include: Bump-up version to 1.0

This patch updates OpenSBI version to 1.0 as part of
release preparation.

Signed-off-by: Anup Patel <anup.patel@wdc.com>

Makefile

@@ -76,26 +76,54 @@ OPENSBI_VERSION_MINOR=`grep "define OPENSBI_VERSION_MINOR" $(include_dir)/sbi/sb
 OPENSBI_VERSION_GIT=$(shell if [ -d $(src_dir)/.git ]; then git describe 2> /dev/null; fi)
 
 # Setup compilation commands
+ifneq ($(LLVM),)
+CC		=	clang
+AR		=	llvm-ar
+LD		=	ld.lld
+OBJCOPY		=	llvm-objcopy
+else
 ifdef CROSS_COMPILE
 CC		=	$(CROSS_COMPILE)gcc
-CPP		=	$(CROSS_COMPILE)cpp
 AR		=	$(CROSS_COMPILE)ar
 LD		=	$(CROSS_COMPILE)ld
 OBJCOPY		=	$(CROSS_COMPILE)objcopy
 else
 CC		?=	gcc
-CPP		?=	cpp
 AR		?=	ar
 LD		?=	ld
 OBJCOPY		?=	objcopy
 endif
+endif
+CPP		=	$(CC) -E
 AS		=	$(CC)
 DTC		=	dtc
 
-# Guess the compillers xlen
-OPENSBI_CC_XLEN := $(shell TMP=`$(CC) -dumpmachine | sed 's/riscv\([0-9][0-9]\).*/\1/'`; echo $${TMP})
+ifneq ($(shell $(CC) --version 2>&1 | head -n 1 | grep clang),)
+CC_IS_CLANG	=	y
+else
+CC_IS_CLANG	=	n
+endif
+
+ifneq ($(shell $(LD) --version 2>&1 | head -n 1 | grep LLD),)
+LD_IS_LLD	=	y
+else
+LD_IS_LLD	=	n
+endif
+
+ifeq ($(CC_IS_CLANG),y)
+ifneq ($(CROSS_COMPILE),)
+CLANG_TARGET	=	--target=$(notdir $(CROSS_COMPILE:%-=%))
+endif
+endif
+
+# Guess the compiler's XLEN
+OPENSBI_CC_XLEN := $(shell TMP=`$(CC) $(CLANG_TARGET) -dumpmachine | sed 's/riscv\([0-9][0-9]\).*/\1/'`; echo $${TMP})
+
+# Guess the compiler's ABI and ISA
+ifneq ($(CC_IS_CLANG),y)
 OPENSBI_CC_ABI := $(shell TMP=`$(CC) -v 2>&1 | sed -n 's/.*\(with\-abi=\([a-zA-Z0-9]*\)\).*/\2/p'`; echo $${TMP})
 OPENSBI_CC_ISA := $(shell TMP=`$(CC) -v 2>&1 | sed -n 's/.*\(with\-arch=\([a-zA-Z0-9]*\)\).*/\2/p'`; echo $${TMP})
+endif
 
 # Setup platform XLEN
 ifndef PLATFORM_RISCV_XLEN
@@ -106,6 +134,44 @@ ifndef PLATFORM_RISCV_XLEN
   endif
 endif
 
+ifeq ($(CC_IS_CLANG),y)
+ifeq ($(CROSS_COMPILE),)
+CLANG_TARGET	=	--target=riscv$(PLATFORM_RISCV_XLEN)-unknown-elf
+endif
+endif
+
+ifeq ($(LD_IS_LLD),y)
+RELAX_FLAG	=	-mno-relax
+USE_LD_FLAG	=	-fuse-ld=lld
+else
+USE_LD_FLAG	=	-fuse-ld=bfd
+endif
+
+# Check whether the linker supports creating PIEs
+OPENSBI_LD_PIE := $(shell $(CC) $(CLANG_TARGET) $(RELAX_FLAG) $(USE_LD_FLAG) -fPIE -nostdlib -Wl,-pie -x c /dev/null -o /dev/null >/dev/null 2>&1 && echo y || echo n)
+
+# Check whether the compiler supports -m(no-)save-restore
+CC_SUPPORT_SAVE_RESTORE := $(shell $(CC) $(CLANG_TARGET) $(RELAX_FLAG) -nostdlib -mno-save-restore -x c /dev/null -o /dev/null 2>&1 | grep "\-save\-restore" >/dev/null && echo n || echo y)
+
+# Build Info:
+# OPENSBI_BUILD_TIME_STAMP -- the compilation time stamp
+# OPENSBI_BUILD_COMPILER_VERSION -- the compiler version info
+BUILD_INFO ?= n
+ifeq ($(BUILD_INFO),y)
+OPENSBI_BUILD_DATE_FMT = +%Y-%m-%d %H:%M:%S %z
+ifdef SOURCE_DATE_EPOCH
+	OPENSBI_BUILD_TIME_STAMP ?= $(shell date -u -d "@$(SOURCE_DATE_EPOCH)" \
+		"$(OPENSBI_BUILD_DATE_FMT)" 2>/dev/null || \
+		date -u -r "$(SOURCE_DATE_EPOCH)" \
+		"$(OPENSBI_BUILD_DATE_FMT)" 2>/dev/null || \
+		date -u "$(OPENSBI_BUILD_DATE_FMT)")
+else
+	OPENSBI_BUILD_TIME_STAMP ?= $(shell date "$(OPENSBI_BUILD_DATE_FMT)")
+endif
+OPENSBI_BUILD_COMPILER_VERSION=$(shell $(CC) -v 2>&1 | grep ' version ' | \
+	sed 's/[[:space:]]*$$//')
+endif
+
 # Setup list of objects.mk files
 ifdef PLATFORM
 platform-object-mks=$(shell if [ -d $(platform_src_dir)/ ]; then find $(platform_src_dir) -iname "objects.mk" | sort -r; fi)
@@ -194,46 +260,72 @@ else
 endif
 
 # Setup compilation commands flags
-GENFLAGS	=	-I$(platform_src_dir)/include
+ifeq ($(CC_IS_CLANG),y)
+GENFLAGS	+=	$(CLANG_TARGET)
+GENFLAGS	+=	-Wno-unused-command-line-argument
+endif
+GENFLAGS	+=	-I$(platform_src_dir)/include
 GENFLAGS	+=	-I$(include_dir)
 ifneq ($(OPENSBI_VERSION_GIT),)
 GENFLAGS	+=	-DOPENSBI_VERSION_GIT="\"$(OPENSBI_VERSION_GIT)\""
 endif
+ifeq ($(BUILD_INFO),y)
+GENFLAGS	+=	-DOPENSBI_BUILD_TIME_STAMP="\"$(OPENSBI_BUILD_TIME_STAMP)\""
+GENFLAGS	+=	-DOPENSBI_BUILD_COMPILER_VERSION="\"$(OPENSBI_BUILD_COMPILER_VERSION)\""
+endif
 GENFLAGS	+=	$(libsbiutils-genflags-y)
 GENFLAGS	+=	$(platform-genflags-y)
 GENFLAGS	+=	$(firmware-genflags-y)
 
-CFLAGS		=	-g -Wall -Werror -ffreestanding -nostdlib -fno-strict-aliasing -O2
-CFLAGS		+=	-fno-omit-frame-pointer -fno-optimize-sibling-calls
-CFLAGS		+=	-mno-save-restore -mstrict-align
+CFLAGS		=	-g -Wall -Werror -ffreestanding -nostdlib -fno-stack-protector -fno-strict-aliasing -O2
+CFLAGS		+=	-fno-omit-frame-pointer -fno-optimize-sibling-calls -mstrict-align
+# enable -m(no-)save-restore option by CC_SUPPORT_SAVE_RESTORE
+ifeq ($(CC_SUPPORT_SAVE_RESTORE),y)
+CFLAGS		+=	-mno-save-restore
+endif
 CFLAGS		+=	-mabi=$(PLATFORM_RISCV_ABI) -march=$(PLATFORM_RISCV_ISA)
 CFLAGS		+=	-mcmodel=$(PLATFORM_RISCV_CODE_MODEL)
+CFLAGS		+=	$(RELAX_FLAG)
 CFLAGS		+=	$(GENFLAGS)
 CFLAGS		+=	$(platform-cflags-y)
-CFLAGS		+=	$(firmware-cflags-y)
 CFLAGS		+=	-fno-pie -no-pie
+CFLAGS		+=	$(firmware-cflags-y)
 
 CPPFLAGS	+=	$(GENFLAGS)
 CPPFLAGS	+=	$(platform-cppflags-y)
 CPPFLAGS	+=	$(firmware-cppflags-y)
 
-ASFLAGS		=	-g -Wall -nostdlib -D__ASSEMBLY__
-ASFLAGS		+=	-fno-omit-frame-pointer -fno-optimize-sibling-calls
-ASFLAGS		+=	-mno-save-restore -mstrict-align
+ASFLAGS		=	-g -Wall -nostdlib
+ASFLAGS		+=	-fno-omit-frame-pointer -fno-optimize-sibling-calls -mstrict-align
+# enable -m(no-)save-restore option by CC_SUPPORT_SAVE_RESTORE
+ifeq ($(CC_SUPPORT_SAVE_RESTORE),y)
+ASFLAGS		+=	-mno-save-restore
+endif
 ASFLAGS		+=	-mabi=$(PLATFORM_RISCV_ABI) -march=$(PLATFORM_RISCV_ISA)
 ASFLAGS		+=	-mcmodel=$(PLATFORM_RISCV_CODE_MODEL)
+ASFLAGS		+=	$(RELAX_FLAG)
+ifneq ($(CC_IS_CLANG),y)
+ifneq ($(RELAX_FLAG),)
+ASFLAGS		+=	-Wa,$(RELAX_FLAG)
+endif
+endif
 ASFLAGS		+=	$(GENFLAGS)
 ASFLAGS		+=	$(platform-asflags-y)
 ASFLAGS		+=	$(firmware-asflags-y)
 
 ARFLAGS		=	rcs
 
-ELFFLAGS	+=	-Wl,--build-id=none -N -static-libgcc -lgcc
+ELFFLAGS	+=	$(USE_LD_FLAG)
+ELFFLAGS	+=	-Wl,--build-id=none -Wl,-N
 ELFFLAGS	+=	$(platform-ldflags-y)
 ELFFLAGS	+=	$(firmware-ldflags-y)
 
 MERGEFLAGS	+=	-r
+ifeq ($(LD_IS_LLD),y)
+MERGEFLAGS	+=	-b elf
+else
 MERGEFLAGS	+=	-b elf$(PLATFORM_RISCV_XLEN)-littleriscv
+endif
 MERGEFLAGS	+=	-m elf$(PLATFORM_RISCV_XLEN)lriscv
 
 DTSCPPFLAGS	=	$(CPPFLAGS) -nostdinc -nostdlib -fno-builtin -D__DTS__ -x assembler-with-cpp
@@ -338,6 +430,11 @@ $(build_dir)/%.dep: $(src_dir)/%.c
 $(build_dir)/%.o: $(src_dir)/%.c
 	$(call compile_cc,$@,$<)
 
+ifeq ($(BUILD_INFO),y)
+$(build_dir)/lib/sbi/sbi_init.o: $(libsbi_dir)/sbi_init.c FORCE
+	$(call compile_cc,$@,$<)
+endif
+
 $(build_dir)/%.dep: $(src_dir)/%.S
 	$(call compile_as_dep,$@,$<)
 
@@ -491,3 +588,6 @@ ifeq ($(install_root_dir),$(install_root_dir_default)/usr)
 	$(if $(V), @echo " RM        $(install_root_dir_default)")
 	$(CMD_PREFIX)rm -rf $(install_root_dir_default)
 endif
+
+.PHONY: FORCE
+FORCE:

README.md

@@ -96,8 +96,21 @@ Required Toolchain
 ------------------
 
 OpenSBI can be compiled natively or cross-compiled on a x86 host. For
-cross-compilation, you can build your own toolchain or just download
-a prebuilt one from the [Bootlin toolchain repository].
+cross-compilation, you can build your own toolchain, download a prebuilt one
+from the [Bootlin toolchain repository] or install a distribution-provided
+toolchain; if you opt to use LLVM/Clang, most distribution toolchains will
+support cross-compiling for RISC-V using the same toolchain as your native
+LLVM/Clang toolchain due to LLVM's ability to support multiple backends in the
+same binary, so is often an easy way to obtain a working cross-compilation
+toolchain.
+
+Basically, we prefer toolchains with Position Independent Executable (PIE)
+support like *riscv64-linux-gnu-gcc*, *riscv64-unknown-freebsd-gcc*, or
+*Clang/LLVM* as they generate PIE firmware images that can run at arbitrary
+address with appropriate alignment. If a bare-metal GNU toolchain (e.g.
+*riscv64-unknown-elf-gcc*) is used, static linked firmware images are
+generated instead. *Clang/LLVM* can still generate PIE images if a bare-metal
+triple is used (e.g. *-target riscv64-unknown-elf*).
 
 Please note that only a 64-bit version of the toolchain is available in
 the Bootlin toolchain repository for now.
@@ -111,7 +124,7 @@ architecture than RISC-V.
 
 For cross-compiling, the environment variable *CROSS_COMPILE* must be defined
 to specify the name prefix of the RISC-V compiler toolchain executables, e.g.
-*riscv64-unknown-elf-* if the gcc executable used is *riscv64-unknown-elf-gcc*.
+*riscv64-linux-gnu-* if the gcc executable used is *riscv64-linux-gnu-gcc*.
 
 To build *libsbi.a* simply execute:
 ```
@@ -187,21 +200,83 @@ Building 32-bit / 64-bit OpenSBI Images
 ---------------------------------------
 By default, building OpenSBI generates 32-bit or 64-bit images based on the
 supplied RISC-V cross-compile toolchain. For example if *CROSS_COMPILE* is set
-to *riscv64-unknown-elf-*, 64-bit OpenSBI images will be generated. If building
+to *riscv64-linux-gnu-*, 64-bit OpenSBI images will be generated. If building
 32-bit OpenSBI images, *CROSS_COMPILE* should be set to a toolchain that is
-pre-configured to generate 32-bit RISC-V codes, like *riscv32-unknown-elf-*.
+pre-configured to generate 32-bit RISC-V codes, like *riscv32-linux-gnu-*.
 
 However it's possible to explicitly specify the image bits we want to build with
 a given RISC-V toolchain. This can be done by setting the environment variable
 *PLATFORM_RISCV_XLEN* to the desired width, for example:
 
 ```
-export CROSS_COMPILE=riscv64-unknown-elf-
+export CROSS_COMPILE=riscv64-linux-gnu-
 export PLATFORM_RISCV_XLEN=32
 ```
 
 will generate 32-bit OpenSBI images. And vice vesa.
 
+Building with Clang/LLVM
+------------------------
+
+OpenSBI can also be built with Clang/LLVM. To build with just Clang but keep
+the default binutils (which will still use the *CROSS_COMPILE* prefix if
+defined), override the *CC* make variable with:
+```
+make CC=clang
+```
+
+To build with a full LLVM-based toolchain, not just Clang, enable the *LLVM*
+option with:
+```
+make LLVM=1
+```
+
+When using Clang, *CROSS_COMPILE* often does not need to be defined unless
+using GNU binutils with prefixed binary names. *PLATFORM_RISCV_XLEN* will be
+used to infer a default triple to pass to Clang, so if *PLATFORM_RISCV_XLEN*
+itself defaults to an undesired value then prefer setting that rather than the
+full triple via *CROSS_COMPILE*. If *CROSS_COMPILE* is nonetheless defined,
+rather than being used as a prefix for the executable name, it will instead be
+passed via the `--target` option with the trailing `-` removed, so must be a
+valid triple.
+
+These can also be mixed; for example using a GCC cross-compiler but LLVM
+binutils would be:
+```
+make CC=riscv64-linux-gnu-gcc LLVM=1
+```
+
+These variables must be passed for all the make invocations described in this
+document.
+
+NOTE: Using Clang with a `riscv*-linux-gnu` GNU binutils linker has been seen
+to produce broken binaries with missing relocations; it is therefore currently
+recommended that this combination be avoided or *FW_PIC=n* be used to disable
+building OpenSBI as a position-independent binary.
+
+Building with timestamp and compiler info
+-----------------------------------------
+
+When doing development, we may want to know the build time and compiler info
+for debug purpose. OpenSBI can also be built with timestamp and compiler info.
+To build with those info and print it out at boot time, we can just simply add
+`BUILD_INFO=y`, like:
+```
+make BUILD_INFO=y
+```
+
+But if you have used `BUILD_INFO=y`, and want to switch back to `BUILD_INFO=n`,
+you must do
+```
+make clean
+```
+before the next build.
+
+NOTE: Using `BUILD_INFO=y` without specifying SOURCE_DATE_EPOCH will violate
+[reproducible builds]. This definition is ONLY for development and debug
+purpose, and should NOT be used in a product which follows "reproducible
+builds".
+
 Contributing to OpenSBI
 -----------------------
 
@@ -284,3 +359,4 @@ make I=<install_directory> install_docs
 [Doxygen manual]: http://www.doxygen.nl/manual/index.html
 [Kendryte standalone SDK]: https://github.com/kendryte/kendryte-standalone-sdk
 [third party notices]: ThirdPartyNotices.md
+[reproducible builds]: https://reproducible-builds.org

docs/firmware/fw.md

@@ -9,6 +9,13 @@ OpenSBI generic library code. The supported firmwares type will differ in how
 the arguments passed by the platform early boot stage are handled, as well as
 how the boot stage following the firmware will be handled and executed.
 
+The previous booting stage will pass information via the following registers
+of RISC-V CPU:
+
+* hartid via *a0* register
+* device tree blob address in memory via *a1* register. The address must
+  be aligned to 8 bytes.
+
 OpenSBI currently supports three different types of firmwares.
 
 Firmware with Dynamic Information (*FW_DYNAMIC*)
@@ -62,6 +69,12 @@ parameters:
   argument by the prior booting stage.
 * **FW_FDT_PADDING** - Optional zero bytes padding to the embedded flattened
   device tree binary file specified by **FW_FDT_PATH** option.
+* **FW_PIC** - "FW_PIC=y" generates position independent executable firmware
+  images. OpenSBI can run at arbitrary address with appropriate alignment.
+  Therefore, the original relocation mechanism ("FW_PIC=n") will be skipped.
+  In other words, OpenSBI will directly run at the load address without any
+  code movement. This option requires a toolchain with PIE support, and it
+  is on by default.
 
 Additionally, each firmware type as a set of type specific configuration
 parameters. Detailed information for each firmware type can be found in the

docs/firmware/fw_dynamic.md

@@ -6,8 +6,9 @@ information about next booting stage (e.g. a bootloader or an OS) and runtime
 OpenSBI library options from previous booting stage.
 
 The previous booting stage will pass information to *FW_DYNAMIC* by creating
-*struct fw_dynamic_info* in memory and passing it's address to *FW_DYNAMIC*
-via *a2* register of RISC-V CPU.
+*struct fw_dynamic_info* in memory and passing its address to *FW_DYNAMIC*
+via *a2* register of RISC-V CPU. The address must be aligned to 8 bytes on
+RV64 and 4 bytes on RV32.
 
 A *FW_DYNAMIC* firmware is particularly useful when the booting stage executed
 prior to OpenSBI firmware is capable of loading both the OpenSBI firmware and

docs/platform/generic.md

@@ -46,9 +46,13 @@ RISC-V Platforms Using Generic Platform
 ---------------------------------------
 
 * **QEMU RISC-V Virt Machine** (*[qemu_virt.md]*)
-* **Spike** (*[spike.md]*)
 * **Shakti C-class SoC Platform** (*[shakti_cclass.md]*)
+* **SiFive HiFive Unleashed** (*[sifive_fu540.md]*)
+* **Spike** (*[spike.md]*)
+* **T-HEAD C9xx series Processors** (*[thead-c9xx.md]*)
 
 [qemu_virt.md]: qemu_virt.md
-[spike.md]: spike.md
 [shakti_cclass.md]: shakti_cclass.md
+[sifive_fu540.md]: sifive_fu540.md
+[spike.md]: spike.md
+[thead-c9xx.md]: thead-c9xx.md

docs/platform/platform.md

@@ -28,15 +28,12 @@ OpenSBI currently supports the following virtual and hardware platforms:
 * **Andes AE350 SoC**: Platform support for the Andes's SoC (AE350). More
   details on this platform can be found in the file *[andes-ae350.md]*.
 
-* **T-HEAD C910**: Platform support for the T-HEAD C910 Processor. More
-  details on this platform can be found in the file *[thead-c910.md]*.
-
 * **Spike**: Platform support for the Spike emulator. More
   details on this platform can be found in the file *[spike.md]*.
 
 * **OpenPiton FPGA SoC**: Platform support OpenPiton research platform based
   on ariane core. More details on this platform can be found in the file
-  *[fpga_openpiton.md]*.
+  *[fpga-openpiton.md]*.
 
 * **Shakti C-class SoC Platform**: Platform support for Shakti C-class
   processor based SOCs. More details on this platform can be found in the
@@ -55,5 +52,5 @@ facilitate the implementation.
 [andes-ae350.md]: andes-ae350.md
 [thead-c910.md]: thead-c910.md
 [spike.md]: spike.md
-[fpga_openpiton.md]: fpga_openpiton.md
+[fpga-openpiton.md]: fpga-openpiton.md
 [shakti_cclass.md]: shakti_cclass.md

docs/platform/qemu_virt.md

@@ -147,3 +147,27 @@ qemu-system-riscv32 -M virt -m 256M -nographic \
 	-device virtio-blk-device,drive=hd0 \
 	-append "root=/dev/vda rw console=ttyS0"
 ```
+
+Debugging with GDB
+------------------
+
+In a first console start OpenSBI with QEMU:
+
+```
+qemu-system-riscv64 -M virt -m 256M -nographic \
+	-bios build/platform/generic/firmware/fw_payload.bin \
+	-gdb tcp::1234 \
+	-S
+
+```
+
+Parameter *-gdb tcp::1234* specifies 1234 as the debug port.
+Parameter *-S* lets QEMU wait at the first instruction.
+
+In a second console start GDB:
+
+```
+gdb build/platform/generic/firmware/fw_payload.elf \
+	-ex 'target remote localhost:1234'
+
+```

docs/platform/sifive_fu540.md

@@ -8,7 +8,7 @@ With QEMU v4.2 or above release, the 'sifive_u' machine can be used to test
 OpenSBI image built for the real hardware as well.
 
 To build platform specific library and firmwares, provide the
-*PLATFORM=sifive/fu540* parameter to the top level `make` command.
+*PLATFORM=generic* parameter to the top level `make` command.
 
 Platform Options
 ----------------
@@ -32,20 +32,20 @@ To use Linux v5.2 (or higher), the pre-built DTB (DT binary) from Linux v5.2
 the compile time option *FW_FDT_PATH*.
 
 ```
-make PLATFORM=sifive/fu540 FW_PAYLOAD_PATH=<linux_build_directory>/arch/riscv/boot/Image
+make PLATFORM=generic FW_PAYLOAD_PATH=<linux_build_directory>/arch/riscv/boot/Image
 or
 (For Linux v5.2 or higher)
-make PLATFORM=sifive/fu540 FW_PAYLOAD_PATH=<linux_build_directory>/arch/riscv/boot/Image FW_FDT_PATH=<hifive-unleashed-a00.dtb path from Linux kernel>
+make PLATFORM=generic FW_PAYLOAD_PATH=<linux_build_directory>/arch/riscv/boot/Image FW_FDT_PATH=<hifive-unleashed-a00.dtb path from Linux kernel>
 ```
 
 **U-Boot Payload**
 
 The command-line example here assumes that U-Boot was compiled using the
 sifive_fu540_defconfig configuration and with U-Boot v2020.01, and up to
-v2020.07-rc3.
+v2021.04. sifive_unleashed_defconfig shall be used with v2021.07 or above.
 
 ```
-make PLATFORM=sifive/fu540 FW_PAYLOAD_PATH=<u-boot_build_dir>/u-boot-dtb.bin
+make PLATFORM=generic FW_PAYLOAD_PATH=<u-boot_build_dir>/u-boot-dtb.bin
 ```
 For U-Boot v2020.07-rc4 or later releases, SPL support was added in U-Boot.
 Please refer to the detailed U-Boot booting guide available at [U-Boot].
@@ -64,7 +64,7 @@ That's why the generated firmware binary in above steps should be copied to
 the partition of the sdcard with above GUID.
 
 ```
-dd if=build/platform/sifive/fu540/firmware/fw_payload.bin of=/dev/disk2s1 bs=1024
+dd if=build/platform/generic/firmware/fw_payload.bin of=/dev/disk2s1 bs=1024
 ```
 
 In my case, it is the first partition is **disk2s1** that has been formatted
@@ -153,14 +153,15 @@ This is because QEMU generates a device tree blob on the fly based on the
 command line parameters and it's compatible with the one used in the upstream
 Linux kernel.
 
