_compute_platform(
#endif
#else
os_cc_enabled = NV_FALSE;
os_cc_sev_snp_enabled = NV_FALSE;
os_cc_snp_vtom_enabled = NV_FALSE;
os_cc_tdx_enabled = NV_FALSE;
#endif
#endif //NV_CC_PLATFORM_PRESENT
}
static

View File

@ -160,6 +160,8 @@ NV_CONFTEST_FUNCTION_COMPILE_TESTS += full_name_hash
NV_CONFTEST_FUNCTION_COMPILE_TESTS += pci_enable_atomic_ops_to_root
NV_CONFTEST_FUNCTION_COMPILE_TESTS += vga_tryget
NV_CONFTEST_FUNCTION_COMPILE_TESTS += cc_platform_has
NV_CONFTEST_FUNCTION_COMPILE_TESTS += cc_attr_guest_sev_snp
NV_CONFTEST_FUNCTION_COMPILE_TESTS += hv_get_isolation_type
NV_CONFTEST_FUNCTION_COMPILE_TESTS += seq_read_iter
NV_CONFTEST_FUNCTION_COMPILE_TESTS += follow_pfn
NV_CONFTEST_FUNCTION_COMPILE_TESTS += drm_gem_object_get

View File

@ -52,6 +52,8 @@ NvU32 os_page_size = PAGE_SIZE;
NvU64 os_page_mask = NV_PAGE_MASK;
NvU8 os_page_shift = PAGE_SHIFT;
NvBool os_cc_enabled = 0;
NvBool os_cc_sev_snp_enabled = 0;
NvBool os_cc_snp_vtom_enabled = 0;
NvBool os_cc_tdx_enabled = 0;
#if defined(CONFIG_DMA_SHARED_BUFFER)

View File

@ -282,6 +282,7 @@ namespace DisplayPort
virtual void markDeviceForDeletion() = 0;
virtual bool getRawDscCaps(NvU8 *buffer, NvU32 bufferSize) = 0;
virtual bool setRawDscCaps(NvU8 *buffer, NvU32 bufferSize) = 0;
// This interface is still nascent. Please don't use it. Read size limit is 16 bytes.
virtual AuxBus::status getDpcdData(unsigned offset, NvU8 * buffer,

View File

@ -44,6 +44,7 @@ namespace DisplayPort
#define HDCP_BCAPS_DDC_EN_BIT 0x80
#define HDCP_BCAPS_DP_EN_BIT 0x01
#define HDCP_I2C_CLIENT_ADDR 0x74
#define DSC_CAPS_SIZE 16
struct GroupImpl;
struct ConnectorImpl;
@ -421,6 +422,7 @@ namespace DisplayPort
virtual void markDeviceForDeletion() {bisMarkedForDeletion = true;};
virtual bool isMarkedForDeletion() {return bisMarkedForDeletion;};
virtual bool getRawDscCaps(NvU8 *buffer, NvU32 bufferSize);
virtual bool setRawDscCaps(NvU8 *buffer, NvU32 bufferSize);
virtual AuxBus::status dscCrcControl(NvBool bEnable, gpuDscCrc *dataGpu, sinkDscCrc *dataSink);

View File

@ -472,6 +472,15 @@ bool DeviceImpl::getRawDscCaps(NvU8 *buffer, NvU32 bufferSize)
return true;
}
bool DeviceImpl::setRawDscCaps(NvU8 *buffer, NvU32 bufferSize)
{
if (bufferSize < sizeof(rawDscCaps))
return false;
dpMemCopy(&rawDscCaps, buffer, sizeof(rawDscCaps));
return parseDscCaps(&rawDscCaps[0], sizeof(rawDscCaps));
}
AuxBus::status DeviceImpl::transaction(Action action, Type type, int address,
NvU8 * buffer, unsigned sizeRequested,
unsigned * sizeCompleted,

View File

@ -36,25 +36,25 @@
// and then checked back in. You cannot make changes to these sections without
// corresponding changes to the buildmeister script
#ifndef NV_BUILD_BRANCH
#define NV_BUILD_BRANCH r550_00
#define NV_BUILD_BRANCH r552_86
#endif
#ifndef NV_PUBLIC_BRANCH
#define NV_PUBLIC_BRANCH r550_00
#define NV_PUBLIC_BRANCH r552_86
#endif
#if defined(NV_LINUX) || defined(NV_BSD) || defined(NV_SUNOS)
#define NV_BUILD_BRANCH_VERSION "rel/gpu_drv/r550/r550_00-326"
#define NV_BUILD_CHANGELIST_NUM (34471492)
#define NV_BUILD_BRANCH_VERSION "rel/gpu_drv/r550/r552_86-355"
#define NV_BUILD_CHANGELIST_NUM (34618165)
#define NV_BUILD_TYPE "Official"
#define NV_BUILD_NAME "rel/gpu_drv/r550/r550_00-326"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34471492)
#define NV_BUILD_NAME "rel/gpu_drv/r550/r552_86-355"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34618165)
#else /* Windows builds */
#define NV_BUILD_BRANCH_VERSION "r550_00-324"
#define NV_BUILD_CHANGELIST_NUM (34468048)
#define NV_BUILD_TYPE "Nightly"
#define NV_BUILD_NAME "r550_00-240627"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34454921)
#define NV_BUILD_BRANCH_VERSION "r552_86-1"
#define NV_BUILD_CHANGELIST_NUM (34615400)
#define NV_BUILD_TYPE "Official"
#define NV_BUILD_NAME "552.87"
#define NV_LAST_OFFICIAL_CHANGELIST_NUM (34615400)
#define NV_BUILD_BRANCH_BASE_VERSION R550
#endif
// End buildmeister python edited section

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 "550.100"
#define NV_VERSION_STRING "550.107.02"
#else

View File

@ -57,7 +57,9 @@
#define NV_PFALCON_FALCON_DMATRFCMD 0x00000118 /* RW-4R */
#define NV_PFALCON_FALCON_DMATRFCMD_FULL 0:0 /* R-XVF */
#define NV_PFALCON_FALCON_DMATRFCMD_FULL_TRUE 0x00000001 /* R---V */
#define NV_PFALCON_FALCON_DMATRFCMD_FULL_FALSE 0x00000000 /* R---V */
#define NV_PFALCON_FALCON_DMATRFCMD_IDLE 1:1 /* R-XVF */
#define NV_PFALCON_FALCON_DMATRFCMD_IDLE_TRUE 0x00000001 /* R---V */
#define NV_PFALCON_FALCON_DMATRFCMD_IDLE_FALSE 0x00000000 /* R---V */
#define NV_PFALCON_FALCON_DMATRFCMD_SEC 3:2 /* RWXVF */
#define NV_PFALCON_FALCON_DMATRFCMD_IMEM 4:4 /* RWXVF */

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2018-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2018-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -49,6 +49,7 @@
#include "soe/soeifsmbpbi.h"
#include "soe/soeifcore.h"
#include "soe/soeifchnmgmt.h"
#include "soe/soeiftnvl.h"
#include "soe/soeifcci.h"
#include "soe/soeifheartbeat.h"
@ -71,6 +72,7 @@ typedef struct
RM_SOE_BIF_CMD bif;
RM_SOE_CORE_CMD core;
RM_SOE_CHNMGMT_CMD chnmgmt;
RM_SOE_TNVL_CMD tnvl;
RM_SOE_CCI_CMD cci;
} cmd;
} RM_FLCN_CMD_SOE,
@ -126,8 +128,9 @@ typedef struct
#define RM_SOE_TASK_ID_CCI 0x0D
#define RM_SOE_TASK_ID_FSPMGMT 0x0E
#define RM_SOE_TASK_ID_HEARTBEAT 0x0F
#define RM_SOE_TASK_ID_TNVL 0x10
// Add new task ID here...
#define RM_SOE_TASK_ID__END 0x10
#define RM_SOE_TASK_ID__END 0x11
/*!
* Unit-identifiers:
@ -151,8 +154,9 @@ typedef struct
#define RM_SOE_UNIT_CHNMGMT (0x0D)
#define RM_SOE_UNIT_CCI (0x0E)
#define RM_SOE_UNIT_HEARTBEAT (0x0F)
#define RM_SOE_UNIT_TNVL (0x10)
// Add new unit ID here...
#define RM_SOE_UNIT_END (0x10)
#define RM_SOE_UNIT_END (0x11)
#endif // _RMSOECMDIF_H_

View File

@ -0,0 +1,76 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
#ifndef _SOEIFTNVL_H_
#define _SOEIFTNVL_H_
/*!
* @file soeiftnvl.h
* @brief SOE TNVL Command Queue
*
* The TNVL unit ID will be used for sending and recieving
* Command Messages between driver and TNVL unit of SOE
*/
/*!
* Commands offered by the SOE Tnvl Interface.
*/
enum
{
/*
* Issue register write command
*/
RM_SOE_TNVL_CMD_ISSUE_REGISTER_WRITE = 0x0,
/*
* Issue pre-lock sequence
*/
RM_SOE_TNVL_CMD_ISSUE_PRE_LOCK_SEQUENCE = 0x1,
};
/*!
* TNVL queue command payload
*/
typedef struct
{
NvU8 cmdType;
NvU32 offset;
NvU32 data;
} RM_SOE_TNVL_CMD_REGISTER_WRITE;
typedef struct
{
NvU8 cmdType;
} RM_SOE_TNVL_CMD_PRE_LOCK_SEQUENCE;
typedef union
{
NvU8 cmdType;
RM_SOE_TNVL_CMD_REGISTER_WRITE registerWrite;
RM_SOE_TNVL_CMD_PRE_LOCK_SEQUENCE preLockSequence;
} RM_SOE_TNVL_CMD;
#endif // _SOEIFTNVL_H_

View File

@ -325,9 +325,17 @@ cciInit
NvU32 pci_device_id
)
{
nvswitch_task_create(device, _nvswitch_cci_poll_callback,
NVSWITCH_INTERVAL_1SEC_IN_NS / NVSWITCH_CCI_POLLING_RATE_HZ,
0);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_task_create(device, _nvswitch_cci_poll_callback,
NVSWITCH_INTERVAL_1SEC_IN_NS / NVSWITCH_CCI_POLLING_RATE_HZ,
0);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping CCI background task when TNVL is enabled\n");
}
return NVL_SUCCESS;
}

View File

@ -295,6 +295,8 @@
_op(NvlStatus, nvswitch_tnvl_get_attestation_report, (nvswitch_device *device, NVSWITCH_GET_ATTESTATION_REPORT_PARAMS *params), _arch) \
_op(NvlStatus, nvswitch_tnvl_send_fsp_lock_config, (nvswitch_device *device), _arch) \
_op(NvlStatus, nvswitch_tnvl_get_status, (nvswitch_device *device, NVSWITCH_GET_TNVL_STATUS_PARAMS *params), _arch) \
_op(NvlStatus, nvswitch_send_tnvl_prelock_cmd, (nvswitch_device *device), _arch) \
_op(void, nvswitch_tnvl_disable_interrupts, (nvswitch_device *device), _arch) \
NVSWITCH_HAL_FUNCTION_LIST_FEATURE_0(_op, _arch) \
#define NVSWITCH_HAL_FUNCTION_LIST_LS10(_op, _arch) \

