mirror of
git://nv-tegra.nvidia.com/linux-nvgpu.git
synced 2025-12-25 02:52:51 +03:00
86b31c4f7c
Historically, nvgpu has supported a struct gk20a_dmabuf_priv and associated it with a dmabuf instance. This was aided by Nvmap's dma_buf_set_drv_data() and dma_buf_get_drvdata() APIs. gk20a_dmabuf_priv is used to store Comptag IDs i.e. (1 per 64 kb) as well as can store the dmabuf attachments to avoid multiple attach/detach calls. dma_buf_set_drv_data() allows Nvgpu to associate an instance of struct gk20a_dmabuf_priv with the instance of the dmabuf and also provide a release callback to delete the instance when the last reference to the dmabuf is put. Nvmap accomplishes this by modifying the struct dma_buf_ops definition to include the set_drv_data and get_drv_data callbacks in the kernel code. The above approach won't work for upstream Kstable and Nvmap plans to remove these APIs for upcoming newer downstream kernels as well. In order to implement the same functionality without depending on Nvmap, Nvgpu will implement a release chaining mechanism. Dmabuf's 'ops' pointer points to a constant struct and hence a whole copy of the ops is made followed by altering the new copy's release pointer. struct gk20a_dmabuf_priv stores the new copy and the dmabuf's 'ops' is changed to point to this. This allows Nvgpu to retrieve the corresponding gk20a_dmabuf_priv instance using container_of. Nvgpu's custom release callback will invoke the original release callback of the dmabuf's producer as a last step, thus completing the full circle. In case, the driver is removed, Nvgpu restores the dmabuf's 'ops' back to the original state. In order to accomplish this, every instance of a struct nvgpu_os_linux maintains a linkedlist of the gk20a_dma_buf instances. During the driver removal, this linkedlist is traversed and the corresponding dmabuf's 'ops' pointer is put back to its original state followed by freeing of this instance. Nvgpu is a producer of dmabuf's for vidmem and needs a way to check whether the given dmabuf belongs to itself. Its no longer reliable to depend on a comparision of the 'ops' pointer. Instead dmabuf_export_info() allows a name to be set by the exporter and this can be used to compare with a memory location that belongs to Nvgpu. Similarly for sysmem dmabufs, Nvmap makes a similar change in the way it identifies whether a dmabuf belongs to itself. Removed NVGPU_DMABUF_HAS_DRVDATA and moved to a unified mechanism for both downstream as well as upstream kernel. Some of the other changes in this file include the following. 1) Deletion of dmabuf.c and moving its contents over to dmabuf_priv.c 2) Replacing gk20a_mm_pin_has_drvdata with nvgpu_mm_pin_privdata and vice-versa for unpin. Bug 2878569 Change-Id: Icf8e79b05a25ad5a85f478c3ee0fc1eb7747e22d Signed-off-by: Debarshi Dutta <ddutta@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2341001 Reviewed-by: automaticguardword <automaticguardword@nvidia.com> Reviewed-by: Puneet Saxena <puneets@nvidia.com> Reviewed-by: Konsta Holtta <kholtta@nvidia.com> Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com> Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> GVS: Gerrit_Virtual_Submit Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
261 lines
6.3 KiB
Plaintext
261 lines
6.3 KiB
Plaintext
config GK20A
|
|
tristate "Nvidia GK20A GPU support"
|
|
default n
|
|
help
|
|
Choose this option if you have an SoC with integrated
|
|
Nvidia GPU IP.
|
|
|
|
config GK20A_PMU
|
|
bool "Support GK20A PMU"
|
|
depends on GK20A
|
|
default n
|
|
help
|
|
Say Y here to enable GK20A PMU features.
|
|
|
|
choice
|
|
depends on GK20A
|
|
prompt "Enable GK20A frequency scaling"
|
|
default GK20A_DEVFREQ
|
|
optional
|
|
help
|
|
Select this entry to enable gk20a scaling
|
|
|
|
config GK20A_DEVFREQ
|
|
depends on COMMON_CLK && PM_DEVFREQ
|
|
bool "Use Devfreq"
|
|
help
|
|
Select this to use devfreq based scaling.