-When U-Boot v2020.01 (or higher) is used as the payload, as the SiFive FU540
+When U-Boot v2021.07 (or higher) is used as the payload, as the SiFive FU540
 DTB for the real hardware is embedded in U-Boot binary itself, due to the same
-reason above, we need to switch the U-Boot sifive_fu540_defconfig configuration
-from **CONFIG_OF_SEPARATE** to **CONFIG_OF_PRIOR_STAGE** so that U-Boot uses the
-DTB generated by QEMU, and u-boot.bin should be used as the payload image, like:
+reason above, we need to switch the U-Boot sifive_unleashed_defconfig
+configuration from **CONFIG_OF_SEPARATE** to **CONFIG_OF_PRIOR_STAGE** so that
+U-Boot uses the DTB generated by QEMU, and u-boot.bin should be used as the
+payload image, like:
 
 ```
-make PLATFORM=sifive/fu540 FW_PAYLOAD_PATH=<u-boot_build_dir>/u-boot.bin
+make PLATFORM=generic FW_PAYLOAD_PATH=<u-boot_build_dir>/u-boot.bin
 ```
 
 U-Boot v2020.07 release added SPL support to SiFive HiFive Unleashed board,
@@ -178,12 +179,12 @@ The above errors can be safely ignored as we don't run U-Boot SPL under QEMU.
 Run:
 ```
 qemu-system-riscv64 -M sifive_u -m 256M -nographic \
-	-bios build/platform/sifive/fu540/firmware/fw_payload.bin
+	-bios build/platform/generic/firmware/fw_payload.bin
 ```
 or
 ```
 qemu-system-riscv64 -M sifive_u -m 256M -nographic \
-	-bios build/platform/sifive/fu540/firmware/fw_jump.bin \
+	-bios build/platform/generic/firmware/fw_jump.bin \
 	-kernel <uboot_build_dir>/u-boot.bin
 ```
 

docs/platform/spike.md

@@ -43,7 +43,18 @@ make PLATFORM=generic FW_PAYLOAD_PATH=<linux_build_directory>/arch/riscv/boot/Im
 
 Run:
 ```
-spike --initrd <path_to_cpio_ramdisk> build/platform/generic/firmware/fw_payload.elf
+spike -m256 \
+	--initrd <path_to_cpio_ramdisk> \
+	--bootargs 'root=/dev/ram rw console=hvc0 earlycon=sbi' \
+	build/platform/generic/firmware/fw_payload.elf
+```
+or
+```
+spike -m256 \
+	--kernel <linux_build_directory>/arch/riscv/boot/Image \
+	--initrd <path_to_cpio_ramdisk> \
+	--bootargs 'root=/dev/ram rw console=hvc0 earlycon=sbi' \
+	build/platform/generic/firmware/fw_jump.elf
 ```
 
 Execution on QEMU RISC-V 64-bit

docs/platform/thead-c910.md

@@ -1,34 +0,0 @@
-T-HEAD C910 Processor
-=====================
-C910 is a 12-stage, 3 issues, 8 executions, out-of-order 64-bit RISC-V CPU which
-supports 16 cores, runs with 2.5GHz, and is capable of running Linux.
-
-To build platform specific library and firmwares, provide the
-*PLATFORM=thead/c910* parameter to the top level make command.
-
-Platform Options
-----------------
-
-The *T-HEAD C910* platform does not have any platform-specific options.
-
-Building T-HEAD C910 Platform
------------------------------
-
-```
-make PLATFORM=thead/c910
-```
-
-Booting T-HEAD C910 Platform
-----------------------------
-
-**No Payload**
-
-As there's no payload, you may download vmlinux or u-boot to FW_JUMP_ADDR which
-specified in config.mk or compile commands with GDB. And the execution flow will
-turn to vmlinux or u-boot when opensbi ends.
-
-**Linux Kernel Payload**
-
-You can also choose to use Linux kernel as payload by enabling FW_PAYLOAD=y
-along with specifying FW_PAYLOAD_OFFSET. The kernel image will be embedded in
-the OPENSBI firmware binary, T-head will directly boot into Linux after OpenSBI.

docs/platform/thead-c9xx.md

@@ -0,0 +1,198 @@
+T-HEAD C9xx Series Processors
+=============================
+
+The **C9xx** series processors are high-performance RISC-V architecture
+multi-core processors with AI vector acceleration engine.
+
+For more details, refer [T-HEAD.CN](https://www.t-head.cn/)
+
+To build the platform-specific library and firmware images, provide the
+*PLATFORM=generic* parameter to the top level `make` command.
+
+Platform Options
+----------------
+
+The *T-HEAD C9xx* does not have any platform-specific compile options
+because it use generic platform.
+
+```
+CROSS_COMPILE=riscv64-linux-gnu- PLATFORM=generic /usr/bin/make
+```
+
+The *T-HEAD C9xx* DTB provided to OpenSBI generic firmwares will usually have
+"riscv,clint0", "riscv,plic0", "thead,reset-sample" compatible strings.
+
+DTS Example1: (Single core, eg: Allwinner D1 - c906)
+----------------------------------------------------
+
+```
+	cpus {
+		#address-cells = <1>;
+		#size-cells = <0>;
+		timebase-frequency = <3000000>;
+		cpu@0 {
+			device_type = "cpu";
+			reg = <0>;
+			status = "okay";
+			compatible = "riscv";
+			riscv,isa = "rv64imafdcv";
+			mmu-type = "riscv,sv39";
+			cpu0_intc: interrupt-controller {
+				#interrupt-cells = <1>;
+				compatible = "riscv,cpu-intc";
+				interrupt-controller;
+			};
+		};
+	};
+
+	soc {
+		#address-cells = <2>;
+		#size-cells = <2>;
+		compatible = "simple-bus";
+		ranges;
+
+		clint0: clint@14000000 {
+			compatible = "riscv,clint0";
+			interrupts-extended = <
+				&cpu0_intc  3 &cpu0_intc  7
+				>;
+			reg = <0x0 0x14000000 0x0 0x04000000>;
+			clint,has-no-64bit-mmio;
+		};
+
+		intc: interrupt-controller@10000000 {
+			#interrupt-cells = <1>;
+			compatible = "allwinner,sun20i-d1-plic",
+				     "thead,c900-plic";
+			interrupt-controller;
+			interrupts-extended = <
+				&cpu0_intc  0xffffffff &cpu0_intc  9
+				>;
+			reg = <0x0 0x10000000 0x0 0x04000000>;
+			reg-names = "control";
+			riscv,max-priority = <7>;
+			riscv,ndev = <200>;
+		};
+	}
+```
+
+DTS Example2: (Multi cores with soc reset-regs)
+-----------------------------------------------
+
+```
+	cpus {
+		#address-cells = <1>;
+		#size-cells = <0>;
+		timebase-frequency = <3000000>;
+		cpu@0 {
+			device_type = "cpu";
+			reg = <0>;
+			status = "okay";
+			compatible = "riscv";
+			riscv,isa = "rv64imafdc";
+			mmu-type = "riscv,sv39";
+			cpu0_intc: interrupt-controller {
+				#interrupt-cells = <1>;
+				compatible = "riscv,cpu-intc";
+				interrupt-controller;
+			};
+		};
+		cpu@1 {
+			device_type = "cpu";
+			reg = <1>;
+			status = "fail";
+			compatible = "riscv";
+			riscv,isa = "rv64imafdc";
+			mmu-type = "riscv,sv39";
+			cpu1_intc: interrupt-controller {
+				#interrupt-cells = <1>;
+				compatible = "riscv,cpu-intc";
+				interrupt-controller;
+			};
+		};
+		cpu@2 {
+			device_type = "cpu";
+			reg = <2>;
+			status = "fail";
+			compatible = "riscv";
+			riscv,isa = "rv64imafdc";
+			mmu-type = "riscv,sv39";
+			cpu2_intc: interrupt-controller {
+				#interrupt-cells = <1>;
+				compatible = "riscv,cpu-intc";
+				interrupt-controller;
+			};
+		};
+		cpu@3 {
+			device_type = "cpu";
+			reg = <3>;
+			status = "fail";
+			compatible = "riscv";
+			riscv,isa = "rv64imafdc";
+			mmu-type = "riscv,sv39";
+			cpu3_intc: interrupt-controller {
+				#interrupt-cells = <1>;
+				compatible = "riscv,cpu-intc";
+				interrupt-controller;
+			};
+		};
+	};
+
+	soc {
+		#address-cells = <2>;
+		#size-cells = <2>;
+		compatible = "simple-bus";
+		ranges;
+
+		reset: reset-sample {
+			compatible = "thead,reset-sample";
+			entry-reg = <0xff 0xff019050>;
+			entry-cnt = <4>;
+			control-reg = <0xff 0xff015004>;
+			control-val = <0x1c>;
+			csr-copy = <0x7f3 0x7c0 0x7c1 0x7c2 0x7c3 0x7c5 0x7cc>;
+		};
+
+		clint0: clint@ffdc000000 {
+			compatible = "riscv,clint0";
+			interrupts-extended = <
+				&cpu0_intc  3 &cpu0_intc  7
+				&cpu1_intc  3 &cpu1_intc  7
+				&cpu2_intc  3 &cpu2_intc  7
+				&cpu3_intc  3 &cpu3_intc  7
+				&cpu4_intc  3 &cpu4_intc  7
+				>;
+			reg = <0xff 0xdc000000 0x0 0x04000000>;
+			clint,has-no-64bit-mmio;
+		};
+
+		intc: interrupt-controller@ffd8000000 {
+			#interrupt-cells = <1>;
+			compatible = "thead,c900-plic";
+			interrupt-controller;
+			interrupts-extended = <
+				&cpu0_intc  0xffffffff &cpu0_intc  9
+				&cpu1_intc  0xffffffff &cpu1_intc  9
+				&cpu2_intc  0xffffffff &cpu2_intc  9
+				&cpu3_intc  0xffffffff &cpu3_intc  9
+				>;
+			reg = <0xff 0xd8000000 0x0 0x04000000>;
+			reg-names = "control";
+			riscv,max-priority = <7>;
+			riscv,ndev = <80>;
+		};
+	}
+```
+
+DTS Example2: (Multi cores with old reset csrs)
+-----------------------------------------------
+```
+	reset: reset-sample {
+		compatible = "thead,reset-sample";
+		using-csr-reset;
+		csr-copy = <0x7c0 0x7c1 0x7c2 0x7c3 0x7c5 0x7cc
+			    0x3b0 0x3b1 0x3b2 0x3b3
+			    0x3b4 0x3b5 0x3b6 0x3b7
+			    0x3a0>;
+	};
+```

docs/platform_guide.md

@@ -17,7 +17,7 @@ the supported platforms. These firmwares are linked against *libplatsbi.a*.
 Firmware binaries are installed in
 *<install_directory>/platform/<platform_subdir>/bin*. These firmwares can be
 used as executable runtime firmwares on the supported platforms as a replacement
-for the legacy *riskv-pk* boot loader (BBL).
+for the legacy *riscv-pk* boot loader (BBL).
 
 A complete doxygen-style documentation of *struct sbi_platform* and related
 APIs is available in the file *include/sbi/sbi_platform.h*.
@@ -25,18 +25,18 @@ APIs is available in the file *include/sbi/sbi_platform.h*.
 Adding support for a new platform
 ---------------------------------
 
-Support for a new platform named *<xyz>* can be added as follows:
+Support for a new platform named *&lt;xyz&gt;* can be added as follows:
 
-1. Create a directory named *<xyz>* under the *platform/* directory.
+1. Create a directory named *&lt;xyz&gt;* under the *platform/* directory.
 2. Create a platform configuration file named *config.mk* under the
-   *platform/<xyz>/* directory. This configuration file will provide
+   *platform/&lt;xyz&gt;/* directory. This configuration file will provide
    compiler flags, and select firmware options.
-3. Create a *platform/<xyz>/objects.mk* file for listing the
+3. Create a *platform/&lt;xyz&gt;/objects.mk* file for listing the
    platform-specific object files to be compiled.
-4. Create a *platform/<xyz>/platform.c* file providing a *struct sbi_platform*
-   instance.
+4. Create a *platform/&lt;xyz&gt;/platform.c* file providing a
+   *struct sbi_platform* instance.
 
 A platform support code template is available under the *platform/template*
 directory. Copying this directory and its content as a new directory named
-*<xyz>* under the *platform/* directory will create all the files mentioned
-above.
+*&lt;xyz&gt;* under the *platform/* directory will create all the files
+mentioned above.

docs/pmu_support.md

@@ -0,0 +1,87 @@
+OpenSBI SBI PMU extension support
+==================================
+SBI PMU extension supports allow supervisor software to configure/start/stop
+any performance counter at anytime. Thus, an user can leverage full
+capability of performance analysis tools such as perf if SBI PMU extension is
+enabled. The OpenSBI implementation makes the following assumptions about the
+hardware platform.
+
+ * MCOUNTINHIBIT CSR must be implemented in the hardware. Otherwise, SBI PMU
+extension will not be enabled.
+
+ * The platform must provide information about PMU event to counter mapping
+via device tree or platform specific hooks. Otherwise, SBI PMU extension will
+not be enabled.
+
+ * The platforms should provide information about the PMU event selector values
+that should be encoded in the expected value of MHPMEVENTx while configuring
+MHPMCOUNTERx for that specific event. This can be done via a device tree or
+platform specific hooks. The exact value to be written to he MHPMEVENTx is
+completely depends on platform. Generic platform writes the zero-extended event_idx
+as the expected value for hardware cache/generic events as suggested by the SBI
+specification.
+
+SBI PMU Device Tree Bindings
+----------------------------
+
+Platforms may choose to describe PMU event selector and event to counter mapping
+values via device tree. The following sections describes the PMU DT node
+bindings in details.
+
+* **compatible** (Mandatory) - The compatible string of SBI PMU device tree node.
+This DT property must have the value **riscv,pmu**.
+
+* **riscv,event-to-mhpmevent**(Optional) - It represents an ONE-to-ONE mapping
+between a PMU event and the event selector value that platform expects to be
+written to the MHPMEVENTx CSR for that event. The mapping is encoded in a
+table format where each row represents an event. The first column represent the
+event idx where the 2nd & 3rd column represent the event selector value that
+should be encoded in the expected value to be written in MHPMEVENTx.
+This property shouldn't encode any raw hardware event.
+
+* **riscv,event-to-mhpmcounters**(Optional) - It represents a MANY-to-MANY
+mapping between a range of events and all the MHPMCOUNTERx in a bitmap format
+that can be used to monitor these range of events. The information is encoded in
+a table format where each row represent a certain range of events and
+corresponding counters. The first column represents starting of the pmu event id
+and 2nd column represents the end of the pmu event id. The third column
+represent a bitmap of all the MHPMCOUNTERx. This property is mandatory if
+event-to-mhpmevent is present. Otherwise, it can be omitted. This property
+shouldn't encode any raw event.
+
+* **riscv,raw-event-to-mhpmcounters**(Optional) - It represents an ONE-to-MANY
+or MANY-to-MANY mapping between the raw event(s) and all the MHPMCOUNTERx in
+a bitmap format that can be used to monitor that raw event, which depends on
+how the platform encodes the monitor events. Currently, only the following three
+encoding methods are supported, encoding each event as a number, using a bitmap
+to encode monitor events, and mixing the previous two methods. The information
+is encoded in a table format where each row represent the specific raw event(s).
+The first column represents a 64-bit selector value which can indicate an
+monitor event ID (encoded by a number) or an event set (encoded by a bitmap).
+In case of the latter, the lower bits used to encode a set of events should be
+set to zero. The second column is a 64-bit selector mask where any bits used
+for event encoding will be cleared. If a platform directly encodes each raw PMU
+event as a unique ID, the value of select_mask will be 0xffffffff_ffffffff.
+The third column represent a bitmap of all the MHPMCOUNTERx that can be used for
+monitoring the specified event(s).
+
+*Note:* A platform may choose to provide the mapping between event & counters
+via platform hooks rather than the device tree.
+
+### Example
+
+```
+pmu {
+	compatible 			= "riscv,pmu";
+	interrupts 			= <0x100>;
+	interrupt-parent 			= <&plic>
+	riscv,event-to-mhpmevent 		= <0x0000B  0x0000 0x0001>,
+	riscv,event-to-mhpmcounters 	= <0x00001 0x00001 0x00000001>,
+						  <0x00002 0x00002 0x00000004>,
+						  <0x00003 0x0000A 0x00000ff8>,
+						  <0x10000 0x10033 0x000ff000>,
+	riscv,raw-event-to-mhpmcounters = <0x0000     0x0002 0xffffffff 0xffffffff 0x00000f8>,
+					  <0xffffffff 0xfffffff0 0xffffffff 0xfffffff0 0x00000ff0>,
+};
+
+```

firmware/fw_base.S

@@ -9,6 +9,7 @@
 
 #include <sbi/riscv_asm.h>
 #include <sbi/riscv_encoding.h>
+#include <sbi/riscv_elf.h>
 #include <sbi/sbi_platform.h>
 #include <sbi/sbi_scratch.h>
 #include <sbi/sbi_trap.h>
@@ -57,39 +58,91 @@ _start:
 	bne	a0, a6, _wait_relocate_copy_done
 _try_lottery:
 	/* Jump to relocation wait loop if we don't get relocation lottery */
-	la	a6, _relocate_lottery
+	lla	a6, _relocate_lottery
 	li	a7, 1
 	amoadd.w a6, a7, (a6)
 	bnez	a6, _wait_relocate_copy_done
 
 	/* Save load address */
-	la	t0, _load_start
-	la	t1, _start
+	lla	t0, _load_start
+	lla	t1, _fw_start
 	REG_S	t1, 0(t0)
 
+#ifdef FW_PIC
+	/* relocate the global table content */
+	lla	t0, _link_start
+	REG_L	t0, 0(t0)
+	/* t1 shall has the address of _fw_start */
+	sub	t2, t1, t0
+	lla	t3, _runtime_offset
+	REG_S	t2, (t3)
+	lla	t0, __rel_dyn_start
+	lla	t1, __rel_dyn_end
+	beq	t0, t1, _relocate_done
+	j	5f
+2:
+	REG_L	t5, -(REGBYTES*2)(t0)	/* t5 <-- relocation info:type */
+	li	t3, R_RISCV_RELATIVE	/* reloc type R_RISCV_RELATIVE */
+	bne	t5, t3, 3f
+	REG_L	t3, -(REGBYTES*3)(t0)
+	REG_L	t5, -(REGBYTES)(t0)	/* t5 <-- addend */
+	add	t5, t5, t2
+	add	t3, t3, t2
+	REG_S	t5, 0(t3)		/* store runtime address to the GOT entry */
+	j	5f
+
+3:
+	lla	t4, __dyn_sym_start
+
+4:
+	REG_L	t5, -(REGBYTES*2)(t0)	/* t5 <-- relocation info:type */
+	srli	t6, t5, SYM_INDEX	/* t6 <--- sym table index */
+	andi	t5, t5, 0xFF		/* t5 <--- relocation type */
+	li	t3, RELOC_TYPE
+	bne	t5, t3, 5f
+
+	/* address R_RISCV_64 or R_RISCV_32 cases*/
+	REG_L	t3, -(REGBYTES*3)(t0)
+	li	t5, SYM_SIZE
+	mul	t6, t6, t5
+	add	s5, t4, t6
+	REG_L	t6, -(REGBYTES)(t0)	/* t0 <-- addend */
+	REG_L	t5, REGBYTES(s5)
+	add	t5, t5, t6
+	add	t5, t5, t2		/* t5 <-- location to fix up in RAM */
+	add	t3, t3, t2		/* t3 <-- location to fix up in RAM */
+	REG_S	t5, 0(t3)		/* store runtime address to the variable */
+
+5:
+	addi	t0, t0, (REGBYTES*3)
+	ble	t0, t1, 2b
+	j	_relocate_done
+_wait_relocate_copy_done:
+	j	_wait_for_boot_hart
+#else
 	/* Relocate if load address != link address */
 _relocate:
-	la	t0, _link_start
+	lla	t0, _link_start
 	REG_L	t0, 0(t0)
-	la	t1, _link_end
+	lla	t1, _link_end
 	REG_L	t1, 0(t1)
-	la	t2, _load_start
+	lla	t2, _load_start
 	REG_L	t2, 0(t2)
 	sub	t3, t1, t0
 	add	t3, t3, t2
 	beq	t0, t2, _relocate_done
-	la	t4, _relocate_done
+	lla	t4, _relocate_done
 	sub	t4, t4, t2
 	add	t4, t4, t0
 	blt	t2, t0, _relocate_copy_to_upper
 _relocate_copy_to_lower:
 	ble	t1, t2, _relocate_copy_to_lower_loop
-	la	t3, _relocate_lottery
+	lla	t3, _relocate_lottery
 	BRANGE	t2, t1, t3, _start_hang
-	la	t3, _boot_status
+	lla	t3, _boot_status
 	BRANGE	t2, t1, t3, _start_hang
-	la	t3, _relocate
-	la	t5, _relocate_done
+	lla	t3, _relocate
+	lla	t5, _relocate_done
 	BRANGE	t2, t1, t3, _start_hang
 	BRANGE	t2, t1, t5, _start_hang
 	BRANGE  t3, t5, t2, _start_hang
@@ -102,12 +155,12 @@ _relocate_copy_to_lower_loop:
 	jr	t4
 _relocate_copy_to_upper:
 	ble	t3, t0, _relocate_copy_to_upper_loop
-	la	t2, _relocate_lottery
+	lla	t2, _relocate_lottery
 	BRANGE	t0, t3, t2, _start_hang
-	la	t2, _boot_status
+	lla	t2, _boot_status
 	BRANGE	t0, t3, t2, _start_hang
-	la	t2, _relocate
-	la	t5, _relocate_done
+	lla	t2, _relocate
+	lla	t5, _relocate_done
 	BRANGE	t0, t3, t2, _start_hang
 	BRANGE	t0, t3, t5, _start_hang
 	BRANGE	t2, t5, t0, _start_hang
@@ -119,12 +172,12 @@ _relocate_copy_to_upper_loop:
 	blt	t0, t1, _relocate_copy_to_upper_loop
 	jr	t4
 _wait_relocate_copy_done:
-	la	t0, _start
-	la	t1, _link_start
+	lla	t0, _fw_start
+	lla	t1, _link_start
 	REG_L	t1, 0(t1)
 	beq	t0, t1, _wait_for_boot_hart
-	la	t2, _boot_status
-	la	t3, _wait_for_boot_hart
+	lla	t2, _boot_status
+	lla	t3, _wait_for_boot_hart
 	sub	t3, t3, t0
 	add	t3, t3, t1
 1:
@@ -137,19 +190,22 @@ _wait_relocate_copy_done:
 	nop
 	bgt     t4, t5, 1b
 	jr	t3
+#endif
 _relocate_done:
 
 	/*
 	 * Mark relocate copy done
 	 * Use _boot_status copy relative to the load address
 	 */
-	la	t0, _boot_status
-	la	t1, _link_start
+	lla	t0, _boot_status
+#ifndef FW_PIC
+	lla	t1, _link_start
 	REG_L	t1, 0(t1)
-	la	t2, _load_start
+	lla	t2, _load_start
 	REG_L	t2, 0(t2)
 	sub	t0, t0, t1
 	add	t0, t0, t2
+#endif
 	li	t1, BOOT_STATUS_RELOCATE_DONE
 	REG_S	t1, 0(t0)
 	fence	rw, rw
@@ -161,19 +217,19 @@ _relocate_done:
 	call	_reset_regs
 
 	/* Zero-out BSS */
-	la	s4, _bss_start
-	la	s5, _bss_end
+	lla	s4, _bss_start
+	lla	s5, _bss_end
 _bss_zero:
 	REG_S	zero, (s4)
 	add	s4, s4, __SIZEOF_POINTER__
 	blt	s4, s5, _bss_zero
 