View File

@ -710,4 +710,5 @@ NvlStatus nvswitch_fsp_error_code_to_nvlstatus_map_lr10(nvswitch_device *device,
NvlStatus nvswitch_tnvl_get_attestation_certificate_chain_lr10(nvswitch_device *device, NVSWITCH_GET_ATTESTATION_CERTIFICATE_CHAIN_PARAMS *params);
NvlStatus nvswitch_tnvl_get_attestation_report_lr10(nvswitch_device *device, NVSWITCH_GET_ATTESTATION_REPORT_PARAMS *params);
NvlStatus nvswitch_tnvl_get_status_lr10(nvswitch_device *device, NVSWITCH_GET_TNVL_STATUS_PARAMS *params);
void nvswitch_tnvl_disable_interrupts_lr10(nvswitch_device *device);
#endif //_LR10_H_

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2020-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -188,7 +188,9 @@
#define SOE_VBIOS_VERSION_MASK 0xFF0000
#define SOE_VBIOS_REVLOCK_DISABLE_NPORT_FATAL_INTR 0x370000
#define SOE_VBIOS_REVLOCK_ISSUE_INGRESS_STOP 0x440000
#define SOE_VBIOS_REVLOCK_ISSUE_INGRESS_STOP 0x4C0000
#define SOE_VBIOS_REVLOCK_ISSUE_REGISTER_WRITE 0x580000
#define SOE_VBIOS_REVLOCK_TNVL_PRELOCK_COMMAND 0x600000
// LS10 Saved LED state
#define ACCESS_LINK_LED_STATE CPLD_MACHXO3_ACCESS_LINK_LED_CTL_NVL_CABLE_LED
@ -1058,6 +1060,9 @@ NvlStatus nvswitch_tnvl_get_attestation_certificate_chain_ls10(nvswitch_device *
NvlStatus nvswitch_tnvl_get_attestation_report_ls10(nvswitch_device *device, NVSWITCH_GET_ATTESTATION_REPORT_PARAMS *params);
NvlStatus nvswitch_tnvl_send_fsp_lock_config_ls10(nvswitch_device *device);
NvlStatus nvswitch_tnvl_get_status_ls10(nvswitch_device *device, NVSWITCH_GET_TNVL_STATUS_PARAMS *params);
void nvswitch_tnvl_reg_wr_32_ls10(nvswitch_device *device, NVSWITCH_ENGINE_ID eng_id, NvU32 eng_bcast, NvU32 eng_instance, NvU32 base_addr, NvU32 offset, NvU32 data);
NvlStatus nvswitch_send_tnvl_prelock_cmd_ls10(nvswitch_device *device);
void nvswitch_tnvl_disable_interrupts_ls10(nvswitch_device *device);
NvlStatus nvswitch_ctrl_get_soe_heartbeat_ls10(nvswitch_device *device, NVSWITCH_GET_SOE_HEARTBEAT_PARAMS *p);
NvlStatus nvswitch_cci_enable_iobist_ls10(nvswitch_device *device, NvU32 linkNumber, NvBool bEnable);

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2020-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -50,5 +50,6 @@ void nvswitch_heartbeat_soe_callback_ls10(nvswitch_device *device, RM_FLCN_
NvlStatus nvswitch_soe_set_nport_interrupts_ls10(nvswitch_device *device, NvU32 nport, NvBool bEnable);
void nvswitch_soe_disable_nport_fatal_interrupts_ls10(nvswitch_device *device, NvU32 nport,
NvU32 nportIntrEnable, NvU8 nportIntrType);
NvlStatus nvswitch_soe_issue_ingress_stop_ls10(nvswitch_device *device, NvU32 nport, NvBool bStop);
NvlStatus nvswitch_soe_reg_wr_32_ls10(nvswitch_device *device, NvU32 offset, NvU32 data);
#endif //_SOE_LS10_H_

View File

@ -272,8 +272,8 @@ const NvU32 soe_ucode_data_lr10_dbg[] = {
0xa6b0001d, 0x240cf409, 0x001da03e, 0x0049190f, 0x009ff711, 0x00f802f8, 0xb50294b6, 0x00f804b9,
0xb602af92, 0xb9bc0294, 0xf400f8f9, 0x82f9d430, 0x301590b4, 0xc1b027e1, 0x0ad1b00b, 0x94b6f4bd,
0x0c91b002, 0x900149fe, 0x9fa04499, 0x20079990, 0x0b99929f, 0x95b29fa0, 0xa0049992, 0x9297b29f,
0x9fa00499, 0x0005ecdf, 0x90ffbf00, 0x4efe1499, 0xa0a6b201, 0x34ee909f, 0xb4b20209, 0x84bde9a0,
0x14bd34bd, 0x001eef3e, 0x277e6ab2, 0x49bf001a, 0x4bfea2b2, 0x014cfe01, 0x9044bb90, 0x95f94bcc,
0x9fa00499, 0x0005ecdf, 0x90ffbf00, 0x4efe1499, 0xa0a6b201, 0x34ee909f, 0xb4b20209, 0x14bde9a0,
0x34bd84bd, 0x001eef3e, 0x277e6ab2, 0x49bf001a, 0x4bfea2b2, 0x014cfe01, 0x9044bb90, 0x95f94bcc,
0xb31100b4, 0x008e0209, 0x9e0309b3, 0x010db300, 0x499800a8, 0xb27cb201, 0xfe5bb22a, 0xdd90014d,
0x3295f938, 0x0be0b40c, 0xa53ed4bd, 0x5fbf001e, 0xf9a6e9bf, 0x34381bf4, 0xe89827b0, 0x987fbf01,
0xb03302e9, 0xb0b40a00, 0x90b9bc0c, 0x1bf4f9a6, 0x1444df1e, 0xf9180000, 0x0094330c, 0x90f1b206,
@ -2269,8 +2269,8 @@ const NvU32 soe_ucode_data_lr10_dbg[] = {
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x69e9060c, 0xe6ca2d91, 0xac20edf2, 0xeafeafcc, 0x294f2cc2, 0x883a9d68, 0x493e2990, 0xc8e27d59,
0x30867660, 0xbc4af25f, 0xbc09e1ed, 0xab87e0fc, 0x8fc5fac6, 0xe1f366be, 0x1ec159bf, 0x352ff984,
0x69e9060c, 0xe6ca2d91, 0xac20edf2, 0xeafeafcc, 0x1de66f4b, 0x98838b38, 0xce342fcf, 0x31422bca,
0x30867660, 0xbc4af25f, 0xbc09e1ed, 0xab87e0fc, 0x154ee848, 0x4d419617, 0xc10ab5e0, 0x5570cfeb,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,

View File

@ -272,8 +272,8 @@ const NvU32 soe_ucode_data_lr10_prd[] = {
0xa6b0001d, 0x240cf409, 0x001da03e, 0x0049190f, 0x009ff711, 0x00f802f8, 0xb50294b6, 0x00f804b9,
0xb602af92, 0xb9bc0294, 0xf400f8f9, 0x82f9d430, 0x301590b4, 0xc1b027e1, 0x0ad1b00b, 0x94b6f4bd,
0x0c91b002, 0x900149fe, 0x9fa04499, 0x20079990, 0x0b99929f, 0x95b29fa0, 0xa0049992, 0x9297b29f,
0x9fa00499, 0x0005ecdf, 0x90ffbf00, 0x4efe1499, 0xa0a6b201, 0x34ee909f, 0xb4b20209, 0x84bde9a0,
0x14bd34bd, 0x001eef3e, 0x277e6ab2, 0x49bf001a, 0x4bfea2b2, 0x014cfe01, 0x9044bb90, 0x95f94bcc,
0x9fa00499, 0x0005ecdf, 0x90ffbf00, 0x4efe1499, 0xa0a6b201, 0x34ee909f, 0xb4b20209, 0x14bde9a0,
0x34bd84bd, 0x001eef3e, 0x277e6ab2, 0x49bf001a, 0x4bfea2b2, 0x014cfe01, 0x9044bb90, 0x95f94bcc,
0xb31100b4, 0x008e0209, 0x9e0309b3, 0x010db300, 0x499800a8, 0xb27cb201, 0xfe5bb22a, 0xdd90014d,
0x3295f938, 0x0be0b40c, 0xa53ed4bd, 0x5fbf001e, 0xf9a6e9bf, 0x34381bf4, 0xe89827b0, 0x987fbf01,
0xb03302e9, 0xb0b40a00, 0x90b9bc0c, 0x1bf4f9a6, 0x1444df1e, 0xf9180000, 0x0094330c, 0x90f1b206,
@ -2269,8 +2269,8 @@ const NvU32 soe_ucode_data_lr10_prd[] = {
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x69e9060c, 0xe6ca2d91, 0xac20edf2, 0xeafeafcc, 0x294f2cc2, 0x883a9d68, 0x493e2990, 0xc8e27d59,
0x30867660, 0xbc4af25f, 0xbc09e1ed, 0xab87e0fc, 0x8fc5fac6, 0xe1f366be, 0x1ec159bf, 0x352ff984,
0x69e9060c, 0xe6ca2d91, 0xac20edf2, 0xeafeafcc, 0x1de66f4b, 0x98838b38, 0xce342fcf, 0x31422bca,
0x30867660, 0xbc4af25f, 0xbc09e1ed, 0xab87e0fc, 0x154ee848, 0x4d419617, 0xc10ab5e0, 0x5570cfeb,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,

View File

@ -212,8 +212,15 @@ _inforom_nvlink_start_correctable_error_recording
pNvlinkState->bCallbackPending = NV_FALSE;
nvswitch_task_create(device, &_nvswitch_nvlink_1hz_callback,
NVSWITCH_INTERVAL_1SEC_IN_NS, 0);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_task_create(device, &_nvswitch_nvlink_1hz_callback,
NVSWITCH_INTERVAL_1SEC_IN_NS, 0);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping NVLINK heartbeat task when TNVL is enabled\n");
}
}
NvlStatus

View File

@ -8186,6 +8186,24 @@ nvswitch_tnvl_get_status_lr10
return -NVL_ERR_NOT_SUPPORTED;
}
NvlStatus
nvswitch_send_tnvl_prelock_cmd_lr10
(
nvswitch_device *device
)
{
return -NVL_ERR_NOT_SUPPORTED;
}
void
nvswitch_tnvl_disable_interrupts_lr10
(
nvswitch_device *device
)
{
return;
}
//
// This function auto creates the lr10 HAL connectivity from the NVSWITCH_INIT_HAL
// macro in haldef_nvswitch.h

View File

@ -386,6 +386,13 @@ nvswitch_is_cci_supported_ls10
nvswitch_device *device
)
{
// Skip CCI on TNVL mode
if (nvswitch_is_tnvl_mode_enabled(device))
{
NVSWITCH_PRINT(device, INFO, "CCI is not supported on TNVL mode\n");
return NV_FALSE;
}
if (FLD_TEST_DRF(_SWITCH_REGKEY, _CCI_CONTROL, _ENABLE, _FALSE,
device->regkeys.cci_control))
{