|
|
Devfreq is a common framework that allows using
|
|
variety of different governors and changing
|
|
between governors on the fly. By default, no
|
|
governor is selected.
|
|
|
|
endchoice
|
|
|
|
config GK20A_PM_QOS
|
|
depends on GK20A_DEVFREQ
|
|
bool "Use PM_QOS constraints for devfreq based scaling"
|
|
default y
|
|
help
|
|
Enable support to pass PM_QOS constraints to devfreq based
|
|
scaling.
|
|
|
|
config NVGPU_TRACK_MEM_USAGE
|
|
bool "Track the usage of system memory in nvgpu"
|
|
depends on GK20A
|
|
default n
|
|
help
|
|
Say Y here to allow nvgpu to track and keep statistics on
|
|
the system memory used by the driver. This does recreate
|
|
some of the kmem_leak tracking but this is also applicable
|
|
to other OSes which do not have Linux' kmem_leak.
|
|
|
|
|
|
config NVGPU_CYCLESTATS
|
|
bool "Support GPU CYCLE STATS"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Say Y here to enable the cycle stats debugging features.
|
|
|
|
config NVGPU_FECS_TRACE
|
|
bool "Support NVGPU FECS Context Switch tracing"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Enable support for the NVGPU Context Switch Tracing. In this mode,
|
|
FECS collects timestamps for contexts loaded on GR engine. This
|
|
allows tracking context switches on GR engine, as well as
|
|
identifying processes that submitted work.
|
|
|
|
config TEGRA_GK20A_NVHOST
|
|
bool "Support interfacing with nvhost hardware"
|
|
depends on TEGRA_GRHOST || TEGRA_HOST1X
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Enable support in GK20A for the nvhost (host1x) dma engine hardware
|
|
that includes things like hardware syncpts. This requires either
|
|
TEGRA_GRHOST or TEGRA_HOST1X.
|
|
|
|
config TEGRA_GK20A
|
|
bool "Enable the GK20A GPU on Tegra"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Enable support for the GK20A graphics engine on Tegra
|
|
by adding a Tegra platfrom interface to the GK20A driver.
|
|
|
|
config TEGRA_ACR
|
|
bool "Enable HS bin support on GM20B GPU on Tegra"
|
|
depends on GK20A && GK20A_PMU
|
|
default n
|
|
help
|
|
Enable Support for Loading High Secure binary, and using
|
|
Write Protected Regions (WPR) for storing ucodes, and bootstrap
|
|
PMU, FECS and GPCCS in Low Secure mode.
|
|
|
|
config TEGRA_USE_NA_GPCPLL
|
|
bool "Enable noise aware mode of GM20B GPCPLL on Tegra"
|
|
depends on GK20A
|
|
depends on COMMON_CLK
|
|
default n
|
|
help
|
|
Enable noise aware (NA) mode of GM20b GPCPLL. In this mode PLL output
|
|
frequency is automatically adjusted when GM20b voltage is fluctuating
|
|
because of transient PMIC or power distribution tree noise.
|
|
|
|
config GK20A_PCI
|
|
bool "Support PCIe NVIDIA GPUs on nvgpu"
|
|
depends on PCI && GK20A
|
|
default y
|
|
help
|
|
Enable support for GPUs on PCIe bus.
|
|
|
|
config GK20A_TRACE_PRINTK
|
|
bool "Support trace_printk debugging"
|
|
depends on GK20A && TRACING
|
|
default n
|
|
help
|
|
Enable nvgpu debug facility to redirect debug spew to ftrace. This
|
|
affects kernel memory use, so should not be enabled by default.
|
|
|
|
config NVGPU_USE_TEGRA_ALLOC_FD
|
|
bool "Use tegra_alloc_fd() for allocating dma_buf fds for vidmem"
|
|
depends on GK20A
|
|
depends on NV_TEGRA_MC
|
|
default y
|
|
help
|
|
Use tegra_alloc_fd() for allocating dma_buf fds. This allocates
|
|
the fds above 1024 which exempts them from counting against process
|
|
fd limit.