 	/* Setup temporary trap handler */
-	la	s4, _start_hang
+	lla	s4, _start_hang
 	csrw	CSR_MTVEC, s4
 
 	/* Setup temporary stack */
-	la	s4, _fw_end
+	lla	s4, _fw_end
 	li	s5, (SBI_SCRATCH_SIZE * 2)
 	add	sp, s4, s5
 
@@ -184,7 +240,7 @@ _bss_zero:
 
 #ifdef FW_FDT_PATH
 	/* Override previous arg1 */
-	la	a1, fw_fdt_bin
+	lla	a1, fw_fdt_bin
 #endif
 
 	/*
@@ -202,7 +258,7 @@ _bss_zero:
 	 * s7 -> HART Count
 	 * s8 -> HART Stack Size
 	 */
-	la	a4, platform
+	lla	a4, platform
 #if __riscv_xlen == 64
 	lwu	s7, SBI_PLATFORM_HART_COUNT_OFFSET(a4)
 	lwu	s8, SBI_PLATFORM_HART_STACK_SIZE_OFFSET(a4)
@@ -212,7 +268,7 @@ _bss_zero:
 #endif
 
 	/* Setup scratch space for all the HARTs*/
-	la	tp, _fw_end
+	lla	tp, _fw_end
 	mul	a5, s7, s8
 	add	tp, tp, a5
 	/* Keep a copy of tp */
@@ -222,6 +278,14 @@ _bss_zero:
 	/* hartid 0 is mandated by ISA */
 	li	t1, 0
 _scratch_init:
+	/*
+	 * The following registers hold values that are computed before
+	 * entering this block, and should remain unchanged.
+	 *
+	 * t3 -> the firmware end address
+	 * s7 -> HART count
+	 * s8 -> HART stack size
+	 */
 	add	tp, t3, zero
 	mul	a5, s8, t1
 	sub	tp, tp, a5
@@ -230,11 +294,8 @@ _scratch_init:
 
 	/* Initialize scratch space */
 	/* Store fw_start and fw_size in scratch space */
-	la	a4, _fw_start
-	la	a5, _fw_end
-	mul	t0, s7, s8
-	add	a5, a5, t0
-	sub	a5, a5, a4
+	lla	a4, _fw_start
+	sub	a5, t3, a4
 	REG_S	a4, SBI_SCRATCH_FW_START_OFFSET(tp)
 	REG_S	a5, SBI_SCRATCH_FW_SIZE_OFFSET(tp)
 	/* Store next arg1 in scratch space */
@@ -253,16 +314,16 @@ _scratch_init:
 	REG_S	a0, SBI_SCRATCH_NEXT_MODE_OFFSET(tp)
 	MOV_3R	a0, s0, a1, s1, a2, s2
 	/* Store warm_boot address in scratch space */
-	la	a4, _start_warm
+	lla	a4, _start_warm
 	REG_S	a4, SBI_SCRATCH_WARMBOOT_ADDR_OFFSET(tp)
 	/* Store platform address in scratch space */
-	la	a4, platform
+	lla	a4, platform
 	REG_S	a4, SBI_SCRATCH_PLATFORM_ADDR_OFFSET(tp)
 	/* Store hartid-to-scratch function address in scratch space */
-	la	a4, _hartid_to_scratch
+	lla	a4, _hartid_to_scratch
 	REG_S	a4, SBI_SCRATCH_HARTID_TO_SCRATCH_OFFSET(tp)
 	/* Store trap-exit function address in scratch space */
-	la	a4, _trap_exit
+	lla	a4, _trap_exit
 	REG_S	a4, SBI_SCRATCH_TRAP_EXIT_OFFSET(tp)
 	/* Clear tmp0 in scratch space */
 	REG_S	zero, SBI_SCRATCH_TMP0_OFFSET(tp)
@@ -288,8 +349,7 @@ _scratch_init:
 	 * previous booting stage.
 	 */
 	beqz	a1, _fdt_reloc_done
-	/* Mask values in a3 and a4 */
-	li	a3, ~(__SIZEOF_POINTER__ - 1)
+	/* Mask values in a4 */
 	li	a4, 0xff
 	/* t1 = destination FDT start address */
 	MOV_3R	s0, a0, s1, a1, s2, a2
@@ -298,10 +358,8 @@ _scratch_init:
 	MOV_3R	a0, s0, a1, s1, a2, s2
 	beqz	t1, _fdt_reloc_done
 	beq	t1, a1, _fdt_reloc_done
-	and	t1, t1, a3
 	/* t0 = source FDT start address */
 	add	t0, a1, zero
-	and	t0, t0, a3
 	/* t2 = source FDT size in big-endian */
 #if __riscv_xlen == 64
 	lwu	t2, 4(t0)
@@ -343,7 +401,7 @@ _fdt_reloc_done:
 
 	/* mark boot hart done */
 	li	t0, BOOT_STATUS_BOOT_HART_DONE
-	la	t1, _boot_status
+	lla	t1, _boot_status
 	REG_S	t0, 0(t1)
 	fence	rw, rw
 	j	_start_warm
@@ -351,7 +409,7 @@ _fdt_reloc_done:
 	/* waiting for boot hart to be done (_boot_status == 2) */
 _wait_for_boot_hart:
 	li	t0, BOOT_STATUS_BOOT_HART_DONE
-	la	t1, _boot_status
+	lla	t1, _boot_status
 	REG_L	t1, 0(t1)
 	/* Reduce the bus traffic so that boot hart may proceed faster */
 	nop
@@ -369,7 +427,7 @@ _start_warm:
 	csrw	CSR_MIP, zero
 
 	/* Find HART count and HART stack size */
-	la	a4, platform
+	lla	a4, platform
 #if __riscv_xlen == 64
 	lwu	s7, SBI_PLATFORM_HART_COUNT_OFFSET(a4)
 	lwu	s8, SBI_PLATFORM_HART_STACK_SIZE_OFFSET(a4)
@@ -400,7 +458,7 @@ _start_warm:
 3:	bge	s6, s7, _start_hang
 
 	/* Find the scratch space based on HART index */
-	la	tp, _fw_end
+	lla	tp, _fw_end
 	mul	a5, s7, s8
 	add	tp, tp, a5
 	mul	a5, s8, s6
@@ -415,13 +473,13 @@ _start_warm:
 	add	sp, tp, zero
 
 	/* Setup trap handler */
-	la	a4, _trap_handler
+	lla	a4, _trap_handler
 #if __riscv_xlen == 32
 	csrr	a5, CSR_MISA
 	srli	a5, a5, ('H' - 'A')
 	andi	a5, a5, 0x1
 	beq	a5, zero, _skip_trap_handler_rv32_hyp
-	la	a4, _trap_handler_rv32_hyp
+	lla	a4, _trap_handler_rv32_hyp
 _skip_trap_handler_rv32_hyp:
 #endif
 	csrw	CSR_MTVEC, a4
@@ -432,7 +490,7 @@ _skip_trap_handler_rv32_hyp:
 	srli	a5, a5, ('H' - 'A')
 	andi	a5, a5, 0x1
 	beq	a5, zero, _skip_trap_exit_rv32_hyp
-	la	a4, _trap_exit_rv32_hyp
+	lla	a4, _trap_exit_rv32_hyp
 	csrr	a5, CSR_MSCRATCH
 	REG_S	a4, SBI_SCRATCH_TRAP_EXIT_OFFSET(a5)
 _skip_trap_exit_rv32_hyp:
@@ -445,7 +503,12 @@ _skip_trap_exit_rv32_hyp:
 	/* We don't expect to reach here hence just hang */
 	j	_start_hang
 
+	.data
 	.align 3
+#ifdef FW_PIC
+_runtime_offset:
+	RISCV_PTR	0
+#endif
 _relocate_lottery:
 	RISCV_PTR	0
 _boot_status:
@@ -453,7 +516,7 @@ _boot_status:
 _load_start:
 	RISCV_PTR	_fw_start
 _link_start:
-	RISCV_PTR	_fw_start
+	RISCV_PTR	FW_TEXT_START
 _link_end:
 	RISCV_PTR	_fw_reloc_end
 
@@ -468,7 +531,7 @@ _hartid_to_scratch:
 	 * t1 -> HART Stack End
 	 * t2 -> Temporary
 	 */
-	la	t2, platform
+	lla	t2, platform
 #if __riscv_xlen == 64
 	lwu	t0, SBI_PLATFORM_HART_STACK_SIZE_OFFSET(t2)
 	lwu	t2, SBI_PLATFORM_HART_COUNT_OFFSET(t2)
@@ -478,7 +541,7 @@ _hartid_to_scratch:
 #endif
 	sub	t2, t2, a1
 	mul	t2, t2, t0
-	la	t1, _fw_end
+	lla	t1, _fw_end
 	add	t1, t1, t2
 	li	t2, SBI_SCRATCH_SIZE
 	sub	a0, t1, t2
@@ -493,12 +556,39 @@ _start_hang:
 
 	.section .entry, "ax", %progbits
 	.align 3
-	.globl fw_platform_init
 	.weak fw_platform_init
 fw_platform_init:
 	add	a0, a1, zero
 	ret
 
+	/* Map implicit memcpy() added by compiler to sbi_memcpy() */
+	.section .text
+	.align 3
+	.globl memcpy
+memcpy:
+	tail	sbi_memcpy
+
+	/* Map implicit memset() added by compiler to sbi_memset() */
+	.section .text
+	.align 3
+	.globl memset
+memset:
+	tail	sbi_memset
+
+	/* Map implicit memmove() added by compiler to sbi_memmove() */
+	.section .text
+	.align 3
+	.globl memmove
+memmove:
+	tail	sbi_memmove
+
+	/* Map implicit memcmp() added by compiler to sbi_memcmp() */
+	.section .text
+	.align 3
+	.globl memcmp
+memcmp:
+	tail	sbi_memcmp
+
 .macro	TRAP_SAVE_AND_SETUP_SP_T0
 	/* Swap TP and MSCRATCH */
 	csrrw	tp, CSR_MSCRATCH, tp
@@ -595,62 +685,63 @@ fw_platform_init:
 	call	sbi_trap_handler
 .endm
 
-.macro	TRAP_RESTORE_GENERAL_REGS_EXCEPT_SP_T0
-	/* Restore all general regisers except SP and T0 */
-	REG_L	ra, SBI_TRAP_REGS_OFFSET(ra)(sp)
-	REG_L	gp, SBI_TRAP_REGS_OFFSET(gp)(sp)
-	REG_L	tp, SBI_TRAP_REGS_OFFSET(tp)(sp)
-	REG_L	t1, SBI_TRAP_REGS_OFFSET(t1)(sp)
-	REG_L	t2, SBI_TRAP_REGS_OFFSET(t2)(sp)
-	REG_L	s0, SBI_TRAP_REGS_OFFSET(s0)(sp)
-	REG_L	s1, SBI_TRAP_REGS_OFFSET(s1)(sp)
-	REG_L	a0, SBI_TRAP_REGS_OFFSET(a0)(sp)
-	REG_L	a1, SBI_TRAP_REGS_OFFSET(a1)(sp)
-	REG_L	a2, SBI_TRAP_REGS_OFFSET(a2)(sp)
-	REG_L	a3, SBI_TRAP_REGS_OFFSET(a3)(sp)
-	REG_L	a4, SBI_TRAP_REGS_OFFSET(a4)(sp)
-	REG_L	a5, SBI_TRAP_REGS_OFFSET(a5)(sp)
-	REG_L	a6, SBI_TRAP_REGS_OFFSET(a6)(sp)
-	REG_L	a7, SBI_TRAP_REGS_OFFSET(a7)(sp)
-	REG_L	s2, SBI_TRAP_REGS_OFFSET(s2)(sp)
-	REG_L	s3, SBI_TRAP_REGS_OFFSET(s3)(sp)
-	REG_L	s4, SBI_TRAP_REGS_OFFSET(s4)(sp)
-	REG_L	s5, SBI_TRAP_REGS_OFFSET(s5)(sp)
-	REG_L	s6, SBI_TRAP_REGS_OFFSET(s6)(sp)
-	REG_L	s7, SBI_TRAP_REGS_OFFSET(s7)(sp)
-	REG_L	s8, SBI_TRAP_REGS_OFFSET(s8)(sp)
-	REG_L	s9, SBI_TRAP_REGS_OFFSET(s9)(sp)
-	REG_L	s10, SBI_TRAP_REGS_OFFSET(s10)(sp)
-	REG_L	s11, SBI_TRAP_REGS_OFFSET(s11)(sp)
-	REG_L	t3, SBI_TRAP_REGS_OFFSET(t3)(sp)
-	REG_L	t4, SBI_TRAP_REGS_OFFSET(t4)(sp)
-	REG_L	t5, SBI_TRAP_REGS_OFFSET(t5)(sp)
-	REG_L	t6, SBI_TRAP_REGS_OFFSET(t6)(sp)
+.macro	TRAP_RESTORE_GENERAL_REGS_EXCEPT_A0_T0
+	/* Restore all general regisers except A0 and T0 */
+	REG_L	ra, SBI_TRAP_REGS_OFFSET(ra)(a0)
+	REG_L	sp, SBI_TRAP_REGS_OFFSET(sp)(a0)
+	REG_L	gp, SBI_TRAP_REGS_OFFSET(gp)(a0)
+	REG_L	tp, SBI_TRAP_REGS_OFFSET(tp)(a0)
+	REG_L	t1, SBI_TRAP_REGS_OFFSET(t1)(a0)
+	REG_L	t2, SBI_TRAP_REGS_OFFSET(t2)(a0)
+	REG_L	s0, SBI_TRAP_REGS_OFFSET(s0)(a0)
+	REG_L	s1, SBI_TRAP_REGS_OFFSET(s1)(a0)
+	REG_L	a1, SBI_TRAP_REGS_OFFSET(a1)(a0)
+	REG_L	a2, SBI_TRAP_REGS_OFFSET(a2)(a0)
+	REG_L	a3, SBI_TRAP_REGS_OFFSET(a3)(a0)
+	REG_L	a4, SBI_TRAP_REGS_OFFSET(a4)(a0)
+	REG_L	a5, SBI_TRAP_REGS_OFFSET(a5)(a0)
+	REG_L	a6, SBI_TRAP_REGS_OFFSET(a6)(a0)
+	REG_L	a7, SBI_TRAP_REGS_OFFSET(a7)(a0)
+	REG_L	s2, SBI_TRAP_REGS_OFFSET(s2)(a0)
+	REG_L	s3, SBI_TRAP_REGS_OFFSET(s3)(a0)
+	REG_L	s4, SBI_TRAP_REGS_OFFSET(s4)(a0)
+	REG_L	s5, SBI_TRAP_REGS_OFFSET(s5)(a0)
+	REG_L	s6, SBI_TRAP_REGS_OFFSET(s6)(a0)
+	REG_L	s7, SBI_TRAP_REGS_OFFSET(s7)(a0)
+	REG_L	s8, SBI_TRAP_REGS_OFFSET(s8)(a0)
+	REG_L	s9, SBI_TRAP_REGS_OFFSET(s9)(a0)
+	REG_L	s10, SBI_TRAP_REGS_OFFSET(s10)(a0)
+	REG_L	s11, SBI_TRAP_REGS_OFFSET(s11)(a0)
+	REG_L	t3, SBI_TRAP_REGS_OFFSET(t3)(a0)
+	REG_L	t4, SBI_TRAP_REGS_OFFSET(t4)(a0)
+	REG_L	t5, SBI_TRAP_REGS_OFFSET(t5)(a0)
+	REG_L	t6, SBI_TRAP_REGS_OFFSET(t6)(a0)
 .endm
 
 .macro	TRAP_RESTORE_MEPC_MSTATUS have_mstatush
 	/* Restore MEPC and MSTATUS CSRs */
-	REG_L	t0, SBI_TRAP_REGS_OFFSET(mepc)(sp)
+	REG_L	t0, SBI_TRAP_REGS_OFFSET(mepc)(a0)
 	csrw	CSR_MEPC, t0
-	REG_L	t0, SBI_TRAP_REGS_OFFSET(mstatus)(sp)
+	REG_L	t0, SBI_TRAP_REGS_OFFSET(mstatus)(a0)
 	csrw	CSR_MSTATUS, t0
 	.if \have_mstatush
-	REG_L	t0, SBI_TRAP_REGS_OFFSET(mstatusH)(sp)
+	REG_L	t0, SBI_TRAP_REGS_OFFSET(mstatusH)(a0)
 	csrw	CSR_MSTATUSH, t0
 	.endif
 .endm
 
-.macro TRAP_RESTORE_SP_T0
+.macro TRAP_RESTORE_A0_T0
 	/* Restore T0 */
-	REG_L	t0, SBI_TRAP_REGS_OFFSET(t0)(sp)
+	REG_L	t0, SBI_TRAP_REGS_OFFSET(t0)(a0)
 
-	/* Restore SP */
-	REG_L	sp, SBI_TRAP_REGS_OFFSET(sp)(sp)
+	/* Restore A0 */
+	REG_L	a0, SBI_TRAP_REGS_OFFSET(a0)(a0)
 .endm
 
 	.section .entry, "ax", %progbits
 	.align 3
 	.globl _trap_handler
+	.globl _trap_exit
 _trap_handler:
 	TRAP_SAVE_AND_SETUP_SP_T0
 
@@ -660,25 +751,12 @@ _trap_handler:
 
 	TRAP_CALL_C_ROUTINE
 
-	TRAP_RESTORE_GENERAL_REGS_EXCEPT_SP_T0
-
-	TRAP_RESTORE_MEPC_MSTATUS 0
-
-	TRAP_RESTORE_SP_T0
-
-	mret
-
-	.section .entry, "ax", %progbits
-	.align 3
-	.globl _trap_exit
 _trap_exit:
-	add	sp, a0, zero
-
-	TRAP_RESTORE_GENERAL_REGS_EXCEPT_SP_T0
+	TRAP_RESTORE_GENERAL_REGS_EXCEPT_A0_T0
 
 	TRAP_RESTORE_MEPC_MSTATUS 0
 
-	TRAP_RESTORE_SP_T0
+	TRAP_RESTORE_A0_T0
 
 	mret
 
@@ -686,6 +764,7 @@ _trap_exit:
 	.section .entry, "ax", %progbits
 	.align 3
 	.globl _trap_handler_rv32_hyp
+	.globl _trap_exit_rv32_hyp
 _trap_handler_rv32_hyp:
 	TRAP_SAVE_AND_SETUP_SP_T0
 
@@ -695,25 +774,12 @@ _trap_handler_rv32_hyp:
 
 	TRAP_CALL_C_ROUTINE
 
-	TRAP_RESTORE_GENERAL_REGS_EXCEPT_SP_T0
-
-	TRAP_RESTORE_MEPC_MSTATUS 1
-
-	TRAP_RESTORE_SP_T0
-
-	mret
-
-	.section .entry, "ax", %progbits
-	.align 3
-	.globl _trap_exit_rv32_hyp
 _trap_exit_rv32_hyp:
-	add	sp, a0, zero
-
-	TRAP_RESTORE_GENERAL_REGS_EXCEPT_SP_T0
+	TRAP_RESTORE_GENERAL_REGS_EXCEPT_A0_T0
 
 	TRAP_RESTORE_MEPC_MSTATUS 1
 
-	TRAP_RESTORE_SP_T0
+	TRAP_RESTORE_A0_T0
 
 	mret
 #endif

firmware/fw_base.ldS

@@ -8,7 +8,7 @@
  */
 
 	. = FW_TEXT_START;
-
+	/* Don't add any section between FW_TEXT_START and _fw_start */
 	PROVIDE(_fw_start = .);
 
 	. = ALIGN(0x1000); /* Need this to create proper sections */
@@ -61,6 +61,19 @@
 		PROVIDE(_data_end = .);
 	}
 
+	.dynsym : {
+		PROVIDE(__dyn_sym_start = .);
+		*(.dynsym)
+		PROVIDE(__dyn_sym_end = .);
+	}
+
+	.rela.dyn : {
+		PROVIDE(__rel_dyn_start = .);
+		*(.rela*)
+		. = ALIGN(8);
+		PROVIDE(__rel_dyn_end = .);
+	}
+
 	. = ALIGN(0x1000); /* Ensure next section is page aligned */
 
 	.bss :

firmware/fw_dynamic.S

@@ -36,7 +36,7 @@ fw_boot_hart:
 	bgt	a0, a1, _bad_dynamic_info
 
 	/* Read boot HART id */
-	li	a1, 0x2
+	li	a1, FW_DYNAMIC_INFO_VERSION_2
 	blt	a0, a1, 2f
 	REG_L	a0, FW_DYNAMIC_INFO_BOOT_HART_OFFSET(a2)
 	ret
@@ -54,33 +54,25 @@ fw_boot_hart:
 	 */
 fw_save_info:
 	/* Save next arg1 in 'a1' */
-	la	a4, _dynamic_next_arg1
+	lla	a4, _dynamic_next_arg1
 	REG_S	a1, (a4)
 
-	/* Sanity checks */
-	li	a4, FW_DYNAMIC_INFO_MAGIC_VALUE
-	REG_L	a3, FW_DYNAMIC_INFO_MAGIC_OFFSET(a2)
-	bne	a3, a4, _bad_dynamic_info
-	li	a4, FW_DYNAMIC_INFO_VERSION_MAX
-	REG_L	a3, FW_DYNAMIC_INFO_VERSION_OFFSET(a2)
-	bgt	a3, a4, _bad_dynamic_info
-
 	/* Save version == 0x1 fields */
-	la	a4, _dynamic_next_addr
+	lla	a4, _dynamic_next_addr
 	REG_L	a3, FW_DYNAMIC_INFO_NEXT_ADDR_OFFSET(a2)
 	REG_S	a3, (a4)
-	la	a4, _dynamic_next_mode
+	lla	a4, _dynamic_next_mode
 	REG_L	a3, FW_DYNAMIC_INFO_NEXT_MODE_OFFSET(a2)
 	REG_S	a3, (a4)
-	la	a4, _dynamic_options
+	lla	a4, _dynamic_options
 	REG_L	a3, FW_DYNAMIC_INFO_OPTIONS_OFFSET(a2)
 	REG_S	a3, (a4)
 
 	/* Save version == 0x2 fields */
-	li	a4, 0x2
+	li	a4, FW_DYNAMIC_INFO_VERSION_2
 	REG_L	a3, FW_DYNAMIC_INFO_VERSION_OFFSET(a2)
 	blt	a3, a4, 2f
-	la	a4, _dynamic_boot_hart
+	lla	a4, _dynamic_boot_hart
 	REG_L	a3, FW_DYNAMIC_INFO_BOOT_HART_OFFSET(a2)
 	REG_S	a3, (a4)
 2:
@@ -96,7 +88,7 @@ fw_save_info:
 	 * The next arg1 should be returned in 'a0'.
 	 */
 fw_next_arg1:
-	la	a0, _dynamic_next_arg1
+	lla	a0, _dynamic_next_arg1
 	REG_L	a0, (a0)
 	ret
 
@@ -108,7 +100,7 @@ fw_next_arg1:
 	 * The next address should be returned in 'a0'.
 	 */
 fw_next_addr:
-	la	a0, _dynamic_next_addr
+	lla	a0, _dynamic_next_addr
 	REG_L	a0, (a0)
 	ret
 
@@ -120,7 +112,7 @@ fw_next_addr:
 	 * The next address should be returned in 'a0'
 	 */
 fw_next_mode:
-	la	a0, _dynamic_next_mode
+	lla	a0, _dynamic_next_mode
 	REG_L	a0, (a0)
 	ret
 
@@ -133,7 +125,7 @@ fw_next_mode:
 	 * The next address should be returned in 'a0'.
 	 */
 fw_options:
-	la	a0, _dynamic_options
+	lla	a0, _dynamic_options
 	REG_L	a0, (a0)
 	ret
 

firmware/fw_jump.S

@@ -59,7 +59,7 @@ fw_next_arg1:
 	 * The next address should be returned in 'a0'.
 	 */
 fw_next_addr:
-	la	a0, _jump_addr
+	lla	a0, _jump_addr
 	REG_L	a0, (a0)
 	ret
 

firmware/fw_payload.S

@@ -59,7 +59,7 @@ fw_next_arg1:
 	 * The next address should be returned in 'a0'.
 	 */
 fw_next_addr:
-	la	a0, payload_bin
+	lla	a0, payload_bin
 	ret
 