View File

@ -5928,11 +5928,18 @@ nvswitch_create_deferred_link_state_check_task_ls10
pErrorReportParams->nvlipt_instance = nvlipt_instance;
pErrorReportParams->link = link;
status = nvswitch_task_create_args(device, (void*)pErrorReportParams,
&_nvswitch_deferred_link_state_check_ls10,
NVSWITCH_DEFERRED_LINK_STATE_CHECK_INTERVAL_NS,
NVSWITCH_TASK_TYPE_FLAGS_RUN_ONCE |
NVSWITCH_TASK_TYPE_FLAGS_VOID_PTR_ARGS);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
status = nvswitch_task_create_args(device, (void*)pErrorReportParams,
&_nvswitch_deferred_link_state_check_ls10,
NVSWITCH_DEFERRED_LINK_STATE_CHECK_INTERVAL_NS,
NVSWITCH_TASK_TYPE_FLAGS_RUN_ONCE |
NVSWITCH_TASK_TYPE_FLAGS_VOID_PTR_ARGS);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping Deferred link state background task when TNVL is enabled\n");
}
}
if (status == NVL_SUCCESS)
@ -6013,11 +6020,14 @@ _nvswitch_create_deferred_link_errors_task_ls10
pErrorReportParams->nvlipt_instance = nvlipt_instance;
pErrorReportParams->link = link;
status = nvswitch_task_create_args(device, (void*)pErrorReportParams,
&_nvswitch_deferred_link_errors_check_ls10,
NVSWITCH_DEFERRED_FAULT_UP_CHECK_INTERVAL_NS,
NVSWITCH_TASK_TYPE_FLAGS_RUN_ONCE |
NVSWITCH_TASK_TYPE_FLAGS_VOID_PTR_ARGS);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
status = nvswitch_task_create_args(device, (void*)pErrorReportParams,
&_nvswitch_deferred_link_errors_check_ls10,
NVSWITCH_DEFERRED_FAULT_UP_CHECK_INTERVAL_NS,
NVSWITCH_TASK_TYPE_FLAGS_RUN_ONCE |
NVSWITCH_TASK_TYPE_FLAGS_VOID_PTR_ARGS);
}
}
if (status == NVL_SUCCESS)
@ -7416,7 +7426,7 @@ nvswitch_lib_service_interrupts_ls10
// 3. Run leaf specific interrupt handler
//
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NVLW_NON_FATAL);
val = DRF_NUM(_CTRL, _CPU_INTR_NVLW_NON_FATAL, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NVLW_NON_FATAL, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, INFO, "%s: NVLW NON_FATAL interrupts pending = 0x%x\n",
@ -7438,7 +7448,7 @@ nvswitch_lib_service_interrupts_ls10
}
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NVLW_FATAL);
val = DRF_NUM(_CTRL, _CPU_INTR_NVLW_FATAL, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NVLW_FATAL, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, INFO, "%s: NVLW FATAL interrupts pending = 0x%x\n",
@ -7462,7 +7472,7 @@ nvswitch_lib_service_interrupts_ls10
}
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NVLW_CORRECTABLE);
val = DRF_NUM(_CTRL, _CPU_INTR_NVLW_CORRECTABLE, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NVLW_CORRECTABLE, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, ERROR, "%s: NVLW CORRECTABLE interrupts pending = 0x%x\n",
@ -7472,7 +7482,7 @@ nvswitch_lib_service_interrupts_ls10
// Check NPG
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NPG_FATAL);
val = DRF_NUM(_CTRL, _CPU_INTR_NPG_FATAL, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NPG_FATAL, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, INFO, "%s: NPG FATAL interrupts pending = 0x%x\n",
@ -7494,7 +7504,7 @@ nvswitch_lib_service_interrupts_ls10
}
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NPG_NON_FATAL);
val = DRF_NUM(_CTRL, _CPU_INTR_NPG_NON_FATAL, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NPG_NON_FATAL, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, INFO, "%s: NPG NON_FATAL interrupts pending = 0x%x\n",
@ -7516,7 +7526,7 @@ nvswitch_lib_service_interrupts_ls10
}
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NPG_CORRECTABLE);
val = DRF_NUM(_CTRL, _CPU_INTR_NPG_CORRECTABLE, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NPG_CORRECTABLE, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, ERROR, "%s: NPG CORRECTABLE interrupts pending = 0x%x\n",
@ -7526,7 +7536,7 @@ nvswitch_lib_service_interrupts_ls10
// Check NXBAR
val = NVSWITCH_ENG_RD32(device, GIN, , 0, _CTRL, _CPU_INTR_NXBAR_FATAL);
val = DRF_NUM(_CTRL, _CPU_INTR_NXBAR_FATAL, _MASK, val);
val = DRF_VAL(_CTRL, _CPU_INTR_NXBAR_FATAL, _MASK, val);
if (val != 0)
{
NVSWITCH_PRINT(device, INFO, "%s: NXBAR FATAL interrupts pending = 0x%x\n",
@ -7605,9 +7615,9 @@ nvswitch_lib_service_interrupts_ls10
NVSWITCH_PRINT(device, ERROR, "%s: Problem servicing SOE",
__FUNCTION__);
return_status = status;
}
}
}
}
}
}
// step 4 -- retrigger engine interrupts

View File

@ -2979,13 +2979,6 @@ nvswitch_is_soe_supported_ls10
NVSWITCH_PRINT(device, WARN, "SOE can not be disabled via regkey.\n");
}
if (nvswitch_is_tnvl_mode_locked(device))
{
NVSWITCH_PRINT(device, INFO,
"SOE is not supported when TNVL mode is locked\n");
return NV_FALSE;
}
return NV_TRUE;
}
@ -3033,13 +3026,6 @@ nvswitch_is_inforom_supported_ls10
return NV_FALSE;
}
if (nvswitch_is_tnvl_mode_enabled(device))
{
NVSWITCH_PRINT(device, INFO,
"INFOROM is not supported when TNVL mode is enabled\n");
return NV_FALSE;
}
if (!nvswitch_is_soe_supported(device))
{
NVSWITCH_PRINT(device, INFO,
@ -4421,7 +4407,14 @@ nvswitch_eng_wr_ls10
return;
}
nvswitch_reg_write_32(device, base_addr + offset, data);
if (nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_tnvl_reg_wr_32_ls10(device, eng_id, eng_bcast, eng_instance, base_addr, offset, data);
}
else
{
nvswitch_reg_write_32(device, base_addr + offset, data);
}
#if defined(DEVELOP) || defined(DEBUG) || defined(NV_MODS)
{

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2020-2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2020-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -559,6 +559,76 @@ nvswitch_soe_disable_nport_fatal_interrupts_ls10
}
}
/*
* @Brief : Perform register writes in SOE during TNVL
*
* @param[in] device
* @param[in] offset
* @param[in] data
*/
NvlStatus
nvswitch_soe_reg_wr_32_ls10
(
nvswitch_device *device,
NvU32 offset,
NvU32 data
)
{
FLCN *pFlcn;
NvU32 cmdSeqDesc = 0;
NV_STATUS status;
RM_FLCN_CMD_SOE cmd;
NVSWITCH_TIMEOUT timeout;
RM_SOE_TNVL_CMD_REGISTER_WRITE *pRegisterWrite;
NVSWITCH_GET_BIOS_INFO_PARAMS params = { 0 };
if (!nvswitch_is_soe_supported(device))
{
NVSWITCH_PRINT(device, INFO,
"%s: SOE is not supported\n",
__FUNCTION__);
return NVL_SUCCESS; // -NVL_ERR_NOT_SUPPORTED
}
status = device->hal.nvswitch_ctrl_get_bios_info(device, &params);
if ((status != NVL_SUCCESS) || ((params.version & SOE_VBIOS_VERSION_MASK) <
SOE_VBIOS_REVLOCK_ISSUE_REGISTER_WRITE))
{
nvswitch_reg_write_32(device, offset, data);
return NVL_SUCCESS;
}
pFlcn = device->pSoe->pFlcn;
nvswitch_os_memset(&cmd, 0, sizeof(cmd));
cmd.hdr.unitId = RM_SOE_UNIT_TNVL;
cmd.hdr.size = RM_SOE_CMD_SIZE(TNVL, REGISTER_WRITE);
pRegisterWrite = &cmd.cmd.tnvl.registerWrite;
pRegisterWrite->cmdType = RM_SOE_TNVL_CMD_ISSUE_REGISTER_WRITE;
pRegisterWrite->offset = offset;
pRegisterWrite->data = data;
nvswitch_timeout_create(NVSWITCH_INTERVAL_5MSEC_IN_NS, &timeout);
status = flcnQueueCmdPostBlocking(device, pFlcn,
(PRM_FLCN_CMD)&cmd,
NULL, // pMsg
NULL, // pPayload
SOE_RM_CMDQ_LOG_ID,
&cmdSeqDesc,
&timeout);
if (status != NV_OK)
{
NVSWITCH_PRINT(device, ERROR,
"%s: Failed to send REGISTER_WRITE command to SOE, offset = 0x%x, data = 0x%x\n",
__FUNCTION__, offset, data);
return -NVL_ERR_GENERIC;
}
return NVL_SUCCESS;
}
/*
* @Brief : Init sequence for SOE FSP RISCV image
*
@ -609,14 +679,21 @@ nvswitch_init_soe_ls10
}
// Register SOE callbacks
status = nvswitch_soe_register_event_callbacks(device);
if (status != NVL_SUCCESS)
if (!nvswitch_is_tnvl_mode_enabled(device))
{
NVSWITCH_PRINT_SXID(device, NVSWITCH_ERR_HW_SOE_COMMAND_QUEUE,
"Failed to register SOE events\n");
NVSWITCH_PRINT_SXID(device, NVSWITCH_ERR_HW_SOE_BOOTSTRAP,
"SOE init failed(2)\n");
return status;
status = nvswitch_soe_register_event_callbacks(device);
if (status != NVL_SUCCESS)
{
NVSWITCH_PRINT_SXID(device, NVSWITCH_ERR_HW_SOE_COMMAND_QUEUE,
"Failed to register SOE events\n");
NVSWITCH_PRINT_SXID(device, NVSWITCH_ERR_HW_SOE_BOOTSTRAP,
"SOE init failed(2)\n");
return status;
}
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping registering SOE callbacks since TNVL is enabled\n");
}
// Sanity the command and message queues as a final check
@ -1363,6 +1440,71 @@ _soeI2CAccess_LS10
return ret;
}
/*
* @Brief : Send TNVL Pre Lock command to SOE
*
* @param[in] device
*/
NvlStatus
nvswitch_send_tnvl_prelock_cmd_ls10
(
nvswitch_device *device
)
{
FLCN *pFlcn;
NvU32 cmdSeqDesc = 0;
NV_STATUS status;
RM_FLCN_CMD_SOE cmd;
NVSWITCH_TIMEOUT timeout;
RM_SOE_TNVL_CMD_PRE_LOCK_SEQUENCE *pTnvlPreLock;
NVSWITCH_GET_BIOS_INFO_PARAMS params = { 0 };
if (!nvswitch_is_soe_supported(device))
{
NVSWITCH_PRINT(device, INFO, "%s: SOE is not supported\n",
__FUNCTION__);
return -NVL_ERR_NOT_SUPPORTED;
}
status = device->hal.nvswitch_ctrl_get_bios_info(device, &params);
if ((status != NVL_SUCCESS) || ((params.version & SOE_VBIOS_VERSION_MASK) <
SOE_VBIOS_REVLOCK_TNVL_PRELOCK_COMMAND))
{
NVSWITCH_PRINT(device, INFO,
"%s: Skipping TNVL_CMD_PRE_LOCK_SEQUENCE command to SOE. Update firmware "
"from .%02X to .%02X\n",
__FUNCTION__, (NvU32)((params.version & SOE_VBIOS_VERSION_MASK) >> 16),
SOE_VBIOS_REVLOCK_TNVL_PRELOCK_COMMAND);
return -NVL_ERR_NOT_SUPPORTED;
}
pFlcn = device->pSoe->pFlcn;
nvswitch_os_memset(&cmd, 0, sizeof(cmd));
cmd.hdr.unitId = RM_SOE_UNIT_TNVL;
cmd.hdr.size = RM_SOE_CMD_SIZE(TNVL, PRE_LOCK_SEQUENCE);
pTnvlPreLock = &cmd.cmd.tnvl.preLockSequence;
pTnvlPreLock->cmdType = RM_SOE_TNVL_CMD_ISSUE_PRE_LOCK_SEQUENCE;
nvswitch_timeout_create(NVSWITCH_INTERVAL_5MSEC_IN_NS, &timeout);
status = flcnQueueCmdPostBlocking(device, pFlcn,
(PRM_FLCN_CMD)&cmd,
NULL, // pMsg
NULL, // pPayload
SOE_RM_CMDQ_LOG_ID,
&cmdSeqDesc,
&timeout);
if (status != NV_OK)
{
NVSWITCH_PRINT(device, ERROR, "%s: Failed to send PRE_LOCK_SEQUENCE command to SOE, status 0x%x\n",
__FUNCTION__, status);
return -NVL_ERR_GENERIC;
}
return NVL_SUCCESS;
}
/**
* @brief set hal function pointers for functions defined in LR10 (i.e. this file)
*