|
|
|
|
It is only available in Tegra kernel.
|
|
|
|
config NVGPU_SUPPORT_TURING
|
|
bool "Turing family GPU"
|
|
depends on GK20A && ARCH_TEGRA_19x_SOC
|
|
default y
|
|
help
|
|
Support for NVIDIA Turing family of GPU
|
|
|
|
config NVGPU_DEBUGGER
|
|
bool "Support debugger APIs"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Support for debugger APIs
|
|
|
|
config NVGPU_LS_PMU
|
|
bool "LS PMU support"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Support for iGPU LS PMU enable/disable
|
|
|
|
config NVGPU_LOGGING
|
|
bool "NVGPU logging"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Enable/Disable NVGPU logging
|
|
|
|
config NVGPU_HAL_NON_FUSA
|
|
bool "Support non-functionally safe HALs"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Enable/Disable the support of HALs from chips that do not have functional
|
|
safety certification
|
|
|
|
config NVGPU_RECOVERY
|
|
bool "Recovery support"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Support recovery on failure (which may involve engine reset)
|
|
|
|
config NVGPU_NVLINK
|
|
bool "Nvlink support"
|
|
depends on TEGRA_NVLINK
|
|
default y
|
|
help
|
|
Support Nvlink
|
|
|
|
config NVGPU_GR_VIRTUALIZATION
|
|
bool "NVGPU Virtualization Support"
|
|
depends on TEGRA_GR_VIRTUALIZATION
|
|
default y
|
|
help
|
|
Support NVGPU Virtualization
|
|
|
|
config NVGPU_COMPRESSION
|
|
bool "Compression support"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Support for compression
|
|
|
|
config NVGPU_SUPPORT_CDE
|
|
bool "Support extraction of comptags for CDE"
|
|
depends on GK20A
|
|
depends on NVGPU_COMPRESSION
|
|
default n
|
|
help
|
|
Enable support for extraction of comptags for CDE.
|
|
|
|
config NVGPU_DGPU
|
|
bool "NVGPU DGPU support"
|
|
depends on GK20A
|
|
depends on GK20A_PCI
|
|
default y
|
|
help
|
|
Support for NVGPU DGPU
|
|
|
|
config NVGPU_VPR
|
|
bool "NVGPU VPR support"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Support for NVGPU VPR
|
|
|
|
config NVGPU_TEGRA_FUSE
|
|
bool "Tegra fuse Support"
|
|
depends on GK20A
|
|
default y
|
|
help
|
|
Support Tegra fuse
|
|
|
|
choice
|
|
prompt "Supported sync fence backend"
|
|
default NVGPU_SYNCFD_ANDROID
|
|
depends on GK20A
|
|
help
|
|
GPU job synchronization (fences before and after submits) can use raw
|
|
syncpoints if available and sync fds if chosen. Without syncpoints,
|
|
nvgpu also provides semaphore-backed sync fds to userspace.
|
|
|
|
Select which kernel-provided API is used for sync fds. Matching
|
|
support is required for the userspace drivers too.
|
|
|
|
config NVGPU_SYNCFD_STABLE
|
|
bool "Upstream stabilized SYNC_FILE"
|
|
depends on SYNC_FILE
|
|
help
|
|
Select CONFIG_SYNC_FILE, the cross-driver "Explicit Synchronization
|
|
Framework" stabilized upstream from the Android sync since v4.9.
|
|
|
|
config NVGPU_SYNCFD_ANDROID
|
|
bool "Android SYNC"
|
|
depends on SYNC
|
|
help
|
|
Select CONFIG_SYNC, the legacy synchronization framework provided by
|
|
Android (and deprecated in Linux 4.9+ in favor of dma fences and the
|
|
stabilized SYNC_FILE).
|
|
|
|
config NVGPU_SYNCFD_NONE
|
|
bool "Nothing"
|
|
help
|
|
Do not build in support for sync fences.
|
|
endchoice
|