 	.section .entry, "ax", %progbits

firmware/objects.mk

@@ -13,6 +13,17 @@ firmware-cflags-y +=
 firmware-asflags-y +=
 firmware-ldflags-y +=
 
+ifndef FW_PIC
+FW_PIC := $(OPENSBI_LD_PIE)
+endif
+
+ifeq ($(FW_PIC),y)
+firmware-genflags-y +=	-DFW_PIC
+firmware-asflags-y  +=	-fpic
+firmware-cflags-y   +=	-fPIE -pie
+firmware-ldflags-y  +=	-Wl,--no-dynamic-linker -Wl,-pie
+endif
+
 ifdef FW_TEXT_START
 firmware-genflags-y += -DFW_TEXT_START=$(FW_TEXT_START)
 endif

firmware/payloads/test_head.S

@@ -28,20 +28,20 @@
 	.globl _start
 _start:
 	/* Pick one hart to run the main boot sequence */
-	la	a3, _hart_lottery
+	lla	a3, _hart_lottery
 	li	a2, 1
 	amoadd.w a3, a2, (a3)
 	bnez	a3, _start_hang
 
 	/* Save a0 and a1 */
-	la	a3, _boot_a0
+	lla	a3, _boot_a0
 	REG_S	a0, 0(a3)
-	la	a3, _boot_a1
+	lla	a3, _boot_a1
 	REG_S	a1, 0(a3)
 
 	/* Zero-out BSS */
-	la	a4, _bss_start
-	la	a5, _bss_end
+	lla	a4, _bss_start
+	lla	a5, _bss_end
 _bss_zero:
 	REG_S	zero, (a4)
 	add	a4, a4, __SIZEOF_POINTER__
@@ -53,18 +53,18 @@ _start_warm:
 	csrw	CSR_SIP, zero
 
 	/* Setup exception vectors */
-	la	a3, _start_hang
+	lla	a3, _start_hang
 	csrw	CSR_STVEC, a3
 
 	/* Setup stack */
-	la	a3, _payload_end
+	lla	a3, _payload_end
 	li	a4, 0x2000
 	add	sp, a3, a4
 
 	/* Jump to C main */
-	la	a3, _boot_a0
+	lla	a3, _boot_a0
 	REG_L	a0, 0(a3)
-	la	a3, _boot_a1
+	lla	a3, _boot_a1
 	REG_L	a1, 0(a3)
 	call	test_main
 

firmware/payloads/test_main.c

@@ -9,24 +9,25 @@
 
 #include <sbi/sbi_ecall_interface.h>
 
-#define SBI_ECALL(__num, __a0, __a1, __a2)                                    \
+#define SBI_ECALL(__eid, __fid, __a0, __a1, __a2)                             \
 	({                                                                    \
 		register unsigned long a0 asm("a0") = (unsigned long)(__a0);  \
 		register unsigned long a1 asm("a1") = (unsigned long)(__a1);  \
 		register unsigned long a2 asm("a2") = (unsigned long)(__a2);  \
-		register unsigned long a7 asm("a7") = (unsigned long)(__num); \
+		register unsigned long a6 asm("a6") = (unsigned long)(__fid); \
+		register unsigned long a7 asm("a7") = (unsigned long)(__eid); \
 		asm volatile("ecall"                                          \
 			     : "+r"(a0)                                       \
-			     : "r"(a1), "r"(a2), "r"(a7)                      \
+			     : "r"(a1), "r"(a2), "r"(a6), "r"(a7)             \
 			     : "memory");                                     \
 		a0;                                                           \
 	})
 
-#define SBI_ECALL_0(__num) SBI_ECALL(__num, 0, 0, 0)
-#define SBI_ECALL_1(__num, __a0) SBI_ECALL(__num, __a0, 0, 0)
-#define SBI_ECALL_2(__num, __a0, __a1) SBI_ECALL(__num, __a0, __a1, 0)
+#define SBI_ECALL_0(__eid, __fid) SBI_ECALL(__eid, __fid, 0, 0, 0)
+#define SBI_ECALL_1(__eid, __fid, __a0) SBI_ECALL(__eid, __fid, __a0, 0, 0)
+#define SBI_ECALL_2(__eid, __fid, __a0, __a1) SBI_ECALL(__eid, __fid, __a0, __a1, 0)
 
-#define sbi_ecall_console_putc(c) SBI_ECALL_1(SBI_EXT_0_1_CONSOLE_PUTCHAR, (c))
+#define sbi_ecall_console_putc(c) SBI_ECALL_1(SBI_EXT_0_1_CONSOLE_PUTCHAR, 0, (c))
 
 static inline void sbi_ecall_console_puts(const char *str)
 {

include/sbi/fw_dynamic.h

@@ -15,23 +15,24 @@
 /* clang-format off */
 
 /** Offset of magic member in fw_dynamic_info */
-#define FW_DYNAMIC_INFO_MAGIC_OFFSET		(0 * __SIZEOF_POINTER__)
+#define FW_DYNAMIC_INFO_MAGIC_OFFSET		(0 * __SIZEOF_LONG__)
 /** Offset of version member in fw_dynamic_info */
-#define FW_DYNAMIC_INFO_VERSION_OFFSET		(1 * __SIZEOF_POINTER__)
+#define FW_DYNAMIC_INFO_VERSION_OFFSET		(1 * __SIZEOF_LONG__)
 /** Offset of next_addr member in fw_dynamic_info (version >= 1) */
-#define FW_DYNAMIC_INFO_NEXT_ADDR_OFFSET	(2 * __SIZEOF_POINTER__)
+#define FW_DYNAMIC_INFO_NEXT_ADDR_OFFSET	(2 * __SIZEOF_LONG__)
 /** Offset of next_mode member in fw_dynamic_info  (version >= 1) */
-#define FW_DYNAMIC_INFO_NEXT_MODE_OFFSET	(3 * __SIZEOF_POINTER__)
+#define FW_DYNAMIC_INFO_NEXT_MODE_OFFSET	(3 * __SIZEOF_LONG__)
 /** Offset of options member in fw_dynamic_info  (version >= 1) */
-#define FW_DYNAMIC_INFO_OPTIONS_OFFSET		(4 * __SIZEOF_POINTER__)
+#define FW_DYNAMIC_INFO_OPTIONS_OFFSET		(4 * __SIZEOF_LONG__)
 /** Offset of boot_hart member in fw_dynamic_info  (version >= 2) */
-#define FW_DYNAMIC_INFO_BOOT_HART_OFFSET	(5 * __SIZEOF_POINTER__)
+#define FW_DYNAMIC_INFO_BOOT_HART_OFFSET	(5 * __SIZEOF_LONG__)
 
 /** Expected value of info magic ('OSBI' ascii string in hex) */
 #define FW_DYNAMIC_INFO_MAGIC_VALUE		0x4942534f
 
 /** Maximum supported info version */
-#define FW_DYNAMIC_INFO_VERSION_MAX		0x2
+#define FW_DYNAMIC_INFO_VERSION_2		0x2
+#define FW_DYNAMIC_INFO_VERSION_MAX		FW_DYNAMIC_INFO_VERSION_2
 
 /** Possible next mode values */
 #define FW_DYNAMIC_INFO_NEXT_MODE_U		0x0
@@ -40,7 +41,7 @@
 
 /* clang-format on */
 
-#ifndef __ASSEMBLY__
+#ifndef __ASSEMBLER__
 
 #include <sbi/sbi_types.h>
 

include/sbi/riscv_asm.h

@@ -14,7 +14,7 @@
 
 /* clang-format off */
 
-#ifdef __ASSEMBLY__
+#ifdef __ASSEMBLER__
 #define __ASM_STR(x)	x
 #else
 #define __ASM_STR(x)	#x
@@ -38,7 +38,7 @@
 #define LGREG		__REG_SEL(3, 2)
 
 #if __SIZEOF_POINTER__ == 8
-#ifdef __ASSEMBLY__
+#ifdef __ASSEMBLER__
 #define RISCV_PTR		.dword
 #define RISCV_SZPTR		8
 #define RISCV_LGPTR		3
@@ -48,7 +48,7 @@
 #define RISCV_LGPTR		"3"
 #endif
 #elif __SIZEOF_POINTER__ == 4
-#ifdef __ASSEMBLY__
+#ifdef __ASSEMBLER__
 #define RISCV_PTR		.word
 #define RISCV_SZPTR		4
 #define RISCV_LGPTR		2
@@ -79,7 +79,7 @@
 
 /* clang-format on */
 
-#ifndef __ASSEMBLY__
+#ifndef __ASSEMBLER__
 
 #define csr_swap(csr, val)                                              \
 	({                                                              \
@@ -157,6 +157,11 @@ void csr_write_num(int csr_num, unsigned long val);
 		__asm__ __volatile__("wfi" ::: "memory"); \
 	} while (0)
 
+#define ebreak()                                             \
+	do {                                              \
+		__asm__ __volatile__("ebreak" ::: "memory"); \
+	} while (0)
+
 /* Get current HART id */
 #define current_hartid()	((unsigned int)csr_read(CSR_MHARTID))
 
@@ -182,6 +187,6 @@ int pmp_set(unsigned int n, unsigned long prot, unsigned long addr,
 int pmp_get(unsigned int n, unsigned long *prot_out, unsigned long *addr_out,
 	    unsigned long *log2len);
 
-#endif /* !__ASSEMBLY__ */
+#endif /* !__ASSEMBLER__ */
 
 #endif

include/sbi/riscv_barrier.h

@@ -18,6 +18,9 @@
 #define RISCV_FENCE(p, s) \
 	__asm__ __volatile__ ("fence " #p "," #s : : : "memory")
 
+#define RISCV_FENCE_I \
+	__asm__ __volatile__ ("fence.i" : : : "memory")
+
 /* Read & Write Memory barrier */
 #define mb()			RISCV_FENCE(iorw,iorw)
 

include/sbi/riscv_elf.h

@@ -0,0 +1,14 @@
+#ifndef __RISCV_ELF_H__
+#define __RISCV_ELF_H__
+
+#include <sbi/riscv_asm.h>
+
+#define R_RISCV_32		1
+#define R_RISCV_64		2
+#define R_RISCV_RELATIVE	3
+
+#define RELOC_TYPE		__REG_SEL(R_RISCV_64, R_RISCV_32)
+#define SYM_INDEX		__REG_SEL(0x20,	0x8)
+#define SYM_SIZE		__REG_SEL(0x18,0x10)
+
+#endif

include/sbi/riscv_encoding.h

@@ -86,6 +86,7 @@
 #define IRQ_VS_EXT			10
 #define IRQ_M_EXT			11
 #define IRQ_S_GEXT			12
+#define IRQ_PMU_OVF			13
 
 #define MIP_SSIP			(_UL(1) << IRQ_S_SOFT)
 #define MIP_VSSIP			(_UL(1) << IRQ_VS_SOFT)
@@ -97,6 +98,7 @@
 #define MIP_VSEIP			(_UL(1) << IRQ_VS_EXT)
 #define MIP_MEIP			(_UL(1) << IRQ_M_EXT)
 #define MIP_SGEIP			(_UL(1) << IRQ_S_GEXT)
+#define MIP_LCOFIP			(_UL(1) << IRQ_PMU_OVF)
 
 #define SIP_SSIP			MIP_SSIP
 #define SIP_STIP			MIP_STIP
@@ -171,6 +173,31 @@
 #define HGATP_MODE_SHIFT		HGATP32_MODE_SHIFT
 #endif
 
+#if __riscv_xlen == 64
+#define MHPMEVENT_OF			(_UL(1) << 63)
+#define MHPMEVENT_MINH			(_UL(1) << 62)
+#define MHPMEVENT_SINH			(_UL(1) << 61)
+#define MHPMEVENT_UINH			(_UL(1) << 60)
+#define MHPMEVENT_VSINH			(_UL(1) << 59)
+#define MHPMEVENT_VUINH			(_UL(1) << 58)
+#else
+#define MHPMEVENTH_OF			(_UL(1) << 31)
+#define MHPMEVENTH_MINH			(_ULL(1) << 30)
+#define MHPMEVENTH_SINH			(_ULL(1) << 29)
+#define MHPMEVENTH_UINH			(_ULL(1) << 28)
+#define MHPMEVENTH_VSINH		(_ULL(1) << 27)
+#define MHPMEVENTH_VUINH		(_ULL(1) << 26)
+
+#define MHPMEVENT_MINH			(MHPMEVENTH_MINH << 32)
+#define MHPMEVENT_SINH			(MHPMEVENTH_SINH << 32)
+#define MHPMEVENT_UINH			(MHPMEVENTH_UINH << 32)
+#define MHPMEVENT_VSINH			(MHPMEVENTH_VSINH << 32)
+#define MHPMEVENT_VUINH			(MHPMEVENTH_VUINH << 32)
+
+#endif
+
+#define MHPMEVENT_SSCOF_MASK		_ULL(0xFFFF000000000000)
+
 /* ===== User-level CSRs ===== */
 
 /* User Trap Setup (N-extension) */
@@ -516,6 +543,40 @@
 #define CSR_MHPMEVENT30			0x33e
 #define CSR_MHPMEVENT31			0x33f
 
+/* For RV32 */
+#define CSR_MHPMEVENT3H			0x723
+#define CSR_MHPMEVENT4H			0x724
+#define CSR_MHPMEVENT5H			0x725
+#define CSR_MHPMEVENT6H			0x726
+#define CSR_MHPMEVENT7H			0x727
+#define CSR_MHPMEVENT8H			0x728
+#define CSR_MHPMEVENT9H			0x729
+#define CSR_MHPMEVENT10H		0x72a
+#define CSR_MHPMEVENT11H		0x72b
+#define CSR_MHPMEVENT12H		0x72c
+#define CSR_MHPMEVENT13H		0x72d
+#define CSR_MHPMEVENT14H		0x72e
+#define CSR_MHPMEVENT15H		0x72f
+#define CSR_MHPMEVENT16H		0x730
+#define CSR_MHPMEVENT17H		0x731
+#define CSR_MHPMEVENT18H		0x732
+#define CSR_MHPMEVENT19H		0x733
+#define CSR_MHPMEVENT20H		0x734
+#define CSR_MHPMEVENT21H		0x735
+#define CSR_MHPMEVENT22H		0x736
+#define CSR_MHPMEVENT23H		0x737
+#define CSR_MHPMEVENT24H		0x738
+#define CSR_MHPMEVENT25H		0x739
+#define CSR_MHPMEVENT26H		0x73a
+#define CSR_MHPMEVENT27H		0x73b
+#define CSR_MHPMEVENT28H		0x73c
+#define CSR_MHPMEVENT29H		0x73d
+#define CSR_MHPMEVENT30H		0x73e
+#define CSR_MHPMEVENT31H		0x73f
+
+/* Counter Overflow CSR */
+#define CSR_SCOUNTOVF			0xda0
+
 /* Debug/Trace Registers */
 #define CSR_TSELECT			0x7a0
 #define CSR_TDATA1			0x7a1

include/sbi/riscv_fp.h

@@ -42,15 +42,28 @@
 			: "t0");                                                                            \
 	})
 #define init_fp_reg(i) SET_F32_REG((i) << 3, 3, 0, 0)
+
+#if __riscv_xlen == 64
 #define GET_F64_REG(insn, pos, regs)                                                                    \
 	({                                                                                              \
-		register ulong value asm("a0") =                                                        \
-			SHIFT_RIGHT(insn, (pos)-3) & 0xf8;                                              \
+		register ulong value asm("a0") = SHIFT_RIGHT(insn, (pos)-3) & 0xf8;                     \
 		ulong tmp;                                                                              \
 		asm("1: auipc %0, %%pcrel_hi(get_f64_reg); add %0, %0, %1; jalr t0, %0, %%pcrel_lo(1b)" \
 		    : "=&r"(tmp), "+&r"(value)::"t0");                                                  \
-		sizeof(ulong) == 4 ? *(int64_t *)value : (int64_t)value;                                \
+		value;                                                                                  \
 	})
+#else
+#define GET_F64_REG(insn, pos, regs)                                                                     \
+	({                                                                                               \
+		u64 value;                                                                               \
+		ulong offset = SHIFT_RIGHT(insn, (pos)-3) & 0xf8;                                        \
+		register ulong ptr asm("a0") = (ulong)&value;                                            \
+		asm ("1: auipc t1, %%pcrel_hi(get_f64_reg); add t1, t1, %2; jalr t0, t1, %%pcrel_lo(1b)" \
+		    : "=m"(value) : "r"(ptr), "r"(offset) : "t0", "t1");                                 \
+		value;                                                                                   \
+	})
+#endif
+
 #define SET_F64_REG(insn, pos, regs, val)                                                                   \
 	({                                                                                                  \
 		uint64_t __val = (val);                                                                     \

include/sbi/riscv_locks.h

@@ -2,30 +2,41 @@
  * SPDX-License-Identifier: BSD-2-Clause
  *
  * Copyright (c) 2019 Western Digital Corporation or its affiliates.
- *
- * Authors:
- *   Anup Patel <anup.patel@wdc.com>
+ * Copyright (c) 2021 Christoph Müllner <cmuellner@linux.com>
  */
 
 #ifndef __RISCV_LOCKS_H__
 #define __RISCV_LOCKS_H__
 
+#include <sbi/sbi_types.h>
+
+#define TICKET_SHIFT	16
+
 typedef struct {
-	volatile long lock;
-} spinlock_t;
+#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+       u16 next;
+       u16 owner;
+#else
+       u16 owner;
+       u16 next;
+#endif
+} __aligned(4) spinlock_t;
 
-#define __RISCV_SPIN_UNLOCKED 0
+#define __SPIN_LOCK_UNLOCKED	\
+	(spinlock_t) { 0, 0 }
 
-#define SPIN_LOCK_INIT(_lptr) (_lptr)->lock = __RISCV_SPIN_UNLOCKED
+#define SPIN_LOCK_INIT(x)	\
+	x = __SPIN_LOCK_UNLOCKED
 
 #define SPIN_LOCK_INITIALIZER	\
-	{                                      \
-		.lock = __RISCV_SPIN_UNLOCKED, \
-	}
+	__SPIN_LOCK_UNLOCKED
 
-int spin_lock_check(spinlock_t *lock);
+#define DEFINE_SPIN_LOCK(x)	\
+	spinlock_t SPIN_LOCK_INIT(x)
 
-int spin_trylock(spinlock_t *lock);
+bool spin_lock_check(spinlock_t *lock);
+
+bool spin_trylock(spinlock_t *lock);
 
 void spin_lock(spinlock_t *lock);
 

include/sbi/sbi_console.h

@@ -12,6 +12,17 @@
 
 #include <sbi/sbi_types.h>
 
+struct sbi_console_device {
+	/** Name of the console device */
+	char name[32];
+
+	/** Write a character to the console output */
+	void (*console_putc)(char ch);
+
+	/** Read a character from the console input */
+	int (*console_getc)(void);
+};
+
 #define __printf(a, b) __attribute__((format(printf, a, b)))
 
 bool sbi_isprintable(char ch);
@@ -32,8 +43,19 @@ int __printf(1, 2) sbi_printf(const char *format, ...);
 
 int __printf(1, 2) sbi_dprintf(const char *format, ...);
 
+void __printf(1, 2) __attribute__((noreturn)) sbi_panic(const char *format, ...);
+
+const struct sbi_console_device *sbi_console_get_device(void);
+
+void sbi_console_set_device(const struct sbi_console_device *dev);
+
 struct sbi_scratch;
 
 int sbi_console_init(struct sbi_scratch *scratch);
 
+#define SBI_ASSERT(cond, args) do { \
+	if (unlikely(!(cond))) \
+		sbi_panic args; \
+} while (0)
+
 #endif

include/sbi/sbi_const.h

@@ -22,7 +22,7 @@
 
 /* clang-format off */
 
-#ifdef __ASSEMBLY__
+#ifdef __ASSEMBLER__
 #define _AC(X,Y)	X
 #define _AT(T,X)	X
 #else

include/sbi/sbi_csr_detect.h

@@ -12,6 +12,7 @@
 
 #include <sbi/riscv_encoding.h>
 #include <sbi/sbi_hart.h>
+#include <sbi/sbi_trap.h>
 
 #define csr_read_allowed(csr_num, trap)					\
 	({								\
@@ -19,6 +20,7 @@
 	register ulong ttmp asm("a4");					\
 	register ulong mtvec = sbi_hart_expected_trap_addr();		\
 	register ulong ret = 0;						\
+	((struct sbi_trap_info *)(trap))->cause = 0;			\
 	asm volatile(							\
 		"add %[ttmp], %[tinfo], zero\n"				\
 		"csrrw %[mtvec], " STR(CSR_MTVEC) ", %[mtvec]\n"	\
@@ -36,6 +38,7 @@
 	register ulong tinfo asm("a3") = (ulong)trap;			\
 	register ulong ttmp asm("a4");					\
 	register ulong mtvec = sbi_hart_expected_trap_addr();		\
+	((struct sbi_trap_info *)(trap))->cause = 0;			\
 	asm volatile(							\
 		"add %[ttmp], %[tinfo], zero\n"				\
 		"csrrw %[mtvec], " STR(CSR_MTVEC) ", %[mtvec]\n"	\

include/sbi/sbi_domain.h

@@ -80,6 +80,9 @@ struct sbi_domain {
 	bool system_reset_allowed;
 };
 
+/** The root domain instance */
+extern struct sbi_domain root;
+
 /** HART id to domain table */
 extern struct sbi_domain *hartid_to_domain_table[];
 
@@ -124,8 +127,19 @@ bool sbi_domain_is_assigned_hart(const struct sbi_domain *dom, u32 hartid);
 ulong sbi_domain_get_assigned_hartmask(const struct sbi_domain *dom,
 				       ulong hbase);
 
-/** Initialize a domain memory region as firmware region */
-void sbi_domain_memregion_initfw(struct sbi_domain_memregion *reg);
+/**
+ * Initialize a domain memory region based on it's physical
+ * address and size.
+ *
+ * @param addr start physical address of memory region
+ * @param size physical size of memory region
+ * @param flags memory region flags
+ * @param reg pointer to memory region being initialized
+ */
+void sbi_domain_memregion_init(unsigned long addr,
+				unsigned long size,
+				unsigned long flags,
+				struct sbi_domain_memregion *reg);
 
 /**
  * Check whether we can access specified address for given mode and
@@ -156,6 +170,14 @@ void sbi_domain_dump_all(const char *suffix);
 int sbi_domain_register(struct sbi_domain *dom,
 			const struct sbi_hartmask *assign_mask);
 
+/**
+ * Add a memory region to the root domain
+ * @param reg pointer to the memory region to be added
+ *
+ * @return 0 on success and negative error code on failure
+ */
+int sbi_domain_root_add_memregion(const struct sbi_domain_memregion *reg);
+
 /** Finalize domain tables and startup non-root domains */
 int sbi_domain_finalize(struct sbi_scratch *scratch, u32 cold_hartid);
 

include/sbi/sbi_ecall.h

@@ -14,7 +14,7 @@
 #include <sbi/sbi_list.h>
 
 #define SBI_ECALL_VERSION_MAJOR		0
-#define SBI_ECALL_VERSION_MINOR		2
+#define SBI_ECALL_VERSION_MINOR		3
 #define SBI_OPENSBI_IMPID		1
 
 struct sbi_trap_regs;
@@ -39,6 +39,7 @@ extern struct sbi_ecall_extension ecall_ipi;
 extern struct sbi_ecall_extension ecall_vendor;
 extern struct sbi_ecall_extension ecall_hsm;
 extern struct sbi_ecall_extension ecall_srst;
+extern struct sbi_ecall_extension ecall_pmu;
 
 u16 sbi_ecall_version_major(void);
 

include/sbi/sbi_ecall_interface.h

@@ -28,6 +28,7 @@
 #define SBI_EXT_RFENCE				0x52464E43
 #define SBI_EXT_HSM				0x48534D
 #define SBI_EXT_SRST				0x53525354
+#define SBI_EXT_PMU				0x504D55
 