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2023-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -26,9 +26,14 @@
#include "common_nvswitch.h"
#include "haldef_nvswitch.h"
#include "ls10/ls10.h"
#include "ls10/soe_ls10.h"
#include "nvswitch/ls10/dev_nvlsaw_ip.h"
#include "nvswitch/ls10/dev_nvlsaw_ip_addendum.h"
#include "nvswitch/ls10/dev_ctrl_ip.h"
#include "nvswitch/ls10/dev_ctrl_ip_addendum.h"
#include "nvswitch/ls10/dev_cpr_ip.h"
#include "nvswitch/ls10/dev_npg_ip.h"
#include <stddef.h>
@ -947,6 +952,9 @@ nvswitch_detect_tnvl_mode_ls10
val = NVSWITCH_SAW_RD32_LS10(device, _NVLSAW, _TNVL_MODE);
if (FLD_TEST_DRF(_NVLSAW, _TNVL_MODE, _STATUS, _ENABLED, val))
{
NVSWITCH_PRINT(device, ERROR,
"%s: TNVL Mode Detected\n",
__FUNCTION__);
device->tnvl_mode = NVSWITCH_DEVICE_TNVL_MODE_ENABLED;
}
@ -1048,3 +1056,119 @@ nvswitch_tnvl_get_status_ls10
params->status = device->tnvl_mode;
return NVL_SUCCESS;
}
static NvBool
_nvswitch_reg_cpu_write_allow_list_ls10
(
nvswitch_device *device,
NVSWITCH_ENGINE_ID eng_id,
NvU32 offset
)
{
switch (eng_id)
{
case NVSWITCH_ENGINE_ID_SOE:
case NVSWITCH_ENGINE_ID_GIN:
case NVSWITCH_ENGINE_ID_FSP:
return NV_TRUE;
case NVSWITCH_ENGINE_ID_SAW:
{
if (offset == NV_NVLSAW_DRIVER_ATTACH_DETACH)
return NV_TRUE;
break;
}
case NVSWITCH_ENGINE_ID_NPG:
{
if ((offset == NV_NPG_INTR_RETRIGGER(0)) ||
(offset == NV_NPG_INTR_RETRIGGER(1)))
return NV_TRUE;
break;
}
case NVSWITCH_ENGINE_ID_CPR:
{
if ((offset == NV_CPR_SYS_INTR_RETRIGGER(0)) ||
(offset == NV_CPR_SYS_INTR_RETRIGGER(1)))
return NV_TRUE;
break;
}
default :
return NV_FALSE;
}
return NV_FALSE;
}
void
nvswitch_tnvl_reg_wr_32_ls10
(
nvswitch_device *device,
NVSWITCH_ENGINE_ID eng_id,
NvU32 eng_bcast,
NvU32 eng_instance,
NvU32 base_addr,
NvU32 offset,
NvU32 data
)
{
if (!nvswitch_is_tnvl_mode_enabled(device))
{
NVSWITCH_PRINT(device, ERROR,
"%s: TNVL mode is not enabled\n",
__FUNCTION__);
NVSWITCH_ASSERT(0);
return;
}
if (nvswitch_is_tnvl_mode_locked(device))
{
NVSWITCH_PRINT(device, ERROR,
"%s: TNVL mode is locked\n",
__FUNCTION__);
NVSWITCH_ASSERT(0);
return;
}
if (_nvswitch_reg_cpu_write_allow_list_ls10(device, eng_id, offset))
{
nvswitch_reg_write_32(device, base_addr + offset, data);
}
else
{
if (nvswitch_soe_reg_wr_32_ls10(device, base_addr + offset, data) != NVL_SUCCESS)
{
NVSWITCH_PRINT(device, ERROR,
"%s: SOE ENG_WR failed for 0x%x[%d] %s @0x%08x+0x%06x = 0x%08x\n",
__FUNCTION__,
eng_id, eng_instance,
(
(eng_bcast == NVSWITCH_GET_ENG_DESC_TYPE_UNICAST) ? "UC" :
(eng_bcast == NVSWITCH_GET_ENG_DESC_TYPE_BCAST) ? "BC" :
(eng_bcast == NVSWITCH_GET_ENG_DESC_TYPE_MULTICAST) ? "MC" :
"??"
),
base_addr, offset, data);
NVSWITCH_ASSERT(0);
}
}
}
void
nvswitch_tnvl_disable_interrupts_ls10
(
nvswitch_device *device
)
{
//
// In TNVL locked disable non-fatal NVLW, NPG, and legacy interrupt,
// disable additional non-fatals on those partitions.
//
NVSWITCH_ENG_WR32(device, GIN, , 0, _CTRL, _CPU_INTR_LEAF_EN_CLEAR(NV_CTRL_CPU_INTR_NVLW_NON_FATAL_IDX),
0xFFFF);
NVSWITCH_ENG_WR32(device, GIN, , 0, _CTRL, _CPU_INTR_LEAF_EN_CLEAR(NV_CTRL_CPU_INTR_NPG_NON_FATAL_IDX),
0xFFFF);
NVSWITCH_ENG_WR32(device, GIN, , 0, _CTRL, _CPU_INTR_LEAF_EN_CLEAR(NV_CTRL_CPU_INTR_UNITS_IDX),
0xFFFFFFFF);
}

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2017-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2017-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -1021,6 +1021,15 @@ _nvswitch_ctrl_get_tnvl_status
return device->hal.nvswitch_tnvl_get_status(device, params);
}
void
nvswitch_tnvl_disable_interrupts
(
nvswitch_device *device
)
{
device->hal.nvswitch_tnvl_disable_interrupts(device);
}
static NvlStatus
_nvswitch_construct_soe
(
@ -1860,9 +1869,16 @@ nvswitch_lib_initialize_device
(void)device->hal.nvswitch_read_oob_blacklist_state(device);
(void)device->hal.nvswitch_write_fabric_state(device);
nvswitch_task_create(device, &nvswitch_fabric_state_heartbeat,
NVSWITCH_HEARTBEAT_INTERVAL_NS,
NVSWITCH_TASK_TYPE_FLAGS_RUN_EVEN_IF_DEVICE_NOT_INITIALIZED);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_task_create(device, &nvswitch_fabric_state_heartbeat,
NVSWITCH_HEARTBEAT_INTERVAL_NS,
NVSWITCH_TASK_TYPE_FLAGS_RUN_EVEN_IF_DEVICE_NOT_INITIALIZED);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping Fabric state heartbeat background task when TNVL is enabled\n");
}
//
// Blacklisted devices return successfully in order to preserve the fabric state heartbeat
@ -1966,12 +1982,26 @@ nvswitch_lib_initialize_device
if (device->regkeys.latency_counter == NV_SWITCH_REGKEY_LATENCY_COUNTER_LOGGING_ENABLE)
{
nvswitch_task_create(device, &nvswitch_internal_latency_bin_log,
nvswitch_get_latency_sample_interval_msec(device) * NVSWITCH_INTERVAL_1MSEC_IN_NS * 9/10, 0);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_task_create(device, &nvswitch_internal_latency_bin_log,
nvswitch_get_latency_sample_interval_msec(device) * NVSWITCH_INTERVAL_1MSEC_IN_NS * 9/10, 0);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping Internal latency background task when TNVL is enabled\n");
}
}
nvswitch_task_create(device, &nvswitch_ecc_writeback_task,
(60 * NVSWITCH_INTERVAL_1SEC_IN_NS), 0);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_task_create(device, &nvswitch_ecc_writeback_task,
(60 * NVSWITCH_INTERVAL_1SEC_IN_NS), 0);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping ECC writeback background task when TNVL is enabled\n");
}
if (IS_RTLSIM(device) || IS_EMULATION(device) || IS_FMODEL(device))
{
@ -1981,8 +2011,15 @@ nvswitch_lib_initialize_device
}
else
{
nvswitch_task_create(device, &nvswitch_monitor_thermal_alert,
100*NVSWITCH_INTERVAL_1MSEC_IN_NS, 0);
if (!nvswitch_is_tnvl_mode_enabled(device))
{
nvswitch_task_create(device, &nvswitch_monitor_thermal_alert,
100*NVSWITCH_INTERVAL_1MSEC_IN_NS, 0);
}
else
{
NVSWITCH_PRINT(device, INFO, "Skipping Thermal alert background task when TNVL is enabled\n");
}
}
device->nvlink_device->initialized = 1;
@ -5968,6 +6005,15 @@ nvswitch_tnvl_send_fsp_lock_config
return device->hal.nvswitch_tnvl_send_fsp_lock_config(device);
}
NvlStatus
nvswitch_send_tnvl_prelock_cmd
(
nvswitch_device *device
)
{
return device->hal.nvswitch_send_tnvl_prelock_cmd(device);
}
static NvlStatus
_nvswitch_ctrl_set_device_tnvl_lock
(
@ -6001,8 +6047,18 @@ _nvswitch_ctrl_set_device_tnvl_lock
//
// Disable non-fatal and legacy interrupts
// Disable commands to SOE
//
nvswitch_tnvl_disable_interrupts(device);
//
//
// Send Pre-Lock sequence command to SOE
//
status = nvswitch_send_tnvl_prelock_cmd(device);
if (status != NVL_SUCCESS)
{
return status;
}
// Send lock-config command to FSP
status = nvswitch_tnvl_send_fsp_lock_config(device);