 /* SBI function IDs for BASE extension*/
 #define SBI_EXT_BASE_GET_SPEC_VERSION		0x0
@@ -48,20 +49,37 @@
 #define SBI_EXT_RFENCE_REMOTE_FENCE_I		0x0
 #define SBI_EXT_RFENCE_REMOTE_SFENCE_VMA	0x1
 #define SBI_EXT_RFENCE_REMOTE_SFENCE_VMA_ASID	0x2
-#define SBI_EXT_RFENCE_REMOTE_HFENCE_GVMA	0x3
-#define SBI_EXT_RFENCE_REMOTE_HFENCE_GVMA_VMID	0x4
-#define SBI_EXT_RFENCE_REMOTE_HFENCE_VVMA	0x5
-#define SBI_EXT_RFENCE_REMOTE_HFENCE_VVMA_ASID	0x6
+#define SBI_EXT_RFENCE_REMOTE_HFENCE_GVMA_VMID	0x3
+#define SBI_EXT_RFENCE_REMOTE_HFENCE_GVMA	0x4
+#define SBI_EXT_RFENCE_REMOTE_HFENCE_VVMA_ASID	0x5
+#define SBI_EXT_RFENCE_REMOTE_HFENCE_VVMA	0x6
 
 /* SBI function IDs for HSM extension */
 #define SBI_EXT_HSM_HART_START			0x0
 #define SBI_EXT_HSM_HART_STOP			0x1
 #define SBI_EXT_HSM_HART_GET_STATUS		0x2
+#define SBI_EXT_HSM_HART_SUSPEND		0x3
 
-#define SBI_HSM_HART_STATUS_STARTED		0x0
-#define SBI_HSM_HART_STATUS_STOPPED		0x1
-#define SBI_HSM_HART_STATUS_START_PENDING	0x2
-#define SBI_HSM_HART_STATUS_STOP_PENDING	0x3
+#define SBI_HSM_STATE_STARTED			0x0
+#define SBI_HSM_STATE_STOPPED			0x1
+#define SBI_HSM_STATE_START_PENDING		0x2
+#define SBI_HSM_STATE_STOP_PENDING		0x3
+#define SBI_HSM_STATE_SUSPENDED			0x4
+#define SBI_HSM_STATE_SUSPEND_PENDING		0x5
+#define SBI_HSM_STATE_RESUME_PENDING		0x6
+
+#define SBI_HSM_SUSP_BASE_MASK			0x7fffffff
+#define SBI_HSM_SUSP_NON_RET_BIT		0x80000000
+#define SBI_HSM_SUSP_PLAT_BASE			0x10000000
+
+#define SBI_HSM_SUSPEND_RET_DEFAULT		0x00000000
+#define SBI_HSM_SUSPEND_RET_PLATFORM		SBI_HSM_SUSP_PLAT_BASE
+#define SBI_HSM_SUSPEND_RET_LAST		SBI_HSM_SUSP_BASE_MASK
+#define SBI_HSM_SUSPEND_NON_RET_DEFAULT		SBI_HSM_SUSP_NON_RET_BIT
+#define SBI_HSM_SUSPEND_NON_RET_PLATFORM	(SBI_HSM_SUSP_NON_RET_BIT | \
+						 SBI_HSM_SUSP_PLAT_BASE)
+#define SBI_HSM_SUSPEND_NON_RET_LAST		(SBI_HSM_SUSP_NON_RET_BIT | \
+						 SBI_HSM_SUSP_BASE_MASK)
 
 /* SBI function IDs for SRST extension */
 #define SBI_EXT_SRST_RESET			0x0
@@ -74,6 +92,139 @@
 #define SBI_SRST_RESET_REASON_NONE	0x0
 #define SBI_SRST_RESET_REASON_SYSFAIL	0x1
 
+/* SBI function IDs for PMU extension */
+#define SBI_EXT_PMU_NUM_COUNTERS	0x0
+#define SBI_EXT_PMU_COUNTER_GET_INFO	0x1
+#define SBI_EXT_PMU_COUNTER_CFG_MATCH	0x2
+#define SBI_EXT_PMU_COUNTER_START	0x3
+#define SBI_EXT_PMU_COUNTER_STOP	0x4
+#define SBI_EXT_PMU_COUNTER_FW_READ	0x5
+
+/** General pmu event codes specified in SBI PMU extension */
+enum sbi_pmu_hw_generic_events_t {
+	SBI_PMU_HW_NO_EVENT			= 0,
+	SBI_PMU_HW_CPU_CYCLES			= 1,
+	SBI_PMU_HW_INSTRUCTIONS			= 2,
+	SBI_PMU_HW_CACHE_REFERENCES		= 3,
+	SBI_PMU_HW_CACHE_MISSES			= 4,
+	SBI_PMU_HW_BRANCH_INSTRUCTIONS		= 5,
+	SBI_PMU_HW_BRANCH_MISSES		= 6,
+	SBI_PMU_HW_BUS_CYCLES			= 7,
+	SBI_PMU_HW_STALLED_CYCLES_FRONTEND	= 8,
+	SBI_PMU_HW_STALLED_CYCLES_BACKEND	= 9,
+	SBI_PMU_HW_REF_CPU_CYCLES		= 10,
+
+	SBI_PMU_HW_GENERAL_MAX,
+};
+
+/**
+ * Generalized hardware cache events:
+ *
+ *       { L1-D, L1-I, LLC, ITLB, DTLB, BPU, NODE } x
+ *       { read, write, prefetch } x
+ *       { accesses, misses }
+ */
+enum sbi_pmu_hw_cache_id {
+	SBI_PMU_HW_CACHE_L1D		= 0,
+	SBI_PMU_HW_CACHE_L1I		= 1,
+	SBI_PMU_HW_CACHE_LL		= 2,
+	SBI_PMU_HW_CACHE_DTLB		= 3,
+	SBI_PMU_HW_CACHE_ITLB		= 4,
+	SBI_PMU_HW_CACHE_BPU		= 5,
+	SBI_PMU_HW_CACHE_NODE		= 6,
+
+	SBI_PMU_HW_CACHE_MAX,
+};
+
+enum sbi_pmu_hw_cache_op_id {
+	SBI_PMU_HW_CACHE_OP_READ	= 0,
+	SBI_PMU_HW_CACHE_OP_WRITE	= 1,
+	SBI_PMU_HW_CACHE_OP_PREFETCH	= 2,
+
+	SBI_PMU_HW_CACHE_OP_MAX,
+};
+
+enum sbi_pmu_hw_cache_op_result_id {
+	SBI_PMU_HW_CACHE_RESULT_ACCESS	= 0,
+	SBI_PMU_HW_CACHE_RESULT_MISS	= 1,
+
+	SBI_PMU_HW_CACHE_RESULT_MAX,
+};
+
+/**
+ * Special "firmware" events provided by the OpenSBI, even if the hardware
+ * does not support performance events. These events are encoded as a raw
+ * event type in Linux kernel perf framework.
+ */
+enum sbi_pmu_fw_event_code_id {
+	SBI_PMU_FW_MISALIGNED_LOAD	= 0,
+	SBI_PMU_FW_MISALIGNED_STORE	= 1,
+	SBI_PMU_FW_ACCESS_LOAD		= 2,
+	SBI_PMU_FW_ACCESS_STORE		= 3,
+	SBI_PMU_FW_ILLEGAL_INSN		= 4,
+	SBI_PMU_FW_SET_TIMER		= 5,
+	SBI_PMU_FW_IPI_SENT		= 6,
+	SBI_PMU_FW_IPI_RECVD		= 7,
+	SBI_PMU_FW_FENCE_I_SENT		= 8,
+	SBI_PMU_FW_FENCE_I_RECVD	= 9,
+	SBI_PMU_FW_SFENCE_VMA_SENT	= 10,
+	SBI_PMU_FW_SFENCE_VMA_RCVD	= 11,
+	SBI_PMU_FW_SFENCE_VMA_ASID_SENT	= 12,
+	SBI_PMU_FW_SFENCE_VMA_ASID_RCVD	= 13,
+
+	SBI_PMU_FW_HFENCE_GVMA_SENT	= 14,
+	SBI_PMU_FW_HFENCE_GVMA_RCVD	= 15,
+	SBI_PMU_FW_HFENCE_GVMA_VMID_SENT = 16,
+	SBI_PMU_FW_HFENCE_GVMA_VMID_RCVD = 17,
+
+	SBI_PMU_FW_HFENCE_VVMA_SENT	= 18,
+	SBI_PMU_FW_HFENCE_VVMA_RCVD	= 19,
+	SBI_PMU_FW_HFENCE_VVMA_ASID_SENT = 20,
+	SBI_PMU_FW_HFENCE_VVMA_ASID_RCVD = 21,
+	SBI_PMU_FW_MAX,
+};
+
+/** SBI PMU event idx type */
+enum sbi_pmu_event_type_id {
+	SBI_PMU_EVENT_TYPE_HW				= 0x0,
+	SBI_PMU_EVENT_TYPE_HW_CACHE			= 0x1,
+	SBI_PMU_EVENT_TYPE_HW_RAW			= 0x2,
+	SBI_PMU_EVENT_TYPE_FW				= 0xf,
+	SBI_PMU_EVENT_TYPE_MAX,
+};
+
+/** SBI PMU counter type */
+enum sbi_pmu_ctr_type {
+	SBI_PMU_CTR_TYPE_HW = 0,
+	SBI_PMU_CTR_TYPE_FW,
+};
+
+/* Helper macros to decode event idx */
+#define SBI_PMU_EVENT_IDX_OFFSET 20
+#define SBI_PMU_EVENT_IDX_MASK 0xFFFFF
+#define SBI_PMU_EVENT_IDX_CODE_MASK 0xFFFF
+#define SBI_PMU_EVENT_IDX_TYPE_MASK 0xF0000
+#define SBI_PMU_EVENT_RAW_IDX 0x20000
+
+#define SBI_PMU_EVENT_IDX_INVALID 0xFFFFFFFF
+
+/* Flags defined for config matching function */
+#define SBI_PMU_CFG_FLAG_SKIP_MATCH	(1 << 0)
+#define SBI_PMU_CFG_FLAG_CLEAR_VALUE	(1 << 1)
+#define SBI_PMU_CFG_FLAG_AUTO_START	(1 << 2)
+#define SBI_PMU_CFG_FLAG_SET_VUINH	(1 << 3)
+#define SBI_PMU_CFG_FLAG_SET_VSINH	(1 << 4)
+#define SBI_PMU_CFG_FLAG_SET_UINH	(1 << 5)
+#define SBI_PMU_CFG_FLAG_SET_SINH	(1 << 6)
+#define SBI_PMU_CFG_FLAG_SET_MINH	(1 << 7)
+
+/* Flags defined for counter start function */
+#define SBI_PMU_START_FLAG_SET_INIT_VALUE (1 << 0)
+
+/* Flags defined for counter stop function */
+#define SBI_PMU_STOP_FLAG_RESET (1 << 0)
+
+/* SBI base specification related macros */
 #define SBI_SPEC_VERSION_MAJOR_OFFSET		24
 #define SBI_SPEC_VERSION_MAJOR_MASK		0x7f
 #define SBI_SPEC_VERSION_MINOR_MASK		0xffffff
@@ -90,8 +241,10 @@
 #define SBI_ERR_DENIED				-4
 #define SBI_ERR_INVALID_ADDRESS			-5
 #define SBI_ERR_ALREADY_AVAILABLE		-6
+#define SBI_ERR_ALREADY_STARTED			-7
+#define SBI_ERR_ALREADY_STOPPED			-8
 
-#define SBI_LAST_ERR				SBI_ERR_ALREADY_AVAILABLE
+#define SBI_LAST_ERR				SBI_ERR_ALREADY_STOPPED
 
 /* clang-format on */
 

include/sbi/sbi_error.h

@@ -21,6 +21,8 @@
 #define SBI_EDENIED		SBI_ERR_DENIED
 #define SBI_EINVALID_ADDR	SBI_ERR_INVALID_ADDRESS
 #define SBI_EALREADY		SBI_ERR_ALREADY_AVAILABLE
+#define SBI_EALREADY_STARTED	SBI_ERR_ALREADY_STARTED
+#define SBI_EALREADY_STOPPED	SBI_ERR_ALREADY_STOPPED
 
 #define SBI_ENODEV		-1000
 #define SBI_ENOSYS		-1001

include/sbi/sbi_fifo.h

@@ -33,8 +33,8 @@ int sbi_fifo_dequeue(struct sbi_fifo *fifo, void *data);
 int sbi_fifo_enqueue(struct sbi_fifo *fifo, void *data);
 void sbi_fifo_init(struct sbi_fifo *fifo, void *queue_mem, u16 entries,
 		   u16 entry_size);
-bool sbi_fifo_is_empty(struct sbi_fifo *fifo);
-bool sbi_fifo_is_full(struct sbi_fifo *fifo);
+int sbi_fifo_is_empty(struct sbi_fifo *fifo);
+int sbi_fifo_is_full(struct sbi_fifo *fifo);
 int sbi_fifo_inplace_update(struct sbi_fifo *fifo, void *in,
 			    int (*fptr)(void *in, void *data));
 u16 sbi_fifo_avail(struct sbi_fifo *fifo);

include/sbi/sbi_hart.h

@@ -18,8 +18,12 @@ enum sbi_hart_features {
 	SBI_HART_HAS_SCOUNTEREN = (1 << 0),
 	/** Hart has M-mode counter enable */
 	SBI_HART_HAS_MCOUNTEREN = (1 << 1),
+	/** Hart has counter inhibit CSR */
+	SBI_HART_HAS_MCOUNTINHIBIT = (1 << 2),
+	/** Hart has sscofpmf extension */
+	SBI_HART_HAS_SSCOFPMF = (1 << 3),
 	/** HART has timer csr implementation in hardware */
-	SBI_HART_HAS_TIME = (1 << 2),
+	SBI_HART_HAS_TIME = (1 << 4),
 
 	/** Last index of Hart features*/
 	SBI_HART_HAS_LAST_FEATURE = SBI_HART_HAS_TIME,
@@ -27,6 +31,7 @@ enum sbi_hart_features {
 
 struct sbi_scratch;
 
+int sbi_hart_reinit(struct sbi_scratch *scratch);
 int sbi_hart_init(struct sbi_scratch *scratch, bool cold_boot);
 
 extern void (*sbi_hart_expected_trap)(void);
@@ -41,6 +46,7 @@ void sbi_hart_delegation_dump(struct sbi_scratch *scratch,
 unsigned int sbi_hart_pmp_count(struct sbi_scratch *scratch);
 unsigned long sbi_hart_pmp_granularity(struct sbi_scratch *scratch);
 unsigned int sbi_hart_pmp_addrbits(struct sbi_scratch *scratch);
+unsigned int sbi_hart_mhpm_bits(struct sbi_scratch *scratch);
 int sbi_hart_pmp_configure(struct sbi_scratch *scratch);
 bool sbi_hart_has_feature(struct sbi_scratch *scratch, unsigned long feature);
 void sbi_hart_get_features_str(struct sbi_scratch *scratch,

include/sbi/sbi_hfence.h

@@ -12,13 +12,14 @@
 #define __SBI_FENCE_H__
 
 /** Invalidate Stage2 TLBs for given VMID and guest physical address */
-void __sbi_hfence_gvma_vmid_gpa(unsigned long gpa, unsigned long vmid);
+void __sbi_hfence_gvma_vmid_gpa(unsigned long gpa_divby_4,
+				unsigned long vmid);
 
 /** Invalidate Stage2 TLBs for given VMID */
 void __sbi_hfence_gvma_vmid(unsigned long vmid);
 
 /** Invalidate Stage2 TLBs for given guest physical address */
-void __sbi_hfence_gvma_gpa(unsigned long gpa);
+void __sbi_hfence_gvma_gpa(unsigned long gpa_divby_4);
 
 /** Invalidate all possible Stage2 TLBs */
 void __sbi_hfence_gvma_all(void);

include/sbi/sbi_hsm.h

@@ -12,16 +12,40 @@
 
 #include <sbi/sbi_types.h>
 
-/** Hart state values **/
-#define SBI_HART_STOPPED	0
-#define SBI_HART_STOPPING	1
-#define SBI_HART_STARTING	2
-#define SBI_HART_STARTED	3
-#define SBI_HART_UNKNOWN	4
+/** Hart state managment device */
+struct sbi_hsm_device {
+	/** Name of the hart state managment device */
+	char name[32];
+
+	/** Start (or power-up) the given hart */
+	int (*hart_start)(u32 hartid, ulong saddr);
+
+	/**
+	 * Stop (or power-down) the current hart from running. This call
+	 * doesn't expect to return if success.
+	 */
+	int (*hart_stop)(void);
+
+	/**
+	 * Put the current hart in platform specific suspend (or low-power)
+	 * state.
+	 *
+	 * For successful retentive suspend, the call will return 0 when
+	 * the hart resumes normal execution.
+	 *
+	 * For successful non-retentive suspend, the hart will resume from
+	 * specified resume address
+	 */
+	int (*hart_suspend)(u32 suspend_type, ulong raddr);
+};
 
 struct sbi_domain;
 struct sbi_scratch;
 
+const struct sbi_hsm_device *sbi_hsm_get_device(void);
+
+void sbi_hsm_set_device(const struct sbi_hsm_device *dev);
+
 int sbi_hsm_init(struct sbi_scratch *scratch, u32 hartid, bool cold_boot);
 void __noreturn sbi_hsm_exit(struct sbi_scratch *scratch);
 
@@ -29,9 +53,12 @@ int sbi_hsm_hart_start(struct sbi_scratch *scratch,
 		       const struct sbi_domain *dom,
 		       u32 hartid, ulong saddr, ulong smode, ulong priv);
 int sbi_hsm_hart_stop(struct sbi_scratch *scratch, bool exitnow);
+void sbi_hsm_hart_resume_start(struct sbi_scratch *scratch);
+void sbi_hsm_hart_resume_finish(struct sbi_scratch *scratch);
+int sbi_hsm_hart_suspend(struct sbi_scratch *scratch, u32 suspend_type,
+			 ulong raddr, ulong rmode, ulong priv);
 int sbi_hsm_hart_get_state(const struct sbi_domain *dom, u32 hartid);
-int sbi_hsm_hart_state_to_status(int state);
-int sbi_hsm_hart_started_mask(const struct sbi_domain *dom,
+int sbi_hsm_hart_interruptible_mask(const struct sbi_domain *dom,
 				    ulong hbase, ulong *out_hmask);
 void sbi_hsm_prepare_next_jump(struct sbi_scratch *scratch, u32 hartid);
 

include/sbi/sbi_ipi.h

@@ -18,6 +18,18 @@
 
 /* clang-format on */
 
+/** IPI hardware device */
+struct sbi_ipi_device {
+	/** Name of the IPI device */
+	char name[32];
+
+	/** Send IPI to a target HART */
+	void (*ipi_send)(u32 target_hart);
+
+	/** Clear IPI for a target HART */
+	void (*ipi_clear)(u32 target_hart);
+};
+
 struct sbi_scratch;
 
 /** IPI event operations or callbacks */
@@ -63,6 +75,12 @@ int sbi_ipi_send_halt(ulong hmask, ulong hbase);
 
 void sbi_ipi_process(void);
 
+void sbi_ipi_raw_send(u32 target_hart);
+
+const struct sbi_ipi_device *sbi_ipi_get_device(void);
+
+void sbi_ipi_set_device(const struct sbi_ipi_device *dev);
+
 int sbi_ipi_init(struct sbi_scratch *scratch, bool cold_boot);
 
 void sbi_ipi_exit(struct sbi_scratch *scratch);

include/sbi/sbi_list.h

@@ -43,6 +43,17 @@ static inline void __sbi_list_add(struct sbi_dlist *new,
 	next->prev = new;
 }
 
+/**
+ * Checks if the list is empty or not.
+ * @param head List head
+ *
+ * Retruns TRUE if list is empty, FALSE otherwise.
+ */
+static inline bool sbi_list_empty(struct sbi_dlist *head)
+{
+	return head->next == head;
+}
+
 /**
  * Adds the new node after the given head.
  * @param new New node that needs to be added to list.

include/sbi/sbi_platform.h

@@ -38,7 +38,7 @@
 
 #define SBI_PLATFORM_TLB_RANGE_FLUSH_LIMIT_DEFAULT		(1UL << 12)
 
-#ifndef __ASSEMBLY__
+#ifndef __ASSEMBLER__
 
 #include <sbi/sbi_ecall_interface.h>
 #include <sbi/sbi_error.h>
@@ -51,22 +51,16 @@ struct sbi_trap_regs;
 
 /** Possible feature flags of a platform */
 enum sbi_platform_features {
-	/** Platform has timer value */
-	SBI_PLATFORM_HAS_TIMER_VALUE = (1 << 0),
-	/** Platform has HART hotplug support */
-	SBI_PLATFORM_HAS_HART_HOTPLUG = (1 << 1),
 	/** Platform has fault delegation support */
-	SBI_PLATFORM_HAS_MFAULTS_DELEGATION = (1 << 2),
-	/** Platform has custom secondary hart booting support */
-	SBI_PLATFORM_HAS_HART_SECONDARY_BOOT = (1 << 3),
+	SBI_PLATFORM_HAS_MFAULTS_DELEGATION = (1 << 1),
 
 	/** Last index of Platform features*/
-	SBI_PLATFORM_HAS_LAST_FEATURE = SBI_PLATFORM_HAS_HART_SECONDARY_BOOT,
+	SBI_PLATFORM_HAS_LAST_FEATURE = SBI_PLATFORM_HAS_MFAULTS_DELEGATION,
 };
 
 /** Default feature set for a platform */
 #define SBI_PLATFORM_DEFAULT_FEATURES                                \
-	(SBI_PLATFORM_HAS_TIMER_VALUE | SBI_PLATFORM_HAS_MFAULTS_DELEGATION)
+	(SBI_PLATFORM_HAS_MFAULTS_DELEGATION)
 
 /** Platform functions */
 struct sbi_platform_operations {
@@ -92,15 +86,15 @@ struct sbi_platform_operations {
 	 */
 	int (*misa_get_xlen)(void);
 
-	/** Get platform specific root domain memory regions */
-	struct sbi_domain_memregion *(*domains_root_regions)(void);
 	/** Initialize (or populate) domains for the platform */
 	int (*domains_init)(void);
 
-	/** Write a character to the platform console output */
-	void (*console_putc)(char ch);
-	/** Read a character from the platform console input */
-	int (*console_getc)(void);
+	/** Initialize hw performance counters */
+	int (*pmu_init)(void);
+
+	/** Get platform specific mhpmevent value */
+	uint64_t (*pmu_xlate_to_mhpmevent)(uint32_t event_idx, uint64_t data);
+
 	/** Initialize the platform console */
 	int (*console_init)(void);
 
@@ -109,10 +103,6 @@ struct sbi_platform_operations {
 	/** Exit the platform interrupt controller for current HART */
 	void (*irqchip_exit)(void);
 
-	/** Send IPI to a target HART */
-	void (*ipi_send)(u32 target_hart);
-	/** Clear IPI for a target HART */
-	void (*ipi_clear)(u32 target_hart);
 	/** Initialize IPI for current HART */
 	int (*ipi_init)(bool cold_boot);
 	/** Exit IPI for current HART */
@@ -121,30 +111,11 @@ struct sbi_platform_operations {
 	/** Get tlb flush limit value **/
 	u64 (*get_tlbr_flush_limit)(void);
 
-	/** Get platform timer value */
-	u64 (*timer_value)(void);
-	/** Start platform timer event for current HART */
-	void (*timer_event_start)(u64 next_event);
-	/** Stop platform timer event for current HART */
-	void (*timer_event_stop)(void);
 	/** Initialize platform timer for current HART */
 	int (*timer_init)(bool cold_boot);
 	/** Exit platform timer for current HART */
 	void (*timer_exit)(void);
 
-	/** Bringup the given hart */
-	int (*hart_start)(u32 hartid, ulong saddr);
-	/**
-	 * Stop the current hart from running. This call doesn't expect to
-	 * return if success.
-	 */
-	int (*hart_stop)(void);
-
-	/* Check whether reset type and reason supported by the platform */
-	int (*system_reset_check)(u32 reset_type, u32 reset_reason);
-	/** Reset the platform */
-	void (*system_reset)(u32 reset_type, u32 reset_reason);
-
 	/** platform specific SBI extension implementation probe function */
 	int (*vendor_ext_check)(long extid);
 	/** platform specific SBI extension implementation provider */
@@ -211,18 +182,9 @@ struct sbi_platform {
 #define sbi_platform_ops(__p) \
 	((const struct sbi_platform_operations *)(__p)->platform_ops_addr)
 
-/** Check whether the platform supports timer value */
-#define sbi_platform_has_timer_value(__p) \
-	((__p)->features & SBI_PLATFORM_HAS_TIMER_VALUE)
-/** Check whether the platform supports HART hotplug */
-#define sbi_platform_has_hart_hotplug(__p) \
-	((__p)->features & SBI_PLATFORM_HAS_HART_HOTPLUG)
 /** Check whether the platform supports fault delegation */
 #define sbi_platform_has_mfaults_delegation(__p) \
 	((__p)->features & SBI_PLATFORM_HAS_MFAULTS_DELEGATION)
-/** Check whether the platform supports custom secondary hart booting support */
-#define sbi_platform_has_hart_secondary_boot(__p) \
-	((__p)->features & SBI_PLATFORM_HAS_HART_SECONDARY_BOOT)
 
 /**
  * Get HART index for the given HART
@@ -337,39 +299,6 @@ static inline bool sbi_platform_hart_invalid(const struct sbi_platform *plat,
 	return FALSE;
 }
 
-/**
- * Bringup a given hart from previous stage. Platform should implement this
- * operation if they support a custom mechanism to start a hart. Otherwise,
- * a generic WFI based approach will be used to start/stop a hart in OpenSBI.
- *
- * @param plat pointer to struct sbi_platform
- * @param hartid HART id
- * @param saddr M-mode start physical address for the HART
- *
- * @return 0 if sucessful and negative error code on failure
- */
-static inline int sbi_platform_hart_start(const struct sbi_platform *plat,
-					  u32 hartid, ulong saddr)
-{
-	if (plat && sbi_platform_ops(plat)->hart_start)
-		return sbi_platform_ops(plat)->hart_start(hartid, saddr);
-	return SBI_ENOTSUPP;
-}
-
-/**
- * Stop the current hart in OpenSBI.
- *
- * @param plat pointer to struct sbi_platform
- *
- * @return Negative error code on failure. It doesn't return on success.
- */
-static inline int sbi_platform_hart_stop(const struct sbi_platform *plat)
-{
-	if (plat && sbi_platform_ops(plat)->hart_stop)
-		return sbi_platform_ops(plat)->hart_stop();
-	return SBI_ENOTSUPP;
-}
-
 /**
  * Early initialization for current HART
  *
@@ -454,22 +383,6 @@ static inline int sbi_platform_misa_xlen(const struct sbi_platform *plat)
 	return -1;
 }
 
-/**
- * Get platform specific root domain memory regions
- *
- * @param plat pointer to struct sbi_platform
- *
- * @return an array of memory regions terminated by a region with order zero
- * or NULL for no memory regions
- */
-static inline struct sbi_domain_memregion *
-sbi_platform_domains_root_regions(const struct sbi_platform *plat)
-{
-	if (plat && sbi_platform_ops(plat)->domains_root_regions)
-		return sbi_platform_ops(plat)->domains_root_regions();
-	return NULL;
-}
-
 /**
  * Initialize (or populate) domains for the platform
  *
@@ -485,30 +398,36 @@ static inline int sbi_platform_domains_init(const struct sbi_platform *plat)
 }
 
 /**
- * Write a character to the platform console output
+ * Setup hw PMU events for the platform
  *
  * @param plat pointer to struct sbi_platform
- * @param ch character to write
+ *
+ * @return 0 on success and negative error code on failure
  */
-static inline void sbi_platform_console_putc(const struct sbi_platform *plat,
-						char ch)
+static inline int sbi_platform_pmu_init(const struct sbi_platform *plat)
 {
-	if (plat && sbi_platform_ops(plat)->console_putc)
-		sbi_platform_ops(plat)->console_putc(ch);
+	if (plat && sbi_platform_ops(plat)->pmu_init)
+		return sbi_platform_ops(plat)->pmu_init();
+	return 0;
 }
 
 /**
- * Read a character from the platform console input
+ * Get the value to be written in mhpmeventx for event_idx
  *
  * @param plat pointer to struct sbi_platform
+ * @param event_idx ID of the PMU event
+ * @param data Additional configuration data passed from supervisor software
  *
- * @return character read from console input
+ * @return expected value by the platform or 0 if platform doesn't know about
+ * the event
  */
-static inline int sbi_platform_console_getc(const struct sbi_platform *plat)
+static inline uint64_t sbi_platform_pmu_xlate_to_mhpmevent(const struct sbi_platform *plat,
+						      uint32_t event_idx, uint64_t data)
 {
-	if (plat && sbi_platform_ops(plat)->console_getc)
-		return sbi_platform_ops(plat)->console_getc();
-	return -1;
+	if (plat && sbi_platform_ops(plat)->pmu_xlate_to_mhpmevent)
+		return sbi_platform_ops(plat)->pmu_xlate_to_mhpmevent(event_idx,
+								      data);
+	return 0;
 }
 
 /**
@@ -552,32 +471,6 @@ static inline void sbi_platform_irqchip_exit(const struct sbi_platform *plat)
 		sbi_platform_ops(plat)->irqchip_exit();
 }
 
-/**
- * Send IPI to a target HART
- *
- * @param plat pointer to struct sbi_platform
- * @param target_hart HART ID of IPI target
- */
-static inline void sbi_platform_ipi_send(const struct sbi_platform *plat,
-					 u32 target_hart)
-{
-	if (plat && sbi_platform_ops(plat)->ipi_send)
-		sbi_platform_ops(plat)->ipi_send(target_hart);
-}
-
-/**
- * Clear IPI for a target HART
- *
- * @param plat pointer to struct sbi_platform
- * @param target_hart HART ID of IPI target
- */
-static inline void sbi_platform_ipi_clear(const struct sbi_platform *plat,
-					  u32 target_hart)
-{
-	if (plat && sbi_platform_ops(plat)->ipi_clear)
-		sbi_platform_ops(plat)->ipi_clear(target_hart);
-}
-
 /**
  * Initialize the platform IPI support for current HART
  *
@@ -605,45 +498,6 @@ static inline void sbi_platform_ipi_exit(const struct sbi_platform *plat)
 		sbi_platform_ops(plat)->ipi_exit();
 }
 
-/**
- * Get platform timer value
- *
- * @param plat pointer to struct sbi_platform
- *
- * @return 64-bit timer value
- */
-static inline u64 sbi_platform_timer_value(const struct sbi_platform *plat)
-{
-	if (plat && sbi_platform_ops(plat)->timer_value)
-		return sbi_platform_ops(plat)->timer_value();
-	return 0;
-}
-
-/**
- * Start platform timer event for current HART
- *
- * @param plat pointer to struct struct sbi_platform
- * @param next_event timer value when timer event will happen
- */
-static inline void
-sbi_platform_timer_event_start(const struct sbi_platform *plat, u64 next_event)
-{
-	if (plat && sbi_platform_ops(plat)->timer_event_start)
-		sbi_platform_ops(plat)->timer_event_start(next_event);
-}
-
-/**
- * Stop platform timer event for current HART
- *
- * @param plat pointer to struct sbi_platform
- */
-static inline void
-sbi_platform_timer_event_stop(const struct sbi_platform *plat)
-{
-	if (plat && sbi_platform_ops(plat)->timer_event_stop)
-		sbi_platform_ops(plat)->timer_event_stop();
-}
-
 /**
  * Initialize the platform timer for current HART
  *
@@ -671,41 +525,6 @@ static inline void sbi_platform_timer_exit(const struct sbi_platform *plat)
 		sbi_platform_ops(plat)->timer_exit();
 }
 
-/**
- * Check whether reset type and reason supported by the platform
- *
- * @param plat pointer to struct sbi_platform
- * @param reset_type type of reset
- * @param reset_reason reason for reset
- *
- * @return 0 if reset type and reason not supported and 1 if supported
- */
-static inline int sbi_platform_system_reset_check(
-					    const struct sbi_platform *plat,
-					    u32 reset_type, u32 reset_reason)
-{
-	if (plat && sbi_platform_ops(plat)->system_reset_check)
-		return sbi_platform_ops(plat)->system_reset_check(reset_type,
-								  reset_reason);
-	return 0;
-}
-
-/**
- * Reset the platform
- *
- * This function will not return for supported reset type and reset reason
- *
- * @param plat pointer to struct sbi_platform
- * @param reset_type type of reset
- * @param reset_reason reason for reset
- */
-static inline void sbi_platform_system_reset(const struct sbi_platform *plat,
-					     u32 reset_type, u32 reset_reason)
-{
-	if (plat && sbi_platform_ops(plat)->system_reset)
-		sbi_platform_ops(plat)->system_reset(reset_type, reset_reason);
-}
-
 /**
  * Check if a vendor extension is implemented or not.
  *

include/sbi/sbi_pmu.h

@@ -0,0 +1,74 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Atish Patra <atish.patra@wdc.com>
+ */
+
+#ifndef __SBI_PMU_H__
+#define __SBI_PMU_H__
+
+#include <sbi/sbi_types.h>
+#include <sbi/sbi_hartmask.h>
+#include <sbi/sbi_scratch.h>
+#include <sbi/sbi_ecall_interface.h>
+
+/* Event related macros */
+/* Maximum number of hardware events that can mapped by OpenSBI */
+#define SBI_PMU_HW_EVENT_MAX 64
+
+/* Maximum number of firmware events that can mapped by OpenSBI */
+#define SBI_PMU_FW_EVENT_MAX 32
+
+/* Counter related macros */
+#define SBI_PMU_FW_CTR_MAX 16
+#define SBI_PMU_HW_CTR_MAX 32
+#define SBI_PMU_CTR_MAX	   (SBI_PMU_HW_CTR_MAX + SBI_PMU_FW_CTR_MAX)
+#define SBI_PMU_FIXED_CTR_MASK 0x07
+
+/** Initialize PMU */
+int sbi_pmu_init(struct sbi_scratch *scratch, bool cold_boot);
+
+/** Reset PMU during hart exit */
+void sbi_pmu_exit(struct sbi_scratch *scratch);
+
+/**
+ * Add the hardware event to counter mapping information. This should be called
+ * from the platform code to update the mapping table.
+ * @param eidx_start Start of the event idx range for supported counters
+ * @param eidx_end   End of the event idx range for supported counters
+ * @param cmap       A bitmap representing counters supporting the event range
+ * @return 0 on success, error otherwise.
+ */
+int sbi_pmu_add_hw_event_counter_map(u32 eidx_start, u32 eidx_end, u32 cmap);
+
+/**
+ * Add the raw hardware event selector and supported counter information. This
+ * should be called from the platform code to update the mapping table.
+ * @param info  a pointer to the hardware event info
+ * @return 0 on success, error otherwise.
+ */
+
+int sbi_pmu_add_raw_event_counter_map(uint64_t select, uint64_t select_mask, u32 cmap);
+
+int sbi_pmu_ctr_read(uint32_t cidx, unsigned long *cval);
+
+int sbi_pmu_ctr_stop(unsigned long cidx_base, unsigned long cidx_mask,
+		     unsigned long flag);
+
+int sbi_pmu_ctr_start(unsigned long cidx_base, unsigned long cidx_mask,
+		      unsigned long flags, uint64_t ival);
+
+int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info);
+
+unsigned long sbi_pmu_num_ctr(void);
+
+int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,
+			  unsigned long flags, unsigned long event_idx,
+			  uint64_t event_data);
+
+int sbi_pmu_ctr_incr_fw(enum sbi_pmu_fw_event_code_id fw_id);
+
+#endif

include/sbi/sbi_scratch.h

@@ -43,7 +43,7 @@
 
 /* clang-format on */
 
-#ifndef __ASSEMBLY__
+#ifndef __ASSEMBLER__
 
 #include <sbi/sbi_types.h>
 
@@ -98,7 +98,7 @@ int sbi_scratch_init(struct sbi_scratch *scratch);
  * @return zero on failure and non-zero (>= SBI_SCRATCH_EXTRA_SPACE_OFFSET)
  * on success
  */
-unsigned long sbi_scratch_alloc_offset(unsigned long size, const char *owner);
+unsigned long sbi_scratch_alloc_offset(unsigned long size);
 
 /** Free-up extra space in sbi_scratch */
 void sbi_scratch_free_offset(unsigned long offset);

include/sbi/sbi_system.h

@@ -11,6 +11,33 @@
 #define __SBI_SYSTEM_H__
 
 #include <sbi/sbi_types.h>
+#include <sbi/sbi_list.h>
+
+/** System reset hardware device */
+struct sbi_system_reset_device {
+	/** Name of the system reset device */
+	char name[32];
+
+	/* Check whether reset type and reason supported by the device */
+	int (*system_reset_check)(u32 reset_type, u32 reset_reason);
+
+	/** Reset the system */
+	void (*system_reset)(u32 reset_type, u32 reset_reason);
+
+	/** List */
+	struct sbi_dlist node;
+};
+
+static inline struct sbi_system_reset_device *to_system_reset_device(
+						struct sbi_dlist *node)
+{
+	return container_of(node, struct sbi_system_reset_device, node);
+}
+
+const struct sbi_system_reset_device *sbi_system_reset_get_device(
+					u32 reset_type, u32 reset_reason);
+
+void sbi_system_reset_add_device(struct sbi_system_reset_device *dev);
 
 bool sbi_system_reset_supported(u32 reset_type, u32 reset_reason);
 

include/sbi/sbi_timer.h

@@ -12,8 +12,42 @@
 
 #include <sbi/sbi_types.h>
 
+/** Timer hardware device */
+struct sbi_timer_device {
+	/** Name of the timer operations */
+	char name[32];
+
+	/** Frequency of timer in HZ */
+	unsigned long timer_freq;
+
+	/** Get free-running timer value */
+	u64 (*timer_value)(void);
+
+	/** Start timer event for current HART */
+	void (*timer_event_start)(u64 next_event);
+
+	/** Stop timer event for current HART */
+	void (*timer_event_stop)(void);
+};
+
 struct sbi_scratch;
 
+/** Generic delay loop of desired granularity */
+void sbi_timer_delay_loop(ulong units, u64 unit_freq,
+			  void (*delay_fn)(void *), void *opaque);
+
+/** Provide delay in terms of milliseconds */
+static inline void sbi_timer_mdelay(ulong msecs)
+{
+	sbi_timer_delay_loop(msecs, 1000, NULL, NULL);
+}
+
+/** Provide delay in terms of microseconds */
+static inline void sbi_timer_udelay(ulong usecs)
+{
+	sbi_timer_delay_loop(usecs, 1000000, NULL, NULL);
+}
+
 /** Get timer value for current HART */
 u64 sbi_timer_value(void);
 
@@ -35,6 +69,12 @@ void sbi_timer_event_start(u64 next_event);
 /** Process timer event for current HART */
 void sbi_timer_process(void);
 
+/** Get current timer device */
+const struct sbi_timer_device *sbi_timer_get_device(void);
+
+/** Register timer device */
+void sbi_timer_set_device(const struct sbi_timer_device *dev);
+
 /* Initialize timer */
 int sbi_timer_init(struct sbi_scratch *scratch, bool cold_boot);
 

include/sbi/sbi_trap.h

@@ -110,7 +110,7 @@
 /** Size (in bytes) of sbi_trap_info */
 #define SBI_TRAP_INFO_SIZE SBI_TRAP_INFO_OFFSET(last)
 
-#ifndef __ASSEMBLY__
+#ifndef __ASSEMBLER__
 
 #include <sbi/sbi_types.h>
 
@@ -205,7 +205,7 @@ struct sbi_trap_info {
 int sbi_trap_redirect(struct sbi_trap_regs *regs,
 		      struct sbi_trap_info *trap);
 
-void sbi_trap_handler(struct sbi_trap_regs *regs);
+struct sbi_trap_regs *sbi_trap_handler(struct sbi_trap_regs *regs);
 
 void __noreturn sbi_trap_exit(const struct sbi_trap_regs *regs);
 

include/sbi/sbi_types.h

@@ -63,12 +63,19 @@ typedef unsigned long		physical_size_t;
 
 #define __packed		__attribute__((packed))
 #define __noreturn		__attribute__((noreturn))
+#define __aligned(x)		__attribute__((aligned(x)))
 
 #define likely(x) __builtin_expect((x), 1)
 #define unlikely(x) __builtin_expect((x), 0)
 
+#ifndef __has_builtin
+#define __has_builtin(...) 0
+#endif
+
 #undef offsetof
-#ifdef __compiler_offsetof
+#if __has_builtin(__builtin_offsetof)
+#define offsetof(TYPE, MEMBER) __builtin_offsetof(TYPE,MEMBER)
+#elif defined(__compiler_offsetof)
 #define offsetof(TYPE, MEMBER) __compiler_offsetof(TYPE,MEMBER)
 #else
 #define offsetof(TYPE, MEMBER) ((size_t) &((TYPE *)0)->MEMBER)

include/sbi/sbi_version.h

@@ -10,8 +10,8 @@
 #ifndef __SBI_VERSION_H__
 #define __SBI_VERSION_H__
 
-#define OPENSBI_VERSION_MAJOR 0
-#define OPENSBI_VERSION_MINOR 9
+#define OPENSBI_VERSION_MAJOR 1
+#define OPENSBI_VERSION_MINOR 0
 
 /**
  * OpenSBI 32-bit version with:

include/sbi_utils/fdt/fdt_fixup.h

@@ -30,9 +30,8 @@ void fdt_cpu_fixup(void *fdt);
  * It is recommended that platform codes call this helper in their final_init()
  *
  * @param fdt: device tree blob
- * @param compat: PLIC node compatible string
  */
-void fdt_plic_fixup(void *fdt, const char *compat);
+void fdt_plic_fixup(void *fdt);
 
 /**
  * Fix up the reserved memory node in the device tree

include/sbi_utils/fdt/fdt_helper.h

@@ -11,12 +11,20 @@
 #define __FDT_HELPER_H__
 
 #include <sbi/sbi_types.h>
+#include <sbi/sbi_scratch.h>
 
 struct fdt_match {
 	const char *compatible;
 	void *data;
 };
 
+#define FDT_MAX_PHANDLE_ARGS 16
+struct fdt_phandle_args {
+	int node_offset;
+	int args_count;
+	u32 args[FDT_MAX_PHANDLE_ARGS];
+};
+
 struct platform_uart_data {
 	unsigned long addr;
 	unsigned long freq;
@@ -32,13 +40,22 @@ int fdt_find_match(void *fdt, int startoff,
 		   const struct fdt_match *match_table,
 		   const struct fdt_match **out_match);
 
-int fdt_get_node_addr_size(void *fdt, int node, unsigned long *addr,
-			   unsigned long *size);
+int fdt_parse_phandle_with_args(void *fdt, int nodeoff,
+				const char *prop, const char *cells_prop,
+				int index, struct fdt_phandle_args *out_args);
+
+int fdt_get_node_addr_size(void *fdt, int node, int index,
+			   uint64_t *addr, uint64_t *size);
 
 int fdt_parse_hart_id(void *fdt, int cpu_offset, u32 *hartid);
 
 int fdt_parse_max_hart_id(void *fdt, u32 *max_hartid);
 
+int fdt_parse_timebase_frequency(void *fdt, unsigned long *freq);
+
+int fdt_parse_gaisler_uart_node(void *fdt, int nodeoffset,
+				struct platform_uart_data *uart);
+
 int fdt_parse_shakti_uart_node(void *fdt, int nodeoffset,
 			       struct platform_uart_data *uart);
 
@@ -57,12 +74,17 @@ int fdt_parse_plic_node(void *fdt, int nodeoffset, struct plic_data *plic);
 
 int fdt_parse_plic(void *fdt, struct plic_data *plic, const char *compat);
 
-struct clint_data;
+int fdt_parse_aclint_node(void *fdt, int nodeoffset, bool for_timer,
+			  unsigned long *out_addr1, unsigned long *out_size1,
+			  unsigned long *out_addr2, unsigned long *out_size2,
+			  u32 *out_first_hartid, u32 *out_hart_count);
 
-int fdt_parse_clint_node(void *fdt, int nodeoffset, bool for_timer,
-			 struct clint_data *clint);
-
-int fdt_parse_compat_addr(void *fdt, unsigned long *addr,
+int fdt_parse_compat_addr(void *fdt, uint64_t *addr,
 			  const char *compatible);
 
+static inline void *fdt_get_address(void)
+{
+	return sbi_scratch_thishart_arg1_ptr();
+}
+
 #endif /* __FDT_HELPER_H__ */

include/sbi_utils/fdt/fdt_pmu.h

@@ -0,0 +1,46 @@
+// SPDX-License-Identifier: BSD-2-Clause
+/*
+ * fdt_pmu.c - Flat Device Tree PMU helper routines
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Atish Patra <atish.patra@wdc.com>
+ */
+
+#ifndef __FDT_PMU_H__
+#define __FDT_PMU_H__
+
+#include <sbi/sbi_types.h>
+
+/**
+ * Fix up the PMU node in the device tree
+ *
+ * This routine:
+ * 1. Disables opensbi specific properties from the DT
+ *
+ * It is recommended that platform support call this function in
+ * their final_init() platform operation.
+ *
+ * @param fdt device tree blob
+ */
+void fdt_pmu_fixup(void *fdt);
+
+/**
+ * Setup PMU data from device tree
+ *
+ * @param fdt device tree blob
+ *
+ * @return 0 on success and negative error code on failure
+ */
+int fdt_pmu_setup(void *fdt);
+
+/**
+ * Get the mhpmevent select value read from DT for a given event
+ * @param event_idx Event ID of the given event
+ *
+ * @return The select value read from DT or 0 if given index was not found
+ */
+uint64_t fdt_pmu_get_select_value(uint32_t event_idx);
+
+#endif

include/sbi_utils/gpio/fdt_gpio.h

@@ -0,0 +1,34 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Anup Patel <anup.patel@wdc.com>
+ */
+
+#ifndef __FDT_GPIO_H__
+#define __FDT_GPIO_H__
+
+#include <sbi_utils/gpio/gpio.h>
+
+/** FDT based GPIO driver */
+struct fdt_gpio {
+	const struct fdt_match *match_table;
+	int (*xlate)(struct gpio_chip *chip,
+		     const struct fdt_phandle_args *pargs,
+		     struct gpio_pin *out_pin);
+	int (*init)(void *fdt, int nodeoff, u32 phandle,
+		    const struct fdt_match *match);
+};
+
+/** Get a GPIO pin using "gpios" DT property of client DT node */
+int fdt_gpio_pin_get(void *fdt, int nodeoff, int index,
+		     struct gpio_pin *out_pin);
+
+/** Simple xlate function to convert two GPIO FDT cells into GPIO pin */
+int fdt_gpio_simple_xlate(struct gpio_chip *chip,
+			  const struct fdt_phandle_args *pargs,
+			  struct gpio_pin *out_pin);
+
+#endif

include/sbi_utils/gpio/gpio.h

@@ -0,0 +1,107 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Anup Patel <anup.patel@wdc.com>
+ */
+
+#ifndef __GPIO_H__
+#define __GPIO_H__
+
+#include <sbi/sbi_types.h>
+#include <sbi/sbi_list.h>
+
+#define GPIO_LINE_DIRECTION_IN	1
+#define GPIO_LINE_DIRECTION_OUT	0
+
+/** Representation of a GPIO pin */
+struct gpio_pin {
+	/** Pointer to the GPIO chip */
+	struct gpio_chip *chip;
+	/** Identification of GPIO pin within GPIO chip */
+	unsigned int offset;
+	/**
+	 * Additional configuration flags of the GPIO pin desired
+	 * by GPIO clients.
+	 *
+	 * NOTE: GPIO chip can have custom configuration flags.
+	 */
+	unsigned int flags;
+#define GPIO_FLAG_ACTIVE_LOW	0x1
+#define GPIO_FLAG_SINGLE_ENDED	0x2
+#define GPIO_FLAG_OPEN_DRAIN	0x4
+#define GPIO_FLAG_TRANSITORY	0x8
+#define GPIO_FLAG_PULL_UP	0x10
+#define GPIO_FLAG_PULL_DOWN	0x20
+};
+
+/** Representation of a GPIO chip */
+struct gpio_chip {
+	/** Pointer to GPIO driver owning this GPIO chip */
+	void *driver;
+	/** Uniquie ID of the GPIO chip assigned by the driver */
+	unsigned int id;
+	/** Number of GPIOs supported by the GPIO chip */
+	unsigned int ngpio;
+	/**
+	 * Get current direction of GPIO pin
+	 *
+	 * @return 0=output, 1=input, or negative error
+	 */
+	int (*get_direction)(struct gpio_pin *gp);
+	/**
+	 * Set input direction of GPIO pin
+	 *
+	 * @return 0 on success and negative error code on failure
+	 */
+	int (*direction_input)(struct gpio_pin *gp);
+	/**
+	 * Set output direction of GPIO pin with given output value
+	 *
+	 * @return 0 on success and negative error code on failure
+	 */
+	int (*direction_output)(struct gpio_pin *gp, int value);
+	/**
+	 * Get current value of GPIO pin
+	 *
+	 * @return 0=low, 1=high, or negative error
+	 */
+	int (*get)(struct gpio_pin *gp);
+	/** Set output value for GPIO pin */
+	void (*set)(struct gpio_pin *gp, int value);
+	/** List */
+	struct sbi_dlist node;
+};
+
+static inline struct gpio_chip *to_gpio_chip(struct sbi_dlist *node)
+{
+	return container_of(node, struct gpio_chip, node);
+}
+
+/** Find a registered GPIO chip */
+struct gpio_chip *gpio_chip_find(unsigned int id);
+
+/** Register GPIO chip */
+int gpio_chip_add(struct gpio_chip *gc);
+
+/** Un-register GPIO chip */
+void gpio_chip_remove(struct gpio_chip *gc);
+
+/** Get current direction of GPIO pin */
+int gpio_get_direction(struct gpio_pin *gp);
+
+/** Set input direction of GPIO pin */
+int gpio_direction_input(struct gpio_pin *gp);
+
+/** Set output direction of GPIO pin */
+int gpio_direction_output(struct gpio_pin *gp, int value);
+
+/** Get current value of GPIO pin */
+int gpio_get(struct gpio_pin *gp);
+
+/** Set output value of GPIO pin */
+int gpio_set(struct gpio_pin *gp, int value);
+
+#endif