View File

@ -155,24 +155,24 @@ typedef struct NV0000_CTRL_VGPU_DELETE_DEVICE_PARAMS {
} NV0000_CTRL_VGPU_DELETE_DEVICE_PARAMS;
/*
* NV0000_CTRL_CMD_VGPU_VFIO_UNREGISTER_STATUS
* NV0000_CTRL_CMD_VGPU_VFIO_NOTIFY_RM_STATUS
*
* This command informs RM the status vgpu-vfio unregister for a GPU.
* This command informs RM the status of vgpu-vfio GPU operations such as probe and unregister.
*
* returnStatus [IN]
* This parameter provides the status vgpu-vfio unregister operation.
* This parameter provides the status of vgpu-vfio GPU operation.
*
* gpuPciId [IN]
* This parameter provides the gpu id of the GPU
*/
#define NV0000_CTRL_CMD_VGPU_VFIO_UNREGISTER_STATUS (0xc05) /* finn: Evaluated from "(FINN_NV01_ROOT_VGPU_INTERFACE_ID << 8) | NV0000_CTRL_VGPU_VFIO_UNREGISTER_STATUS_PARAMS_MESSAGE_ID" */
#define NV0000_CTRL_CMD_VGPU_VFIO_NOTIFY_RM_STATUS (0xc05) /* finn: Evaluated from "(FINN_NV01_ROOT_VGPU_INTERFACE_ID << 8) | NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS_MESSAGE_ID" */
#define NV0000_CTRL_VGPU_VFIO_UNREGISTER_STATUS_PARAMS_MESSAGE_ID (0x5U)
#define NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS_MESSAGE_ID (0x5U)
typedef struct NV0000_CTRL_VGPU_VFIO_UNREGISTER_STATUS_PARAMS {
typedef struct NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS {
NvU32 returnStatus;
NvU32 gpuId;
} NV0000_CTRL_VGPU_VFIO_UNREGISTER_STATUS_PARAMS;
} NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS;
/* _ctrl0000vgpu_h_ */

View File

@ -108,6 +108,8 @@
#define NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_NONE 0
#define NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_AMD_SEV 1
#define NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_INTEL_TDX 2
#define NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_AMD_SEV_SNP 3
#define NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_AMD_SNP_VTOM 4
#define NV_CONF_COMPUTE_SYSTEM_GPUS_CAPABILITY_NONE 0
#define NV_CONF_COMPUTE_SYSTEM_GPUS_CAPABILITY_APM 1

View File

@ -361,7 +361,8 @@ ARMCSALLOWLISTINFO armChipsetAllowListInfo[] =
{PCI_VENDOR_ID_MELLANOX, 0xA2D0, CS_MELLANOX_BLUEFIELD}, // Mellanox BlueField
{PCI_VENDOR_ID_MELLANOX, 0xA2D4, CS_MELLANOX_BLUEFIELD2},// Mellanox BlueField 2
{PCI_VENDOR_ID_MELLANOX, 0xA2D5, CS_MELLANOX_BLUEFIELD2},// Mellanox BlueField 2 Crypto disabled
{PCI_VENDOR_ID_MELLANOX, 0xA2DB, CS_MELLANOX_BLUEFIELD3},// Mellanox BlueField 3
{PCI_VENDOR_ID_MELLANOX, 0xA2DB, CS_MELLANOX_BLUEFIELD3},// Mellanox BlueField 3 Crypto disabled
{PCI_VENDOR_ID_MELLANOX, 0xA2DA, CS_MELLANOX_BLUEFIELD3},// Mellanox BlueField 3 Crypto enabled
{PCI_VENDOR_ID_AMAZON, 0x0200, CS_AMAZON_GRAVITRON2}, // Amazon Gravitron2
{PCI_VENDOR_ID_FUJITSU, 0x1952, CS_FUJITSU_A64FX}, // Fujitsu A64FX
{PCI_VENDOR_ID_CADENCE, 0xDC01, CS_PHYTIUM_S2500}, // Phytium S2500

View File

@ -1045,7 +1045,7 @@ NV_STATUS NV_API_CALL nv_vgpu_get_bar_info(nvidia_stack_t *, nv_state_t *, con
NvU64 *, NvU64 *, NvU32 *, NvBool *, NvU8 *);
NV_STATUS NV_API_CALL nv_vgpu_get_hbm_info(nvidia_stack_t *, nv_state_t *, const NvU8 *, NvU64 *, NvU64 *);
NV_STATUS NV_API_CALL nv_vgpu_process_vf_info(nvidia_stack_t *, nv_state_t *, NvU8, NvU32, NvU8, NvU8, NvU8, NvBool, void *);
NV_STATUS NV_API_CALL nv_gpu_bind_event(nvidia_stack_t *);
NV_STATUS NV_API_CALL nv_gpu_bind_event(nvidia_stack_t *, NvU32, NvBool *);
NV_STATUS NV_API_CALL nv_gpu_unbind_event(nvidia_stack_t *, NvU32, NvBool *);
NV_STATUS NV_API_CALL nv_get_usermap_access_params(nv_state_t*, nv_usermap_access_params_t*);

View File

@ -218,6 +218,8 @@ extern NvU32 os_page_size;
extern NvU64 os_page_mask;
extern NvU8 os_page_shift;
extern NvBool os_cc_enabled;
extern NvBool os_cc_sev_snp_enabled;
extern NvBool os_cc_snp_vtom_enabled;
extern NvBool os_cc_tdx_enabled;
extern NvBool os_dma_buf_enabled;
extern NvBool os_imex_channel_is_supported;

View File

@ -799,7 +799,9 @@ NV_STATUS NV_API_CALL nv_gpu_unbind_event
}
NV_STATUS NV_API_CALL nv_gpu_bind_event(
nvidia_stack_t *sp
nvidia_stack_t *sp,
NvU32 gpuId,
NvBool *isEventNotified
)
{
THREAD_STATE_NODE threadState;
@ -812,7 +814,7 @@ NV_STATUS NV_API_CALL nv_gpu_bind_event(
// LOCK: acquire API lock
if ((rmStatus = rmapiLockAcquire(API_LOCK_FLAGS_NONE, RM_LOCK_MODULES_HYPERVISOR)) == NV_OK)
{
CliAddSystemEvent(NV0000_NOTIFIERS_GPU_BIND_EVENT, 0, NULL);
CliAddSystemEvent(NV0000_NOTIFIERS_GPU_BIND_EVENT, gpuId, isEventNotified);
// UNLOCK: release API lock
rmapiLockRelease();

View File

@ -2675,6 +2675,8 @@ void osInitSystemStaticConfig(SYS_STATIC_CONFIG *pConfig)
pConfig->bIsNotebook = rm_is_system_notebook();
pConfig->osType = nv_get_os_type();
pConfig->bOsCCEnabled = os_cc_enabled;
pConfig->bOsCCSevSnpEnabled = os_cc_sev_snp_enabled;
pConfig->bOsCCSnpVtomEnabled = os_cc_snp_vtom_enabled;
pConfig->bOsCCTdxEnabled = os_cc_tdx_enabled;
}

View File

@ -1559,24 +1559,6 @@ failed:
return status;
}
static void
RmHandleNvpcfEvents(
nv_state_t *pNv
)
{
OBJGPU *pGpu = NV_GET_NV_PRIV_PGPU(pNv);
THREAD_STATE_NODE threadState;
if (RmUnixRmApiPrologue(pNv, &threadState, RM_LOCK_MODULES_ACPI) == NULL)
{
return;
}
gpuNotifySubDeviceEvent(pGpu, NV2080_NOTIFIERS_NVPCF_EVENTS, NULL, 0, 0, 0);
RmUnixRmApiEpilogue(pNv, &threadState);
}
/*
* ---------------------------------------------------------------------------
*
@ -4312,7 +4294,6 @@ void NV_API_CALL rm_power_source_change_event(
THREAD_STATE_NODE threadState;
void *fp;
nv_state_t *nv;
OBJGPU *pGpu = gpumgrGetGpu(0);
NV_STATUS rmStatus = NV_OK;
NV_ENTER_RM_RUNTIME(sp,fp);
@ -4321,6 +4302,7 @@ void NV_API_CALL rm_power_source_change_event(
// LOCK: acquire API lock
if ((rmStatus = rmapiLockAcquire(API_LOCK_FLAGS_NONE, RM_LOCK_MODULES_EVENT)) == NV_OK)
{
OBJGPU *pGpu = gpumgrGetGpu(0);
if (pGpu != NULL)
{
nv = NV_GET_NV_STATE(pGpu);
@ -5940,16 +5922,32 @@ void NV_API_CALL rm_acpi_nvpcf_notify(
nvidia_stack_t *sp
)
{
void *fp;
OBJGPU *pGpu = gpumgrGetGpu(0);
void *fp;
THREAD_STATE_NODE threadState;
NV_STATUS rmStatus = NV_OK;
NV_ENTER_RM_RUNTIME(sp,fp);
threadStateInit(&threadState, THREAD_STATE_FLAGS_NONE);
if (pGpu != NULL)
// LOCK: acquire API lock
if ((rmStatus = rmapiLockAcquire(API_LOCK_FLAGS_NONE,
RM_LOCK_MODULES_EVENT)) == NV_OK)
{
nv_state_t *nv = NV_GET_NV_STATE(pGpu);
RmHandleNvpcfEvents(nv);
OBJGPU *pGpu = gpumgrGetGpu(0);
if (pGpu != NULL)
{
nv_state_t *nv = NV_GET_NV_STATE(pGpu);
if ((rmStatus = os_ref_dynamic_power(nv, NV_DYNAMIC_PM_FINE)) ==
NV_OK)
{
gpuNotifySubDeviceEvent(pGpu, NV2080_NOTIFIERS_NVPCF_EVENTS,
NULL, 0, 0, 0);
}
os_unref_dynamic_power(nv, NV_DYNAMIC_PM_FINE);
}
rmapiLockRelease();
}
threadStateFree(&threadState, THREAD_STATE_FLAGS_NONE);
NV_EXIT_RM_RUNTIME(sp,fp);
}