include/sbi_utils/i2c/fdt_i2c.h

@@ -0,0 +1,26 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 YADRO
+ *
+ * Authors:
+ *   Nikita Shubin <n.shubin@yadro.com>
+ */
+
+#ifndef __FDT_I2C_H__
+#define __FDT_I2C_H__
+
+#include <sbi_utils/i2c/i2c.h>
+
+/** FDT based I2C adapter driver */
+struct fdt_i2c_adapter {
+	const struct fdt_match *match_table;
+	int (*init)(void *fdt, int nodeoff,
+		    const struct fdt_match *match);
+};
+
+/** Get I2C adapter identified by nodeoff */
+int fdt_i2c_adapter_get(void *fdt, int nodeoff,
+			struct i2c_adapter **out_adapter);
+
+#endif

include/sbi_utils/i2c/i2c.h

@@ -0,0 +1,85 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 YADRO
+ *
+ * Authors:
+ *   Nikita Shubin <n.shubin@yadro.com>
+ */
+
+#ifndef __I2C_H__
+#define __I2C_H__
+
+#include <sbi/sbi_types.h>
+#include <sbi/sbi_list.h>
+
+/** Representation of a I2C adapter */
+struct i2c_adapter {
+	/** Pointer to I2C driver owning this I2C adapter */
+	void *driver;
+
+	/** Unique ID of the I2C adapter assigned by the driver */
+	int id;
+
+	/**
+	 * Send buffer to given address, register
+	 *
+	 * @return 0 on success and negative error code on failure
+	 */
+	int (*write)(struct i2c_adapter *ia, uint8_t addr, uint8_t reg,
+		    uint8_t *buffer, int len);
+
+	/**
+	 * Read buffer from given address, register
+	 *
+	 * @return 0 on success and negative error code on failure
+	 */
+	int (*read)(struct i2c_adapter *ia, uint8_t addr, uint8_t reg,
+		    uint8_t *buffer, int len);
+
+	/** List */
+	struct sbi_dlist node;
+};
+
+static inline struct i2c_adapter *to_i2c_adapter(struct sbi_dlist *node)
+{
+	return container_of(node, struct i2c_adapter, node);
+}
+
+/** Find a registered I2C adapter */
+struct i2c_adapter *i2c_adapter_find(int id);
+
+/** Register I2C adapter */
+int i2c_adapter_add(struct i2c_adapter *ia);
+
+/** Un-register I2C adapter */
+void i2c_adapter_remove(struct i2c_adapter *ia);
+
+/** Send to device on I2C adapter bus */
+int i2c_adapter_write(struct i2c_adapter *ia, uint8_t addr, uint8_t reg,
+		      uint8_t *buffer, int len);
+
+/** Read from device on I2C adapter bus */
+int i2c_adapter_read(struct i2c_adapter *ia, uint8_t addr, uint8_t reg,
+		     uint8_t *buffer, int len);
+
+static inline int i2c_adapter_reg_write(struct i2c_adapter *ia, uint8_t addr,
+				 uint8_t reg, uint8_t val)
+{
+	return i2c_adapter_write(ia, addr, reg, &val, 1);
+}
+
+static inline int i2c_adapter_reg_read(struct i2c_adapter *ia, uint8_t addr,
+				       uint8_t reg, uint8_t *val)
+{
+	uint8_t buf;
+	int ret = i2c_adapter_read(ia, addr, reg, &buf, 1);
+
+	if (ret)
+		return ret;
+
+	*val = buf;
+	return 0;
+}
+
+#endif

include/sbi_utils/ipi/aclint_mswi.h

@@ -0,0 +1,33 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Anup Patel <anup.patel@wdc.com>
+ */
+
+#ifndef __IPI_ACLINT_MSWI_H__
+#define __IPI_ACLINT_MSWI_H__
+
+#include <sbi/sbi_types.h>
+
+#define ACLINT_MSWI_ALIGN		0x1000
+#define ACLINT_MSWI_SIZE		0x4000
+#define ACLINT_MSWI_MAX_HARTS		4095
+
+#define CLINT_MSWI_OFFSET		0x0000
+
+struct aclint_mswi_data {
+	/* Public details */
+	unsigned long addr;
+	unsigned long size;
+	u32 first_hartid;
+	u32 hart_count;
+};
+
+int aclint_mswi_warm_init(void);
+
+int aclint_mswi_cold_init(struct aclint_mswi_data *mswi);
+
+#endif

include/sbi_utils/ipi/fdt_ipi.h

@@ -17,14 +17,8 @@ struct fdt_ipi {
 	int (*cold_init)(void *fdt, int nodeoff, const struct fdt_match *match);
 	int (*warm_init)(void);
 	void (*exit)(void);
-	void (*send)(u32 target_hart);
-	void (*clear)(u32 target_hart);
 };
 
-void fdt_ipi_send(u32 target_hart);
-
-void fdt_ipi_clear(u32 target_hart);
-
 void fdt_ipi_exit(void);
 
 int fdt_ipi_init(bool cold_boot);

include/sbi_utils/reset/fdt_reset.h

@@ -15,14 +15,18 @@
 struct fdt_reset {
 	const struct fdt_match *match_table;
 	int (*init)(void *fdt, int nodeoff, const struct fdt_match *match);
-	int (*system_reset_check)(u32 reset_type, u32 reset_reason);
-	void (*system_reset)(u32 reset_type, u32 reset_reason);
 };
 
-int fdt_system_reset_check(u32 reset_type, u32 reset_reason);
+/**
+ * fdt_reset_driver_init() - initialize reset driver based on the device-tree
+ */
+int fdt_reset_driver_init(void *fdt, struct fdt_reset *drv);
 
-void fdt_system_reset(u32 reset_type, u32 reset_reason);
-
-int fdt_reset_init(void);
+/**
+ * fdt_reset_init() - initialize reset drivers based on the device-tree
+ *
+ * This function shall be invoked in final init.
+ */
+void fdt_reset_init(void);
 
 #endif

include/sbi_utils/serial/fdt_serial.h

@@ -15,14 +15,8 @@
 struct fdt_serial {
 	const struct fdt_match *match_table;
 	int (*init)(void *fdt, int nodeoff, const struct fdt_match *match);
-	void (*putc)(char ch);
-	int (*getc)(void);
 };
 
-void fdt_serial_putc(char ch);
-
-int fdt_serial_getc(void);
-
 int fdt_serial_init(void);
 
 #endif

include/sbi_utils/serial/gaisler-uart.h

@@ -0,0 +1,17 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Cobham Gaisler AB.
+ *
+ * Authors:
+ *   Daniel Cederman <cederman@gaisler.com>
+ */
+
+#ifndef __SERIAL_GAISLER_APBUART_H__
+#define __SERIAL_GAISLER_APBUART_H__
+
+#include <sbi/sbi_types.h>
+
+int gaisler_uart_init(unsigned long base, u32 in_freq, u32 baudrate);
+
+#endif

include/sbi_utils/serial/litex-uart.h

@@ -0,0 +1,17 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Gabriel Somlo
+ *
+ * Authors:
+ *   Gabriel Somlo <gsomlo@gmail.com>
+ */
+
+#ifndef __SERIAL_LITEX_UART_H__
+#define __SERIAL_LITEX_UART_H__
+
+#include <sbi/sbi_types.h>
+
+int litex_uart_init(unsigned long base);
+
+#endif

include/sbi_utils/serial/shakti-uart.h

@@ -9,10 +9,6 @@
 
 #include <sbi/sbi_types.h>
 
-void shakti_uart_putc(char ch);
-
-int shakti_uart_getc(void);
-
 int shakti_uart_init(unsigned long base, u32 in_freq, u32 baudrate);
 
 #endif

include/sbi_utils/serial/sifive-uart.h

@@ -12,10 +12,6 @@
 
 #include <sbi/sbi_types.h>
 
-void sifive_uart_putc(char ch);
-
-int sifive_uart_getc(void);
-
 int sifive_uart_init(unsigned long base, u32 in_freq, u32 baudrate);
 
 #endif

include/sbi_utils/serial/uart8250.h

@@ -12,10 +12,6 @@
 
 #include <sbi/sbi_types.h>
 
-void uart8250_putc(char ch);
-
-int uart8250_getc(void);
-
 int uart8250_init(unsigned long base, u32 in_freq, u32 baudrate, u32 reg_shift,
 		  u32 reg_width);
 

include/sbi_utils/sys/clint.h

@@ -1,51 +0,0 @@
-/*
- * SPDX-License-Identifier: BSD-2-Clause
- *
- * Copyright (c) 2019 Western Digital Corporation or its affiliates.
- *
- * Authors:
- *   Anup Patel <anup.patel@wdc.com>
- */
-
-#ifndef __SYS_CLINT_H__
-#define __SYS_CLINT_H__
-
-#include <sbi/sbi_types.h>
-
-struct clint_data {
-	/* Public details */
-	unsigned long addr;
-	u32 first_hartid;
-	u32 hart_count;
-	bool has_64bit_mmio;
-	/* Private details (initialized and used by CLINT library)*/
-	u32 *ipi;
-	struct clint_data *time_delta_reference;
-	unsigned long time_delta_computed;
-	u64 time_delta;
-	u64 *time_val;
-	u64 *time_cmp;
-	u64 (*time_rd)(volatile u64 *addr);
-	void (*time_wr)(u64 value, volatile u64 *addr);
-};
-
-void clint_ipi_send(u32 target_hart);
-
-void clint_ipi_clear(u32 target_hart);
-
-int clint_warm_ipi_init(void);
-
-int clint_cold_ipi_init(struct clint_data *clint);
-
-u64 clint_timer_value(void);
-
-void clint_timer_event_stop(void);
-
-void clint_timer_event_start(u64 next_event);
-
-int clint_warm_timer_init(void);
-
-int clint_cold_timer_init(struct clint_data *clint,
-			  struct clint_data *reference);
-
-#endif

include/sbi_utils/sys/htif.h

@@ -10,12 +10,8 @@
 
 #include <sbi/sbi_types.h>
 
-void htif_putc(char ch);
+int htif_serial_init(void);
 
-int htif_getc(void);
-
-int htif_system_reset_check(u32 type, u32 reason);
-
-void htif_system_reset(u32 type, u32 reason);
+int htif_system_reset_init(void);
 
 #endif

include/sbi_utils/sys/sifive_test.h

@@ -12,10 +12,6 @@
 
 #include <sbi/sbi_types.h>
 
-int sifive_test_system_reset_check(u32 type, u32 reason);
-
-void sifive_test_system_reset(u32 type, u32 reason);
-
 int sifive_test_init(unsigned long base);
 
 #endif

include/sbi_utils/timer/aclint_mtimer.h

@@ -0,0 +1,53 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Anup Patel <anup.patel@wdc.com>
+ */
+
+#ifndef __TIMER_ACLINT_MTIMER_H__
+#define __TIMER_ACLINT_MTIMER_H__
+
+#include <sbi/sbi_types.h>
+
+#define ACLINT_MTIMER_ALIGN		0x8
+#define ACLINT_MTIMER_MAX_HARTS		4095
+
+#define ACLINT_DEFAULT_MTIME_OFFSET	0x7ff8
+#define ACLINT_DEFAULT_MTIME_SIZE	0x8
+#define ACLINT_DEFAULT_MTIMECMP_OFFSET	0x0000
+#define ACLINT_DEFAULT_MTIMECMP_SIZE	0x7ff8
+
+#define CLINT_MTIMER_OFFSET		0x4000
+
+struct aclint_mtimer_data {
+	/* Public details */
+	unsigned long mtime_freq;
+	unsigned long mtime_addr;
+	unsigned long mtime_size;
+	unsigned long mtimecmp_addr;
+	unsigned long mtimecmp_size;
+	u32 first_hartid;
+	u32 hart_count;
+	bool has_64bit_mmio;
+	bool has_shared_mtime;
+	/* Private details (initialized and used by ACLINT MTIMER library) */
+	struct aclint_mtimer_data *time_delta_reference;
+	unsigned long time_delta_computed;
+	u64 (*time_rd)(volatile u64 *addr);
+	void (*time_wr)(bool timecmp, u64 value, volatile u64 *addr);
+};
+
+void aclint_mtimer_sync(struct aclint_mtimer_data *mt);
+
+void aclint_mtimer_set_reference(struct aclint_mtimer_data *mt,
+				 struct aclint_mtimer_data *ref);
+
+int aclint_mtimer_warm_init(void);
+
+int aclint_mtimer_cold_init(struct aclint_mtimer_data *mt,
+			    struct aclint_mtimer_data *reference);
+
+#endif

include/sbi_utils/timer/fdt_timer.h

@@ -17,17 +17,8 @@ struct fdt_timer {
 	int (*cold_init)(void *fdt, int nodeoff, const struct fdt_match *match);
 	int (*warm_init)(void);
 	void (*exit)(void);
-	u64 (*value)(void);
-	void (*event_stop)(void);
-	void (*event_start)(u64 next_event);
 };
 
-u64 fdt_timer_value(void);
-
-void fdt_timer_event_stop(void);
-
-void fdt_timer_event_start(u64 next_event);
-
 void fdt_timer_exit(void);
 
 int fdt_timer_init(bool cold_boot);

lib/sbi/objects.mk

@@ -20,6 +20,7 @@ libsbi-objs-y += sbi_ecall.o
 libsbi-objs-y += sbi_ecall_base.o
 libsbi-objs-y += sbi_ecall_hsm.o
 libsbi-objs-y += sbi_ecall_legacy.o
+libsbi-objs-y += sbi_ecall_pmu.o
 libsbi-objs-y += sbi_ecall_replace.o
 libsbi-objs-y += sbi_ecall_vendor.o
 libsbi-objs-y += sbi_emulate_csr.o
@@ -33,6 +34,7 @@ libsbi-objs-y += sbi_init.o
 libsbi-objs-y += sbi_ipi.o
 libsbi-objs-y += sbi_misaligned_ldst.o
 libsbi-objs-y += sbi_platform.o
+libsbi-objs-y += sbi_pmu.o
 libsbi-objs-y += sbi_scratch.o
 libsbi-objs-y += sbi_string.o
 libsbi-objs-y += sbi_system.o

lib/sbi/riscv_asm.c

@@ -11,6 +11,7 @@
 #include <sbi/riscv_encoding.h>
 #include <sbi/sbi_error.h>
 #include <sbi/sbi_platform.h>
+#include <sbi/sbi_console.h>
 
 /* determine CPU extension, return non-zero support */
 int misa_extension_imp(char ext)
@@ -75,6 +76,8 @@ void misa_string(int xlen, char *out, unsigned int out_sz)
 			out[pos++] = '8';
 			break;
 		default:
+			sbi_panic("%s: Unknown misa.MXL encoding %d",
+				   __func__, xlen);
 			return;
 		}
 	}
@@ -118,7 +121,32 @@ unsigned long csr_read_num(int csr_num)
 	switch (csr_num) {
 	switchcase_csr_read_16(CSR_PMPCFG0, ret)
 	switchcase_csr_read_64(CSR_PMPADDR0, ret)
+	switchcase_csr_read(CSR_MCYCLE, ret)
+	switchcase_csr_read(CSR_MINSTRET, ret)
+	switchcase_csr_read(CSR_MHPMCOUNTER3, ret)
+	switchcase_csr_read_4(CSR_MHPMCOUNTER4, ret)
+	switchcase_csr_read_8(CSR_MHPMCOUNTER8, ret)
+	switchcase_csr_read_16(CSR_MHPMCOUNTER16, ret)
+	switchcase_csr_read(CSR_MCOUNTINHIBIT, ret)
+	switchcase_csr_read(CSR_MHPMEVENT3, ret)
+	switchcase_csr_read_4(CSR_MHPMEVENT4, ret)
+	switchcase_csr_read_8(CSR_MHPMEVENT8, ret)
+	switchcase_csr_read_16(CSR_MHPMEVENT16, ret)
+#if __riscv_xlen == 32
+	switchcase_csr_read(CSR_MCYCLEH, ret)
+	switchcase_csr_read(CSR_MINSTRETH, ret)
+	switchcase_csr_read(CSR_MHPMCOUNTER3H, ret)
+	switchcase_csr_read_4(CSR_MHPMCOUNTER4H, ret)
+	switchcase_csr_read_8(CSR_MHPMCOUNTER8H, ret)
+	switchcase_csr_read_16(CSR_MHPMCOUNTER16H, ret)
+	switchcase_csr_read(CSR_MHPMEVENT3H, ret)
+	switchcase_csr_read_4(CSR_MHPMEVENT4H, ret)
+	switchcase_csr_read_8(CSR_MHPMEVENT8H, ret)
+	switchcase_csr_read_16(CSR_MHPMEVENT16H, ret)
+#endif
+
 	default:
+		sbi_panic("%s: Unknown CSR %#x", __func__, csr_num);
 		break;
 	};
 
@@ -161,7 +189,32 @@ void csr_write_num(int csr_num, unsigned long val)
 	switch (csr_num) {
 	switchcase_csr_write_16(CSR_PMPCFG0, val)
 	switchcase_csr_write_64(CSR_PMPADDR0, val)
+	switchcase_csr_write(CSR_MCYCLE, val)
+	switchcase_csr_write(CSR_MINSTRET, val)
+	switchcase_csr_write(CSR_MHPMCOUNTER3, val)
+	switchcase_csr_write_4(CSR_MHPMCOUNTER4, val)
+	switchcase_csr_write_8(CSR_MHPMCOUNTER8, val)
+	switchcase_csr_write_16(CSR_MHPMCOUNTER16, val)
+#if __riscv_xlen == 32
+	switchcase_csr_write(CSR_MCYCLEH, val)
+	switchcase_csr_write(CSR_MINSTRETH, val)
+	switchcase_csr_write(CSR_MHPMCOUNTER3H, val)
+	switchcase_csr_write_4(CSR_MHPMCOUNTER4H, val)
+	switchcase_csr_write_8(CSR_MHPMCOUNTER8H, val)
+	switchcase_csr_write_16(CSR_MHPMCOUNTER16H, val)
+	switchcase_csr_write(CSR_MHPMEVENT3H, val)
+	switchcase_csr_write_4(CSR_MHPMEVENT4H, val)
+	switchcase_csr_write_8(CSR_MHPMEVENT8H, val)
+	switchcase_csr_write_16(CSR_MHPMEVENT16H, val)
+#endif
+	switchcase_csr_write(CSR_MCOUNTINHIBIT, val)
+	switchcase_csr_write(CSR_MHPMEVENT3, val)
+	switchcase_csr_write_4(CSR_MHPMEVENT4, val)
+	switchcase_csr_write_8(CSR_MHPMEVENT8, val)
+	switchcase_csr_write_16(CSR_MHPMEVENT16, val)
+
 	default:
+		sbi_panic("%s: Unknown CSR %#x", __func__, csr_num);
 		break;
 	};
 
@@ -178,6 +231,9 @@ static unsigned long ctz(unsigned long x)
 {
 	unsigned long ret = 0;
 
+	if (x == 0)
+		return 8 * sizeof(x);
+
 	while (!(x & 1UL)) {
 		ret++;
 		x = x >> 1;
@@ -205,14 +261,12 @@ int pmp_set(unsigned int n, unsigned long prot, unsigned long addr,
 	pmpcfg_csr   = (CSR_PMPCFG0 + (n >> 2)) & ~1;
 	pmpcfg_shift = (n & 7) << 3;
 #else
-	pmpcfg_csr   = -1;
-	pmpcfg_shift = -1;
+	return SBI_ENOTSUPP;
 #endif
 	pmpaddr_csr = CSR_PMPADDR0 + n;
-	if (pmpcfg_csr < 0 || pmpcfg_shift < 0)
-		return SBI_ENOTSUPP;
 
 	/* encode PMP config */
+	prot &= ~PMP_A;
 	prot |= (log2len == PMP_SHIFT) ? PMP_A_NA4 : PMP_A_NAPOT;
 	cfgmask = ~(0xffUL << pmpcfg_shift);
 	pmpcfg	= (csr_read_num(pmpcfg_csr) & cfgmask);
@@ -258,12 +312,9 @@ int pmp_get(unsigned int n, unsigned long *prot_out, unsigned long *addr_out,
 	pmpcfg_csr   = (CSR_PMPCFG0 + (n >> 2)) & ~1;
 	pmpcfg_shift = (n & 7) << 3;
 #else
-	pmpcfg_csr   = -1;
-	pmpcfg_shift = -1;
+	return SBI_ENOTSUPP;
 #endif
 	pmpaddr_csr = CSR_PMPADDR0 + n;
-	if (pmpcfg_csr < 0 || pmpcfg_shift < 0)
-		return SBI_ENOTSUPP;
 
 	/* decode PMP config */
 	cfgmask = (0xffUL << pmpcfg_shift);

lib/sbi/riscv_atomic.c

@@ -28,25 +28,23 @@ void atomic_write(atomic_t *atom, long value)
 long atomic_add_return(atomic_t *atom, long value)
 {
 	long ret;
-
+#if __SIZEOF_LONG__ == 4
 	__asm__ __volatile__("	amoadd.w.aqrl  %1, %2, %0"
 			     : "+A"(atom->counter), "=r"(ret)
 			     : "r"(value)
 			     : "memory");
-
+#elif __SIZEOF_LONG__ == 8
+	__asm__ __volatile__("	amoadd.d.aqrl  %1, %2, %0"
+			     : "+A"(atom->counter), "=r"(ret)
+			     : "r"(value)
+			     : "memory");
+#endif
 	return ret + value;
 }
 
 long atomic_sub_return(atomic_t *atom, long value)
 {
-	long ret;
-
-	__asm__ __volatile__("	amoadd.w.aqrl  %1, %2, %0"
-			     : "+A"(atom->counter), "=r"(ret)
-			     : "r"(-value)
-			     : "memory");
-
-	return ret - value;
+	return atomic_add_return(atom, -value);
 }
 
 #define __axchg(ptr, new, size)							\

lib/sbi/riscv_locks.c

@@ -2,44 +2,76 @@
  * SPDX-License-Identifier: BSD-2-Clause
  *
  * Copyright (c) 2019 Western Digital Corporation or its affiliates.
- *
- * Authors:
- *   Anup Patel <anup.patel@wdc.com>
+ * Copyright (c) 2021 Christoph Müllner <cmuellner@linux.com>
  */
 
 #include <sbi/riscv_barrier.h>
 #include <sbi/riscv_locks.h>
 
-int spin_lock_check(spinlock_t *lock)
+static inline bool spin_lock_unlocked(spinlock_t lock)
 {
-	return (lock->lock == __RISCV_SPIN_UNLOCKED) ? 0 : 1;
+	return lock.owner == lock.next;
 }
 