View File

@ -1551,6 +1551,21 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
#endif
},
{ /* [90] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x4u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
/*pFunc=*/ (void (*)(void)) cliresCtrlCmdVgpuVfioNotifyRMStatus_IMPL,
#endif // NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x4u)
/*flags=*/ 0x4u,
/*accessRight=*/0x0u,
/*methodId=*/ 0xc05u,
/*paramSize=*/ sizeof(NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS),
/*pClassInfo=*/ &(__nvoc_class_def_RmClientResource.classInfo),
#if NV_PRINTF_STRINGS_ALLOWED
/*func=*/ "cliresCtrlCmdVgpuVfioNotifyRMStatus"
#endif
},
{ /* [91] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x811u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1565,7 +1580,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientGetAddrSpaceType"
#endif
},
{ /* [91] */
{ /* [92] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x811u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1580,7 +1595,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientGetHandleInfo"
#endif
},
{ /* [92] */
{ /* [93] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1595,7 +1610,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientGetAccessRights"
#endif
},
{ /* [93] */
{ /* [94] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1610,7 +1625,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientSetInheritedSharePolicy"
#endif
},
{ /* [94] */
{ /* [95] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1625,7 +1640,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientGetChildHandle"
#endif
},
{ /* [95] */
{ /* [96] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1640,7 +1655,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientShareObject"
#endif
},
{ /* [96] */
{ /* [97] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x811u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1655,7 +1670,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdObjectsAreDuplicates"
#endif
},
{ /* [97] */
{ /* [98] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x811u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1670,7 +1685,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdClientSubscribeToImexChannel"
#endif
},
{ /* [98] */
{ /* [99] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x10u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1685,7 +1700,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdOsUnixFlushUserCache"
#endif
},
{ /* [99] */
{ /* [100] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1700,7 +1715,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdOsUnixExportObjectToFd"
#endif
},
{ /* [100] */
{ /* [101] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1715,7 +1730,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdOsUnixImportObjectFromFd"
#endif
},
{ /* [101] */
{ /* [102] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x813u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1730,7 +1745,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdOsUnixGetExportObjectInfo"
#endif
},
{ /* [102] */
{ /* [103] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1745,7 +1760,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdOsUnixCreateExportObjectFd"
#endif
},
{ /* [103] */
{ /* [104] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1760,7 +1775,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
/*func=*/ "cliresCtrlCmdOsUnixExportObjectsToFd"
#endif
},
{ /* [104] */
{ /* [105] */
#if NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x11u)
/*pFunc=*/ (void (*)(void)) NULL,
#else
@ -1780,7 +1795,7 @@ static const struct NVOC_EXPORTED_METHOD_DEF __nvoc_exported_method_def_RmClient
const struct NVOC_EXPORT_INFO __nvoc_export_info_RmClientResource =
{
/*numEntries=*/ 105,
/*numEntries=*/ 106,
/*pExportEntries=*/ __nvoc_exported_method_def_RmClientResource
};
@ -2219,6 +2234,10 @@ static void __nvoc_init_funcTable_RmClientResource_1(RmClientResource *pThis) {
pThis->__cliresCtrlCmdVgpuSetVgpuVersion__ = &cliresCtrlCmdVgpuSetVgpuVersion_IMPL;
#endif
#if !NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x4u)
pThis->__cliresCtrlCmdVgpuVfioNotifyRMStatus__ = &cliresCtrlCmdVgpuVfioNotifyRMStatus_IMPL;
#endif
#if !NVOC_EXPORTED_METHOD_DISABLED_BY_FLAG(0x10u)
pThis->__cliresCtrlCmdSystemNVPCFGetPowerModeInfo__ = &cliresCtrlCmdSystemNVPCFGetPowerModeInfo_IMPL;
#endif

View File

@ -178,6 +178,7 @@ struct RmClientResource {
NV_STATUS (*__cliresCtrlCmdSyncGpuBoostGroupInfo__)(struct RmClientResource *, NV0000_SYNC_GPU_BOOST_GROUP_INFO_PARAMS *);
NV_STATUS (*__cliresCtrlCmdVgpuGetVgpuVersion__)(struct RmClientResource *, NV0000_CTRL_VGPU_GET_VGPU_VERSION_PARAMS *);
NV_STATUS (*__cliresCtrlCmdVgpuSetVgpuVersion__)(struct RmClientResource *, NV0000_CTRL_VGPU_SET_VGPU_VERSION_PARAMS *);
NV_STATUS (*__cliresCtrlCmdVgpuVfioNotifyRMStatus__)(struct RmClientResource *, NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS *);
NV_STATUS (*__cliresCtrlCmdSystemNVPCFGetPowerModeInfo__)(struct RmClientResource *, NV0000_CTRL_CMD_SYSTEM_NVPCF_GET_POWER_MODE_INFO_PARAMS *);
NV_STATUS (*__cliresCtrlCmdSystemSyncExternalFabricMgmt__)(struct RmClientResource *, NV0000_CTRL_CMD_SYSTEM_SYNC_EXTERNAL_FABRIC_MGMT_PARAMS *);
NV_STATUS (*__cliresCtrlCmdSystemPfmreqhndlrCtrl__)(struct RmClientResource *, NV0000_CTRL_SYSTEM_PFM_REQ_HNDLR_CTRL_PARAMS *);
@ -336,6 +337,7 @@ NV_STATUS __nvoc_objCreate_RmClientResource(RmClientResource**, Dynamic*, NvU32,
#define cliresCtrlCmdSyncGpuBoostGroupInfo(pRmCliRes, pParams) cliresCtrlCmdSyncGpuBoostGroupInfo_DISPATCH(pRmCliRes, pParams)
#define cliresCtrlCmdVgpuGetVgpuVersion(pRmCliRes, vgpuVersionInfo) cliresCtrlCmdVgpuGetVgpuVersion_DISPATCH(pRmCliRes, vgpuVersionInfo)
#define cliresCtrlCmdVgpuSetVgpuVersion(pRmCliRes, vgpuVersionInfo) cliresCtrlCmdVgpuSetVgpuVersion_DISPATCH(pRmCliRes, vgpuVersionInfo)
#define cliresCtrlCmdVgpuVfioNotifyRMStatus(pRmCliRes, pVgpuDeleteParams) cliresCtrlCmdVgpuVfioNotifyRMStatus_DISPATCH(pRmCliRes, pVgpuDeleteParams)
#define cliresCtrlCmdSystemNVPCFGetPowerModeInfo(pRmCliRes, pParams) cliresCtrlCmdSystemNVPCFGetPowerModeInfo_DISPATCH(pRmCliRes, pParams)
#define cliresCtrlCmdSystemSyncExternalFabricMgmt(pRmCliRes, pExtFabricMgmtParams) cliresCtrlCmdSystemSyncExternalFabricMgmt_DISPATCH(pRmCliRes, pExtFabricMgmtParams)
#define cliresCtrlCmdSystemPfmreqhndlrCtrl(pRmCliRes, pParams) cliresCtrlCmdSystemPfmreqhndlrCtrl_DISPATCH(pRmCliRes, pParams)
@ -959,6 +961,12 @@ static inline NV_STATUS cliresCtrlCmdVgpuSetVgpuVersion_DISPATCH(struct RmClient
return pRmCliRes->__cliresCtrlCmdVgpuSetVgpuVersion__(pRmCliRes, vgpuVersionInfo);
}
NV_STATUS cliresCtrlCmdVgpuVfioNotifyRMStatus_IMPL(struct RmClientResource *pRmCliRes, NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS *pVgpuDeleteParams);
static inline NV_STATUS cliresCtrlCmdVgpuVfioNotifyRMStatus_DISPATCH(struct RmClientResource *pRmCliRes, NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS *pVgpuDeleteParams) {
return pRmCliRes->__cliresCtrlCmdVgpuVfioNotifyRMStatus__(pRmCliRes, pVgpuDeleteParams);
}
NV_STATUS cliresCtrlCmdSystemNVPCFGetPowerModeInfo_IMPL(struct RmClientResource *pRmCliRes, NV0000_CTRL_CMD_SYSTEM_NVPCF_GET_POWER_MODE_INFO_PARAMS *pParams);
static inline NV_STATUS cliresCtrlCmdSystemNVPCFGetPowerModeInfo_DISPATCH(struct RmClientResource *pRmCliRes, NV0000_CTRL_CMD_SYSTEM_NVPCF_GET_POWER_MODE_INFO_PARAMS *pParams) {

View File

@ -1098,26 +1098,40 @@ static void __nvoc_init_funcTable_OBJGPU_1(OBJGPU *pThis) {
}
// Hal function -- gpuIsDevModeEnabledInHw
if (( ((chipHal_HalVarIdx >> 5) == 1UL) && ((1UL << (chipHal_HalVarIdx & 0x1f)) & 0x10000000UL) )) /* ChipHal: GH100 */
{
pThis->__gpuIsDevModeEnabledInHw__ = &gpuIsDevModeEnabledInHw_GH100;
}
// default
else
if (( ((rmVariantHal_HalVarIdx >> 5) == 0UL) && ((1UL << (rmVariantHal_HalVarIdx & 0x1f)) & 0x00000001UL) )) /* RmVariantHal: VF */
{
pThis->__gpuIsDevModeEnabledInHw__ = &gpuIsDevModeEnabledInHw_491d52;
}
// Hal function -- gpuIsProtectedPcieEnabledInHw
if (( ((chipHal_HalVarIdx >> 5) == 1UL) && ((1UL << (chipHal_HalVarIdx & 0x1f)) & 0x10000000UL) )) /* ChipHal: GH100 */
{
pThis->__gpuIsProtectedPcieEnabledInHw__ = &gpuIsProtectedPcieEnabledInHw_GH100;
}
// default
else
{
if (( ((chipHal_HalVarIdx >> 5) == 1UL) && ((1UL << (chipHal_HalVarIdx & 0x1f)) & 0x10000000UL) )) /* ChipHal: GH100 */
{
pThis->__gpuIsDevModeEnabledInHw__ = &gpuIsDevModeEnabledInHw_GH100;
}
// default
else
{
pThis->__gpuIsDevModeEnabledInHw__ = &gpuIsDevModeEnabledInHw_491d52;
}
}
// Hal function -- gpuIsProtectedPcieEnabledInHw
if (( ((rmVariantHal_HalVarIdx >> 5) == 0UL) && ((1UL << (rmVariantHal_HalVarIdx & 0x1f)) & 0x00000001UL) )) /* RmVariantHal: VF */
{
pThis->__gpuIsProtectedPcieEnabledInHw__ = &gpuIsProtectedPcieEnabledInHw_491d52;
}
else
{
if (( ((chipHal_HalVarIdx >> 5) == 1UL) && ((1UL << (chipHal_HalVarIdx & 0x1f)) & 0x10000000UL) )) /* ChipHal: GH100 */
{
pThis->__gpuIsProtectedPcieEnabledInHw__ = &gpuIsProtectedPcieEnabledInHw_GH100;
}
// default
else
{
pThis->__gpuIsProtectedPcieEnabledInHw__ = &gpuIsProtectedPcieEnabledInHw_491d52;
}
}
// Hal function -- gpuIsCtxBufAllocInPmaSupported
if (( ((chipHal_HalVarIdx >> 5) == 1UL) && ((1UL << (chipHal_HalVarIdx & 0x1f)) & 0x11f0fc00UL) )) /* ChipHal: GA100 | GA102 | GA103 | GA104 | GA106 | GA107 | AD102 | AD103 | AD104 | AD106 | AD107 | GH100 */

View File

@ -3145,22 +3145,22 @@ static inline NvBool gpuIsCCEnabledInHw_DISPATCH(struct OBJGPU *pGpu) {
return pGpu->__gpuIsCCEnabledInHw__(pGpu);
}
NvBool gpuIsDevModeEnabledInHw_GH100(struct OBJGPU *pGpu);
static inline NvBool gpuIsDevModeEnabledInHw_491d52(struct OBJGPU *pGpu) {
return ((NvBool)(0 != 0));
}
NvBool gpuIsDevModeEnabledInHw_GH100(struct OBJGPU *pGpu);
static inline NvBool gpuIsDevModeEnabledInHw_DISPATCH(struct OBJGPU *pGpu) {
return pGpu->__gpuIsDevModeEnabledInHw__(pGpu);
}
NvBool gpuIsProtectedPcieEnabledInHw_GH100(struct OBJGPU *pGpu);
static inline NvBool gpuIsProtectedPcieEnabledInHw_491d52(struct OBJGPU *pGpu) {
return ((NvBool)(0 != 0));
}
NvBool gpuIsProtectedPcieEnabledInHw_GH100(struct OBJGPU *pGpu);
static inline NvBool gpuIsProtectedPcieEnabledInHw_DISPATCH(struct OBJGPU *pGpu) {
return pGpu->__gpuIsProtectedPcieEnabledInHw__(pGpu);
}

View File

@ -1118,6 +1118,23 @@ static void __nvoc_init_funcTable_KernelGsp_1(KernelGsp *pThis, RmHalspecOwner *
}
}
// Hal function -- kgspPreserveVgpuPartitionLogging
if (( ((rmVariantHal_HalVarIdx >> 5) == 0UL) && ((1UL << (rmVariantHal_HalVarIdx & 0x1f)) & 0x00000001UL) )) /* RmVariantHal: VF */
{
pThis->__kgspPreserveVgpuPartitionLogging__ = &kgspPreserveVgpuPartitionLogging_395e98;
}
else
{
if (( ((chipHal_HalVarIdx >> 5) == 1UL) && ((1UL << (chipHal_HalVarIdx & 0x1f)) & 0x000007e0UL) )) /* ChipHal: TU102 | TU104 | TU106 | TU116 | TU117 | GA100 */
{
pThis->__kgspPreserveVgpuPartitionLogging__ = &kgspPreserveVgpuPartitionLogging_395e98;
}
else
{
pThis->__kgspPreserveVgpuPartitionLogging__ = &kgspPreserveVgpuPartitionLogging_IMPL;
}
}
// Hal function -- kgspFreeVgpuPartitionLogging
if (( ((rmVariantHal_HalVarIdx >> 5) == 0UL) && ((1UL << (rmVariantHal_HalVarIdx & 0x1f)) & 0x00000001UL) )) /* RmVariantHal: VF */
{

View File

@ -361,6 +361,7 @@ struct KernelGsp {
NvU64 (*__kgspGetMaxWprHeapSizeMB__)(struct OBJGPU *, struct KernelGsp *);
NvU32 (*__kgspGetFwHeapParamOsCarveoutSize__)(struct OBJGPU *, struct KernelGsp *);
NV_STATUS (*__kgspInitVgpuPartitionLogging__)(struct OBJGPU *, struct KernelGsp *, NvU32, NvU64, NvU64, NvU64, NvU64);
NV_STATUS (*__kgspPreserveVgpuPartitionLogging__)(struct OBJGPU *, struct KernelGsp *, NvU32);
NV_STATUS (*__kgspFreeVgpuPartitionLogging__)(struct OBJGPU *, struct KernelGsp *, NvU32);
const char *(*__kgspGetSignatureSectionNamePrefix__)(struct OBJGPU *, struct KernelGsp *);
NV_STATUS (*__kgspSetupGspFmcArgs__)(struct OBJGPU *, struct KernelGsp *, GSP_FIRMWARE *);
@ -580,6 +581,8 @@ NV_STATUS __nvoc_objCreate_KernelGsp(KernelGsp**, Dynamic*, NvU32);
#define kgspGetFwHeapParamOsCarveoutSize_HAL(pGpu, pKernelGsp) kgspGetFwHeapParamOsCarveoutSize_DISPATCH(pGpu, pKernelGsp)
#define kgspInitVgpuPartitionLogging(pGpu, pKernelGsp, gfid, initTaskLogBUffOffset, initTaskLogBUffSize, vgpuTaskLogBUffOffset, vgpuTaskLogBuffSize) kgspInitVgpuPartitionLogging_DISPATCH(pGpu, pKernelGsp, gfid, initTaskLogBUffOffset, initTaskLogBUffSize, vgpuTaskLogBUffOffset, vgpuTaskLogBuffSize)
#define kgspInitVgpuPartitionLogging_HAL(pGpu, pKernelGsp, gfid, initTaskLogBUffOffset, initTaskLogBUffSize, vgpuTaskLogBUffOffset, vgpuTaskLogBuffSize) kgspInitVgpuPartitionLogging_DISPATCH(pGpu, pKernelGsp, gfid, initTaskLogBUffOffset, initTaskLogBUffSize, vgpuTaskLogBUffOffset, vgpuTaskLogBuffSize)
#define kgspPreserveVgpuPartitionLogging(pGpu, pKernelGsp, gfid) kgspPreserveVgpuPartitionLogging_DISPATCH(pGpu, pKernelGsp, gfid)
#define kgspPreserveVgpuPartitionLogging_HAL(pGpu, pKernelGsp, gfid) kgspPreserveVgpuPartitionLogging_DISPATCH(pGpu, pKernelGsp, gfid)
#define kgspFreeVgpuPartitionLogging(pGpu, pKernelGsp, gfid) kgspFreeVgpuPartitionLogging_DISPATCH(pGpu, pKernelGsp, gfid)
#define kgspFreeVgpuPartitionLogging_HAL(pGpu, pKernelGsp, gfid) kgspFreeVgpuPartitionLogging_DISPATCH(pGpu, pKernelGsp, gfid)
#define kgspGetSignatureSectionNamePrefix(pGpu, pKernelGsp) kgspGetSignatureSectionNamePrefix_DISPATCH(pGpu, pKernelGsp)
@ -1177,6 +1180,16 @@ static inline NV_STATUS kgspInitVgpuPartitionLogging_DISPATCH(struct OBJGPU *pGp
return pKernelGsp->__kgspInitVgpuPartitionLogging__(pGpu, pKernelGsp, gfid, initTaskLogBUffOffset, initTaskLogBUffSize, vgpuTaskLogBUffOffset, vgpuTaskLogBuffSize);
}
static inline NV_STATUS kgspPreserveVgpuPartitionLogging_395e98(struct OBJGPU *pGpu, struct KernelGsp *pKernelGsp, NvU32 gfid) {
return NV_ERR_NOT_SUPPORTED;
}
NV_STATUS kgspPreserveVgpuPartitionLogging_IMPL(struct OBJGPU *pGpu, struct KernelGsp *pKernelGsp, NvU32 gfid);
static inline NV_STATUS kgspPreserveVgpuPartitionLogging_DISPATCH(struct OBJGPU *pGpu, struct KernelGsp *pKernelGsp, NvU32 gfid) {
return pKernelGsp->__kgspPreserveVgpuPartitionLogging__(pGpu, pKernelGsp, gfid);
}
static inline NV_STATUS kgspFreeVgpuPartitionLogging_395e98(struct OBJGPU *pGpu, struct KernelGsp *pKernelGsp, NvU32 gfid) {
return NV_ERR_NOT_SUPPORTED;
}

View File

@ -125,7 +125,6 @@ typedef struct KERNEL_HOST_VGPU_DEVICE
NvU32 chidOffset[RM_ENGINE_TYPE_LAST];
NvU32 channelCount[RM_ENGINE_TYPE_LAST]; /*Number of channels available to the VF*/
NvU8 vgpuUuid[RM_SHA1_GID_SIZE];
void *pVgpuVfioRef;
struct REQUEST_VGPU_INFO_NODE *pRequestVgpuInfoNode;
struct PhysMemSubAlloc *pPhysMemSubAlloc;
struct HOST_VGPU_DEVICE *pHostVgpuDevice;

View File

@ -119,7 +119,6 @@ typedef struct _def_client_vgpu_ns_intr
NvU64 guestDomainId; // guest ID that we need to use to inject interrupt
NvU64 guestMSIAddr; // MSI address allocated by guest OS
NvU32 guestMSIData; // MSI data value set by guest OS
void *pVgpuVfioRef; // Reference to vgpu device in nvidia-vgpu-vfio module
void *pEventDpc; // DPC event to pass the interrupt
} VGPU_NS_INTR;

View File

@ -981,10 +981,12 @@ static const CHIPS_RELEASED sChipsReleased[] = {
{ 0x25AD, 0x0000, 0x0000, "NVIDIA GeForce RTX 2050" },
{ 0x25B0, 0x1878, 0x1028, "NVIDIA RTX A1000" },
{ 0x25B0, 0x1878, 0x103c, "NVIDIA RTX A1000" },
{ 0x25B0, 0x8d96, 0x103c, "NVIDIA RTX A1000" },
{ 0x25B0, 0x1878, 0x10de, "NVIDIA RTX A1000" },
{ 0x25B0, 0x1878, 0x17aa, "NVIDIA RTX A1000" },
{ 0x25B2, 0x1879, 0x1028, "NVIDIA RTX A400" },
{ 0x25B2, 0x1879, 0x103c, "NVIDIA RTX A400" },
{ 0x25B2, 0x8d95, 0x103c, "NVIDIA RTX A400" },
{ 0x25B2, 0x1879, 0x10de, "NVIDIA RTX A400" },
{ 0x25B2, 0x1879, 0x17aa, "NVIDIA RTX A400" },
{ 0x25B6, 0x14a9, 0x10de, "NVIDIA A16" },
@ -1059,6 +1061,7 @@ static const CHIPS_RELEASED sChipsReleased[] = {
{ 0x2805, 0x0000, 0x0000, "NVIDIA GeForce RTX 4060 Ti" },
{ 0x2808, 0x0000, 0x0000, "NVIDIA GeForce RTX 4060" },
{ 0x2820, 0x0000, 0x0000, "NVIDIA GeForce RTX 4070 Laptop GPU" },
{ 0x2822, 0x0000, 0x0000, "NVIDIA GeForce RTX 3050 A Laptop GPU" },
{ 0x2838, 0x0000, 0x0000, "NVIDIA RTX 3000 Ada Generation Laptop GPU" },
{ 0x2860, 0x0000, 0x0000, "NVIDIA GeForce RTX 4070 Laptop GPU" },
{ 0x2882, 0x0000, 0x0000, "NVIDIA GeForce RTX 4060" },

View File

@ -308,6 +308,12 @@ typedef struct SYS_STATIC_CONFIG
/*! Indicates confidentail compute OS support is enabled or not */
NvBool bOsCCEnabled;
/*! Indicates SEV-SNP confidential compute OS support is enabled or not */
NvBool bOsCCSevSnpEnabled;
/*! Indicates SEV-SNP vTOM confidential compute OS support is enabled or not */
NvBool bOsCCSnpVtomEnabled;
/*! Indicates Intel TDX confidentail compute OS support is enabled or not */
NvBool bOsCCTdxEnabled;
} SYS_STATIC_CONFIG;

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2021-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2021-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -63,4 +63,12 @@
*/
#define FLCN_RESET_PROPAGATION_DELAY_COUNT 10
/*!
* Used by FALCON_DMATRFCMD polling functions to wait for _FULL==FALSE or _IDLE==TRUE
*/
typedef enum {
FLCN_DMA_POLL_QUEUE_NOT_FULL = 0,
FLCN_DMA_POLL_ENGINE_IDLE = 1
} FlcnDmaPollMode;
#endif // FALCON_COMMON_H

View File

@ -80,7 +80,15 @@ confComputeApiCtrlCmdSystemGetCapabilities_IMPL
if ((sysGetStaticConfig(pSys))->bOsCCEnabled)
{
pParams->cpuCapability = NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_AMD_SEV;
if ((sysGetStaticConfig(pSys))->bOsCCTdxEnabled)
if ((sysGetStaticConfig(pSys))->bOsCCSevSnpEnabled)
{
pParams->cpuCapability = NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_AMD_SEV_SNP;
}
else if ((sysGetStaticConfig(pSys))->bOsCCSnpVtomEnabled)
{
pParams->cpuCapability = NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_AMD_SNP_VTOM;
}
else if ((sysGetStaticConfig(pSys))->bOsCCTdxEnabled)
{
pParams->cpuCapability = NV_CONF_COMPUTE_SYSTEM_CPU_CAPABILITY_INTEL_TDX;
}

View File

@ -1,5 +1,5 @@
/*
* SPDX-FileCopyrightText: Copyright (c) 2021-2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-FileCopyrightText: Copyright (c) 2021-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
@ -35,6 +35,67 @@
#include "published/ampere/ga102/dev_falcon_second_pri.h"
#include "published/ampere/ga102/dev_fbif_v4.h"
static GpuWaitConditionFunc s_dmaPollCondFunc;
typedef struct {
KernelFalcon *pKernelFlcn;
NvU32 pollMask;
NvU32 pollValue;
} DmaPollCondData;
static NvBool
s_dmaPollCondFunc
(
OBJGPU *pGpu,
void *pVoid
)
{
DmaPollCondData *pData = (DmaPollCondData *)pVoid;
return ((kflcnRegRead_HAL(pGpu, pData->pKernelFlcn, NV_PFALCON_FALCON_DMATRFCMD) & pData->pollMask) == pData->pollValue);
}
/*!
* Poll on either _FULL or _IDLE field of NV_PFALCON_FALCON_DMATRFCMD
*
* @param[in] pGpu GPU object pointer
* @param[in] pKernelFlcn pKernelFlcn object pointer
* @param[in] mode FLCN_DMA_POLL_QUEUE_NOT_FULL for poll on _FULL; return when _FULL is false
* FLCN_DMA_POLL_ENGINE_IDLE for poll on _IDLE; return when _IDLE is true
*/
static NV_STATUS
s_dmaPoll_GA102
(
OBJGPU *pGpu,
KernelFalcon *pKernelFlcn,
FlcnDmaPollMode mode
)
{
NV_STATUS status;
DmaPollCondData data;
data.pKernelFlcn = pKernelFlcn;
if (mode == FLCN_DMA_POLL_QUEUE_NOT_FULL)
{
data.pollMask = DRF_SHIFTMASK(NV_PFALCON_FALCON_DMATRFCMD_FULL);
data.pollValue = DRF_DEF(_PFALCON, _FALCON_DMATRFCMD, _FULL, _FALSE);
}
else
{
data.pollMask = DRF_SHIFTMASK(NV_PFALCON_FALCON_DMATRFCMD_IDLE);
data.pollValue = DRF_DEF(_PFALCON, _FALCON_DMATRFCMD, _IDLE, _TRUE);
}
status = gpuTimeoutCondWait(pGpu, s_dmaPollCondFunc, &data, NULL);
if (status != NV_OK)
{
NV_PRINTF(LEVEL_ERROR, "Error while waiting for Falcon DMA; mode: %d, status: 0x%08x\n", mode, status);
DBG_BREAKPOINT();
return status;
}
return NV_OK;
}
static NV_STATUS
s_dmaTransfer_GA102
(
@ -48,15 +109,20 @@ s_dmaTransfer_GA102
)
{
NV_STATUS status = NV_OK;
RMTIMEOUT timeout;
NvU32 data;
NvU32 bytesXfered = 0;
// Ensure request queue initially has space or writing base registers will corrupt DMA transfer.
NV_CHECK_OK_OR_RETURN(LEVEL_SILENT, s_dmaPoll_GA102(pGpu, pKernelFlcn, FLCN_DMA_POLL_QUEUE_NOT_FULL));
kflcnRegWrite_HAL(pGpu, pKernelFlcn, NV_PFALCON_FALCON_DMATRFBASE, NvU64_LO32(srcPhysAddr >> 8));
kflcnRegWrite_HAL(pGpu, pKernelFlcn, NV_PFALCON_FALCON_DMATRFBASE1, NvU64_HI32(srcPhysAddr >> 8) & 0x1FF);
while (bytesXfered < sizeInBytes)
{
// Poll for non-full request queue as writing control registers when full will corrupt DMA transfer.
NV_CHECK_OK_OR_RETURN(LEVEL_SILENT, s_dmaPoll_GA102(pGpu, pKernelFlcn, FLCN_DMA_POLL_QUEUE_NOT_FULL));
data = FLD_SET_DRF_NUM(_PFALCON, _FALCON_DMATRFMOFFS, _OFFS, dest, 0);
kflcnRegWrite_HAL(pGpu, pKernelFlcn, NV_PFALCON_FALCON_DMATRFMOFFS, data);
@ -66,28 +132,17 @@ s_dmaTransfer_GA102
// Write the command
kflcnRegWrite_HAL(pGpu, pKernelFlcn, NV_PFALCON_FALCON_DMATRFCMD, dmaCmd);
// Poll for completion
data = kflcnRegRead_HAL(pGpu, pKernelFlcn, NV_PFALCON_FALCON_DMATRFCMD);
gpuSetTimeout(pGpu, GPU_TIMEOUT_DEFAULT, &timeout, 0);
while(FLD_TEST_DRF(_PFALCON_FALCON, _DMATRFCMD, _IDLE, _FALSE, data))
{
status = gpuCheckTimeout(pGpu, &timeout);
if (status == NV_ERR_TIMEOUT)
{
NV_PRINTF(LEVEL_ERROR, "Timeout waiting for Falcon DMA to finish\n");
DBG_BREAKPOINT();
return status;
}
osSpinLoop();
data = kflcnRegRead_HAL(pGpu, pKernelFlcn, NV_PFALCON_FALCON_DMATRFCMD);
}
bytesXfered += FLCN_BLK_ALIGNMENT;
dest += FLCN_BLK_ALIGNMENT;
memOff += FLCN_BLK_ALIGNMENT;
}
//
// Poll for completion. GA10x+ does not have TCM tagging so DMA operations to/from TCM should
// wait for DMA to complete before launching another operation to avoid memory ordering problems.
//
NV_CHECK_OK_OR_RETURN(LEVEL_SILENT, s_dmaPoll_GA102(pGpu, pKernelFlcn, FLCN_DMA_POLL_ENGINE_IDLE));
return status;
}

View File

@ -2390,6 +2390,31 @@ error_cleanup:
return nvStatus;
}
/*!
* Preserve vGPU Partition log buffers between VM reboots
*/
NV_STATUS
kgspPreserveVgpuPartitionLogging_IMPL
(
OBJGPU *pGpu,
KernelGsp *pKernelGsp,
NvU32 gfid
)
{
if ((gfid == 0) || (gfid > MAX_PARTITIONS_WITH_GFID))
{
return NV_ERR_INVALID_ARGUMENT;
}
// Make sure this this NvLog buffer is pushed
kgspDumpGspLogsUnlocked(pKernelGsp, NV_FALSE);
// Preserve any captured vGPU Partition logs
libosPreserveLogs(&pKernelGsp->logDecodeVgpuPartition[gfid - 1]);
return NV_OK;
}
void kgspNvlogFlushCb(void *pKernelGsp)
{
if (pKernelGsp != NULL)
@ -3449,7 +3474,9 @@ kgspDumpGspLogsUnlocked_IMPL
NvBool bSyncNvLog
)
{
if (pKernelGsp->bInInit || pKernelGsp->pLogElf || bSyncNvLog)
if (pKernelGsp->bInInit || pKernelGsp->pLogElf || bSyncNvLog
|| pKernelGsp->bHasVgpuLogs
)
{
libosExtractLogs(&pKernelGsp->logDecode, bSyncNvLog);
@ -3479,7 +3506,9 @@ kgspDumpGspLogs_IMPL
NvBool bSyncNvLog
)
{
if (pKernelGsp->bInInit || pKernelGsp->pLogElf || bSyncNvLog)
if (pKernelGsp->bInInit || pKernelGsp->pLogElf || bSyncNvLog
|| pKernelGsp->bHasVgpuLogs
)
{
if (pKernelGsp->pNvlogFlushMtx != NULL)
portSyncMutexAcquire(pKernelGsp->pNvlogFlushMtx);

View File

@ -241,7 +241,6 @@ subdeviceCtrlCmdEventSetSemaphoreMemory_IMPL
pMemory->vgpuNsIntr.guestMSIAddr = 0;
pMemory->vgpuNsIntr.guestMSIData = 0;
pMemory->vgpuNsIntr.guestDomainId = 0;
pMemory->vgpuNsIntr.pVgpuVfioRef = NULL;
pMemory->vgpuNsIntr.isSemaMemValidationEnabled = NV_TRUE;
return NV_OK;

View File

@ -436,21 +436,22 @@ tmrEventTimeUntilNextCallback_IMPL
NvU64 currentTime;
NvU64 nextAlarmTime;
NV_ASSERT_OK_OR_RETURN(tmrGetCurrentTime(pTmr, &currentTime));
TMR_EVENT_PVT *pEvent = (TMR_EVENT_PVT*)pEventPublic;
if (tmrIsOSTimer(pTmr, pEventPublic))
{
// timens corresponds to relative time for OS timer
osGetCurrentTick(&currentTime);
// timens corresponds to relative time for OS timer
NV_CHECK_OR_RETURN(LEVEL_ERROR, portSafeAddU64(pEvent->timens, pEvent->startTimeNs, &nextAlarmTime),
NV_ERR_INVALID_ARGUMENT);
}
else
{
NV_ASSERT_OK_OR_RETURN(tmrGetCurrentTime(pTmr, &currentTime));
// timens corresponds to abs time in case of ptimer
nextAlarmTime = pEvent->timens;
}
if (currentTime >= nextAlarmTime)
if (currentTime > nextAlarmTime)
return NV_ERR_INVALID_STATE;
*pTimeUntilCallbackNs = nextAlarmTime - currentTime;

View File

@ -4477,6 +4477,23 @@ cliresCtrlCmdSyncGpuBoostGroupInfo_IMPL
return status;
}
NV_STATUS
cliresCtrlCmdVgpuVfioNotifyRMStatus_IMPL
(
RmClientResource *pRmCliRes,
NV0000_CTRL_VGPU_VFIO_NOTIFY_RM_STATUS_PARAMS *pVgpuStatusParams
)
{
if (osIsVgpuVfioPresent() != NV_OK)
return NV_ERR_NOT_SUPPORTED;
osWakeRemoveVgpu(pVgpuStatusParams->gpuId, pVgpuStatusParams->returnStatus);
return NV_OK;
}
NV_STATUS
cliresCtrlCmdVgpuGetVgpuVersion_IMPL
(

View File

@ -273,6 +273,10 @@ kernelhostvgpudeviceapiConstruct_IMPL
status = pRmApi->Control(pRmApi, pGpu->hInternalClient, pGpu->hInternalSubdevice,
NV2080_CTRL_CMD_VGPU_MGR_INTERNAL_BOOTLOAD_GSP_VGPU_PLUGIN_TASK,
pBootloadParams, sizeof(*pBootloadParams));
// Preserve any captured vGPU Partition logs
NV_ASSERT_OK(kgspPreserveVgpuPartitionLogging(pGpu, pKernelGsp, pAllocParams->gfid));
if (status != NV_OK)
{
NV_PRINTF(LEVEL_ERROR, "Failed to call NV2080_CTRL_CMD_VGPU_MGR_INTERNAL_BOOTLOAD_GSP_VGPU_PLUGIN_TASK\n");

View File

@ -1,4 +1,4 @@
NVIDIA_VERSION = 550.100
NVIDIA_VERSION = 550.107.02
# This file.
VERSION_MK_FILE := $(lastword $(MAKEFILE_LIST))