-int spin_trylock(spinlock_t *lock)
+bool spin_lock_check(spinlock_t *lock)
 {
-	int tmp = 1, busy;
+	RISCV_FENCE(r, rw);
+	return !spin_lock_unlocked(*lock);
+}
+
+bool spin_trylock(spinlock_t *lock)
+{
+	unsigned long inc = 1u << TICKET_SHIFT;
+	unsigned long mask = 0xffffu << TICKET_SHIFT;
+	u32 l0, tmp1, tmp2;
 
 	__asm__ __volatile__(
-		"	amoswap.w %0, %2, %1\n" RISCV_ACQUIRE_BARRIER
-		: "=r"(busy), "+A"(lock->lock)
-		: "r"(tmp)
+		/* Get the current lock counters. */
+		"1:	lr.w.aq	%0, %3\n"
+		"	slli	%2, %0, %6\n"
+		"	and	%2, %2, %5\n"
+		"	and	%1, %0, %5\n"
+		/* Is the lock free right now? */
+		"	bne	%1, %2, 2f\n"
+		"	add	%0, %0, %4\n"
+		/* Acquire the lock. */
+		"	sc.w.rl	%0, %0, %3\n"
+		"	bnez	%0, 1b\n"
+		"2:"
+		: "=&r"(l0), "=&r"(tmp1), "=&r"(tmp2), "+A"(*lock)
+		: "r"(inc), "r"(mask), "I"(TICKET_SHIFT)
 		: "memory");
 
-	return !busy;
+	return l0 == 0;
 }
 
 void spin_lock(spinlock_t *lock)
 {
-	while (1) {
-		if (spin_lock_check(lock))
-			continue;
+	unsigned long inc = 1u << TICKET_SHIFT;
+	unsigned long mask = 0xffffu;
+	u32 l0, tmp1, tmp2;
 
-		if (spin_trylock(lock))
-			break;
-	}
+	__asm__ __volatile__(
+		/* Atomically increment the next ticket. */
+		"	amoadd.w.aqrl	%0, %4, %3\n"
+
+		/* Did we get the lock? */
+		"	srli	%1, %0, %6\n"
+		"	and	%1, %1, %5\n"
+		"1:	and	%2, %0, %5\n"
+		"	beq	%1, %2, 2f\n"
+
+		/* If not, then spin on the lock. */
+		"	lw	%0, %3\n"
+		RISCV_ACQUIRE_BARRIER
+		"	j	1b\n"
+		"2:"
+		: "=&r"(l0), "=&r"(tmp1), "=&r"(tmp2), "+A"(*lock)
+		: "r"(inc), "r"(mask), "I"(TICKET_SHIFT)
+		: "memory");
 }
 
 void spin_unlock(spinlock_t *lock)
 {
-	__smp_store_release(&lock->lock, __RISCV_SPIN_UNLOCKED);
+	__smp_store_release(&lock->owner, lock->owner + 1);
 }

lib/sbi/sbi_console.c

@@ -9,10 +9,11 @@
 
 #include <sbi/riscv_locks.h>
 #include <sbi/sbi_console.h>
+#include <sbi/sbi_hart.h>
 #include <sbi/sbi_platform.h>
 #include <sbi/sbi_scratch.h>
 
-static const struct sbi_platform *console_plat = NULL;
+static const struct sbi_console_device *console_dev = NULL;
 static spinlock_t console_out_lock	       = SPIN_LOCK_INITIALIZER;
 
 bool sbi_isprintable(char c)
@@ -26,14 +27,18 @@ bool sbi_isprintable(char c)
 
 int sbi_getc(void)
 {
-	return sbi_platform_console_getc(console_plat);
+	if (console_dev && console_dev->console_getc)
+		return console_dev->console_getc();
+	return -1;
 }
 
 void sbi_putc(char ch)
 {
+	if (console_dev && console_dev->console_putc) {
 		if (ch == '\n')
-		sbi_platform_console_putc(console_plat, '\r');
-	sbi_platform_console_putc(console_plat, ch);
+			console_dev->console_putc('\r');
+		console_dev->console_putc(ch);
+	}
 }
 
 void sbi_puts(const char *str)
@@ -383,16 +388,43 @@ int sbi_dprintf(const char *format, ...)
 	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
 
 	va_start(args, format);
-	if (scratch->options & SBI_SCRATCH_DEBUG_PRINTS)
+	if (scratch->options & SBI_SCRATCH_DEBUG_PRINTS) {
+		spin_lock(&console_out_lock);
 		retval = print(NULL, NULL, format, args);
+		spin_unlock(&console_out_lock);
+	}
 	va_end(args);
 
 	return retval;
 }
 
+void sbi_panic(const char *format, ...)
+{
+	va_list args;
+
+	spin_lock(&console_out_lock);
+	va_start(args, format);
+	print(NULL, NULL, format, args);
+	va_end(args);
+	spin_unlock(&console_out_lock);
+
+	sbi_hart_hang();
+}
+
+const struct sbi_console_device *sbi_console_get_device(void)
+{
+	return console_dev;
+}
+
+void sbi_console_set_device(const struct sbi_console_device *dev)
+{
+	if (!dev || console_dev)
+		return;
+
+	console_dev = dev;
+}
+
 int sbi_console_init(struct sbi_scratch *scratch)
 {
-	console_plat = sbi_platform_ptr(scratch);
-
-	return sbi_platform_console_init(console_plat);
+	return sbi_platform_console_init(sbi_platform_ptr(scratch));
 }

lib/sbi/sbi_domain.c

@@ -19,17 +19,17 @@
 
 struct sbi_domain *hartid_to_domain_table[SBI_HARTMASK_MAX_BITS] = { 0 };
 struct sbi_domain *domidx_to_domain_table[SBI_DOMAIN_MAX_INDEX] = { 0 };
-
 static u32 domain_count = 0;
+static bool domain_finalized = false;
 
 static struct sbi_hartmask root_hmask = { 0 };
 
-#define ROOT_FW_REGION		0
-#define ROOT_ALL_REGION	1
-#define ROOT_END_REGION	2
-static struct sbi_domain_memregion root_memregs[ROOT_END_REGION + 1] = { 0 };
+#define ROOT_REGION_MAX	16
+static u32 root_memregs_count = 0;
+static struct sbi_domain_memregion root_fw_region;
+static struct sbi_domain_memregion root_memregs[ROOT_REGION_MAX + 1] = { 0 };
 
-static struct sbi_domain root = {
+struct sbi_domain root = {
 	.name = "root",
 	.possible_harts = &root_hmask,
 	.regions = root_memregs,
@@ -64,12 +64,41 @@ ulong sbi_domain_get_assigned_hartmask(const struct sbi_domain *dom,
 	return ret;
 }
 
-void sbi_domain_memregion_initfw(struct sbi_domain_memregion *reg)
+static void domain_memregion_initfw(struct sbi_domain_memregion *reg)
 {
 	if (!reg)
 		return;
 
-	sbi_memcpy(reg, &root_memregs[ROOT_FW_REGION], sizeof(*reg));
+	sbi_memcpy(reg, &root_fw_region, sizeof(*reg));
+}
+
+void sbi_domain_memregion_init(unsigned long addr,
+				unsigned long size,
+				unsigned long flags,
+				struct sbi_domain_memregion *reg)
+{
+	unsigned long base = 0, order;
+
+	for (order = log2roundup(size) ; order <= __riscv_xlen; order++) {
+		if (order < __riscv_xlen) {
+			base = addr & ~((1UL << order) - 1UL);
+			if ((base <= addr) &&
+			    (addr < (base + (1UL << order))) &&
+			    (base <= (addr + size - 1UL)) &&
+			    ((addr + size - 1UL) < (base + (1UL << order))))
+				break;
+		} else {
+			base = 0;
+			break;
+		}
+
+	}
+
+	if (reg) {
+		reg->base = base;
+		reg->order = order;
+		reg->flags = flags;
+	}
 }
 
 bool sbi_domain_check_addr(const struct sbi_domain *dom,
@@ -207,9 +236,9 @@ static int sanitize_domain(const struct sbi_platform *plat,
 	count = 0;
 	have_fw_reg = FALSE;
 	sbi_domain_for_each_memregion(dom, reg) {
-		if (reg->order == root_memregs[ROOT_FW_REGION].order &&
-		    reg->base == root_memregs[ROOT_FW_REGION].base &&
-		    reg->flags == root_memregs[ROOT_FW_REGION].flags)
+		if (reg->order == root_fw_region.order &&
+		    reg->base == root_fw_region.base &&
+		    reg->flags == root_fw_region.flags)
 			have_fw_reg = TRUE;
 		count++;
 	}
@@ -266,7 +295,7 @@ static int sanitize_domain(const struct sbi_platform *plat,
 	/* Check next address and next mode*/
 	if (!sbi_domain_check_addr(dom, dom->next_addr, dom->next_mode,
 				   SBI_DOMAIN_EXECUTE)) {
-		sbi_printf("%s: %s next booting stage addres 0x%lx can't "
+		sbi_printf("%s: %s next booting stage address 0x%lx can't "
 			   "execute\n", __func__, dom->name, dom->next_addr);
 		return SBI_EINVAL;
 	}
@@ -299,11 +328,7 @@ void sbi_domain_dump(const struct sbi_domain *dom, const char *suffix)
 		rend = (reg->order < __riscv_xlen) ?
 			rstart + ((1UL << reg->order) - 1) : -1UL;
 
-#if __riscv_xlen == 32
-		sbi_printf("Domain%d Region%02d    %s: 0x%08lx-0x%08lx ",
-#else
-		sbi_printf("Domain%d Region%02d    %s: 0x%016lx-0x%016lx ",
-#endif
+		sbi_printf("Domain%d Region%02d    %s: 0x%" PRILX "-0x%" PRILX " ",
 			   dom->index, i, suffix, rstart, rend);
 
 		k = 0;
@@ -322,18 +347,10 @@ void sbi_domain_dump(const struct sbi_domain *dom, const char *suffix)
 		i++;
 	}
 
-#if __riscv_xlen == 32
-	sbi_printf("Domain%d Next Address%s: 0x%08lx\n",
-#else
-	sbi_printf("Domain%d Next Address%s: 0x%016lx\n",
-#endif
+	sbi_printf("Domain%d Next Address%s: 0x%" PRILX "\n",
 		   dom->index, suffix, dom->next_addr);
 
-#if __riscv_xlen == 32
-	sbi_printf("Domain%d Next Arg1   %s: 0x%08lx\n",
-#else
-	sbi_printf("Domain%d Next Arg1   %s: 0x%016lx\n",
-#endif
+	sbi_printf("Domain%d Next Arg1   %s: 0x%" PRILX "\n",
 		   dom->index, suffix, dom->next_arg1);
 
 	sbi_printf("Domain%d Next Mode   %s: ", dom->index, suffix);
@@ -376,7 +393,8 @@ int sbi_domain_register(struct sbi_domain *dom,
 	u32 cold_hartid = current_hartid();
 	const struct sbi_platform *plat = sbi_platform_thishart_ptr();
 
-	if (!dom || !assign_mask)
+	/* Sanity checks */
+	if (!dom || !assign_mask || domain_finalized)
 		return SBI_EINVAL;
 
 	/* Check if domain already discovered */
@@ -438,6 +456,68 @@ int sbi_domain_register(struct sbi_domain *dom,
 	return 0;
 }
 
+int sbi_domain_root_add_memregion(const struct sbi_domain_memregion *reg)
+{
+	int rc;
+	bool reg_merged;
+	struct sbi_domain_memregion *nreg, *nreg1, *nreg2;
+	const struct sbi_platform *plat = sbi_platform_thishart_ptr();
+
+	/* Sanity checks */
+	if (!reg || domain_finalized ||
+	    (root.regions != root_memregs) ||
+	    (ROOT_REGION_MAX <= root_memregs_count))
+		return SBI_EINVAL;
+
+	/* Check for conflicts */
+	sbi_domain_for_each_memregion(&root, nreg) {
+		if (is_region_conflict(reg, nreg))
+			return SBI_EINVAL;
+	}
+
+	/* Append the memregion to root memregions */
+	nreg = &root_memregs[root_memregs_count];
+	sbi_memcpy(nreg, reg, sizeof(*reg));
+	root_memregs_count++;
+	root_memregs[root_memregs_count].order = 0;
+
+	/* Sort and optimize root regions */
+	do {
+		/* Sanitize the root domain so that memregions are sorted */
+		rc = sanitize_domain(plat, &root);
+		if (rc) {
+			sbi_printf("%s: sanity checks failed for"
+				   " %s (error %d)\n", __func__,
+				   root.name, rc);
+			return rc;
+		}
+
+		/* Merge consecutive memregions with same order and flags */
+		reg_merged = false;
+		sbi_domain_for_each_memregion(&root, nreg) {
+			nreg1 = nreg + 1;
+			if (!nreg1->order)
+				continue;
+
+			if (!(nreg->base & (BIT(nreg->order + 1) - 1)) &&
+			    (nreg->base + BIT(nreg->order)) == nreg1->base &&
+			    nreg->order == nreg1->order &&
+			    nreg->flags == nreg1->flags) {
+				nreg->order++;
+				while (nreg1->order) {
+					nreg2 = nreg1 + 1;
+					sbi_memcpy(nreg1, nreg2, sizeof(*nreg1));
+					nreg1++;
+				}
+				reg_merged = true;
+				root_memregs_count--;
+			}
+		}
+	} while (reg_merged);
+
+	return 0;
+}
+
 int sbi_domain_finalize(struct sbi_scratch *scratch, u32 cold_hartid)
 {
 	int rc;
@@ -490,35 +570,34 @@ int sbi_domain_finalize(struct sbi_scratch *scratch, u32 cold_hartid)
 		}
 	}
 
+	/*
+	 * Set the finalized flag so that the root domain
+	 * regions can't be changed.
+	 */
+	domain_finalized = true;
+
 	return 0;
 }
 
 int sbi_domain_init(struct sbi_scratch *scratch, u32 cold_hartid)
 {
 	u32 i;
-	struct sbi_domain_memregion *memregs;
 	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 
 	/* Root domain firmware memory region */
-	root_memregs[ROOT_FW_REGION].order = log2roundup(scratch->fw_size);
-	root_memregs[ROOT_FW_REGION].base = scratch->fw_start &
-				~((1UL << root_memregs[0].order) - 1UL);
-	root_memregs[ROOT_FW_REGION].flags = 0;
+	sbi_domain_memregion_init(scratch->fw_start, scratch->fw_size, 0,
+				  &root_fw_region);
+	domain_memregion_initfw(&root_memregs[root_memregs_count++]);
 
 	/* Root domain allow everything memory region */
-	root_memregs[ROOT_ALL_REGION].order = __riscv_xlen;
-	root_memregs[ROOT_ALL_REGION].base = 0;
-	root_memregs[ROOT_ALL_REGION].flags = (SBI_DOMAIN_MEMREGION_READABLE |
+	sbi_domain_memregion_init(0, ~0UL,
+				  (SBI_DOMAIN_MEMREGION_READABLE |
 				   SBI_DOMAIN_MEMREGION_WRITEABLE |
-						SBI_DOMAIN_MEMREGION_EXECUTABLE);
+				   SBI_DOMAIN_MEMREGION_EXECUTABLE),
+				  &root_memregs[root_memregs_count++]);
 
 	/* Root domain memory region end */
-	root_memregs[ROOT_END_REGION].order = 0;
-
-	/* Use platform specific root memory regions when available */
-	memregs = sbi_platform_domains_root_regions(plat);
-	if (memregs)
-		root.regions = memregs;
+	root_memregs[root_memregs_count].order = 0;
 
 	/* Root domain boot HART id is same as coldboot HART id */
 	root.boot_hartid = cold_hartid;

lib/sbi/sbi_ecall.c

@@ -162,6 +162,9 @@ int sbi_ecall_init(void)
 	if (ret)
 		return ret;
 	ret = sbi_ecall_register_extension(&ecall_srst);
+	if (ret)
+		return ret;
+	ret = sbi_ecall_register_extension(&ecall_pmu);
 	if (ret)
 		return ret;
 	ret = sbi_ecall_register_extension(&ecall_legacy);

lib/sbi/sbi_ecall_hsm.c

@@ -22,14 +22,13 @@ static int sbi_ecall_hsm_handler(unsigned long extid, unsigned long funcid,
 				 unsigned long *out_val,
 				 struct sbi_trap_info *out_trap)
 {
-	ulong smode;
-	int ret = 0, hstate;
+	int ret = 0;
 	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
+	ulong smode = (csr_read(CSR_MSTATUS) & MSTATUS_MPP) >>
+			MSTATUS_MPP_SHIFT;
 
 	switch (funcid) {
 	case SBI_EXT_HSM_HART_START:
-		smode = csr_read(CSR_MSTATUS);
-		smode = (smode & MSTATUS_MPP) >> MSTATUS_MPP_SHIFT;
 		ret = sbi_hsm_hart_start(scratch, sbi_domain_thishart_ptr(),
 					 regs->a0, regs->a1, smode, regs->a2);
 		break;
@@ -37,9 +36,12 @@ static int sbi_ecall_hsm_handler(unsigned long extid, unsigned long funcid,
 		ret = sbi_hsm_hart_stop(scratch, TRUE);
 		break;
 	case SBI_EXT_HSM_HART_GET_STATUS:
-		hstate = sbi_hsm_hart_get_state(sbi_domain_thishart_ptr(),
+		ret = sbi_hsm_hart_get_state(sbi_domain_thishart_ptr(),
 					     regs->a0);
-		ret = sbi_hsm_hart_state_to_status(hstate);
+		break;
+	case SBI_EXT_HSM_HART_SUSPEND:
+		ret = sbi_hsm_hart_suspend(scratch, regs->a0, regs->a1,
+					   smode, regs->a2);
 		break;
 	default:
 		ret = SBI_ENOTSUPP;

lib/sbi/sbi_ecall_legacy.c

@@ -34,7 +34,7 @@ static int sbi_load_hart_mask_unpriv(ulong *pmask, ulong *hmask,
 		if (uptrap->cause)
 			return SBI_ETRAP;
 	} else {
-		sbi_hsm_hart_started_mask(sbi_domain_thishart_ptr(),
+		sbi_hsm_hart_interruptible_mask(sbi_domain_thishart_ptr(),
 						0, &mask);
 	}
 	*hmask = mask;

lib/sbi/sbi_ecall_pmu.c

@@ -0,0 +1,87 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Atish Patra <atish.patra@wdc.com>
+ */
+
+#include <sbi/sbi_ecall.h>
+#include <sbi/sbi_ecall_interface.h>
+#include <sbi/sbi_error.h>
+#include <sbi/sbi_hart.h>
+#include <sbi/sbi_trap.h>
+#include <sbi/sbi_version.h>
+#include <sbi/sbi_pmu.h>
+#include <sbi/sbi_scratch.h>
+#include <sbi/riscv_asm.h>
+
+static int sbi_ecall_pmu_handler(unsigned long extid, unsigned long funcid,
+				 const struct sbi_trap_regs *regs,
+				 unsigned long *out_val,
+				 struct sbi_trap_info *out_trap)
+{
+	int ret = 0;
+	uint64_t temp;
+
+	switch (funcid) {
+	case SBI_EXT_PMU_NUM_COUNTERS:
+		ret = sbi_pmu_num_ctr();
+		if (ret >= 0) {
+			*out_val = ret;
+			ret = 0;
+		}
+		break;
+	case SBI_EXT_PMU_COUNTER_GET_INFO:
+		ret = sbi_pmu_ctr_get_info(regs->a0, out_val);
+		break;
+	case SBI_EXT_PMU_COUNTER_CFG_MATCH:
+#if __riscv_xlen == 32
+		temp = ((uint64_t)regs->a5 << 32) | regs->a4;
+#else
+		temp = regs->a4;
+#endif
+		ret = sbi_pmu_ctr_cfg_match(regs->a0, regs->a1, regs->a2,
+					    regs->a3, temp);
+		if (ret >= 0) {
+			*out_val = ret;
+			ret = 0;
+		}
+
+		break;
+	case SBI_EXT_PMU_COUNTER_FW_READ:
+		ret = sbi_pmu_ctr_read(regs->a0, out_val);
+		break;
+	case SBI_EXT_PMU_COUNTER_START:
+
+#if __riscv_xlen == 32
+		temp = ((uint64_t)regs->a4 << 32) | regs->a3;
+#else
+		temp = regs->a3;
+#endif
+		ret = sbi_pmu_ctr_start(regs->a0, regs->a1, regs->a2, temp);
+		break;
+	case SBI_EXT_PMU_COUNTER_STOP:
+		ret = sbi_pmu_ctr_stop(regs->a0, regs->a1, regs->a2);
+		break;
+	default:
+		ret = SBI_ENOTSUPP;
+	};
+
+	return ret;
+}
+
+static int sbi_ecall_pmu_probe(unsigned long extid, unsigned long *out_val)
+{
+	/* PMU extension is always enabled */
+	*out_val = 1;
+	return 0;
+}
+
+struct sbi_ecall_extension ecall_pmu = {
+	.extid_start = SBI_EXT_PMU,
+	.extid_end = SBI_EXT_PMU,
+	.handle = sbi_ecall_pmu_handler,
+	.probe = sbi_ecall_pmu_probe,
+};

lib/sbi/sbi_ecall_replace.c

@@ -54,8 +54,8 @@ static int sbi_ecall_rfence_handler(unsigned long extid, unsigned long funcid,
 	struct sbi_tlb_info tlb_info;
 	u32 source_hart = current_hartid();
 
-	if (funcid >= SBI_EXT_RFENCE_REMOTE_HFENCE_GVMA &&
-	    funcid <= SBI_EXT_RFENCE_REMOTE_HFENCE_VVMA_ASID)
+	if (funcid >= SBI_EXT_RFENCE_REMOTE_HFENCE_GVMA_VMID &&
+	    funcid <= SBI_EXT_RFENCE_REMOTE_HFENCE_VVMA)
 		if (!misa_extension('H'))
 			return SBI_ENOTSUPP;
 
@@ -151,7 +151,7 @@ static int sbi_ecall_srst_handler(unsigned long extid, unsigned long funcid,
 		case SBI_SRST_RESET_TYPE_WARM_REBOOT:
 			break;
 		default:
-			return SBI_ENOTSUPP;
+			return SBI_EINVAL;
 		}
 
 		switch (regs->a1) {
@@ -159,7 +159,7 @@ static int sbi_ecall_srst_handler(unsigned long extid, unsigned long funcid,
 		case SBI_SRST_RESET_REASON_SYSFAIL:
 			break;
 		default:
-			return SBI_ENOTSUPP;
+			return SBI_EINVAL;
 		}
 
 		if (sbi_system_reset_supported(regs->a0, regs->a1))

lib/sbi/sbi_emulate_csr.c

@@ -21,14 +21,23 @@
 static bool hpm_allowed(int hpm_num, ulong prev_mode, bool virt)
 {
 	ulong cen = -1UL;
+	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
 
 	if (prev_mode <= PRV_S) {
+		if (sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTEREN)) {
 			cen &= csr_read(CSR_MCOUNTEREN);
 			if (virt)
 				cen &= csr_read(CSR_HCOUNTEREN);
+		} else {
+			cen = 0;
 		}
-	if (prev_mode == PRV_U)
+	}
+	if (prev_mode == PRV_U) {
+		if (sbi_hart_has_feature(scratch, SBI_HART_HAS_SCOUNTEREN))
 			cen &= csr_read(CSR_SCOUNTEREN);
+		else
+			cen = 0;
+	}
 
 	return ((cen >> hpm_num) & 1) ? TRUE : FALSE;
 }

lib/sbi/sbi_fifo.c

@@ -18,7 +18,7 @@ void sbi_fifo_init(struct sbi_fifo *fifo, void *queue_mem, u16 entries,
 	fifo->queue	  = queue_mem;
 	fifo->num_entries = entries;
 	fifo->entry_size  = entry_size;
-	SPIN_LOCK_INIT(&fifo->qlock);
+	SPIN_LOCK_INIT(fifo->qlock);
 	fifo->avail = fifo->tail = 0;
 	sbi_memset(fifo->queue, 0, (size_t)entries * entry_size);
 }
@@ -43,10 +43,13 @@ u16 sbi_fifo_avail(struct sbi_fifo *fifo)
 	return ret;
 }
 
-bool sbi_fifo_is_full(struct sbi_fifo *fifo)
+int sbi_fifo_is_full(struct sbi_fifo *fifo)
 {
 	bool ret;
 
+	if (!fifo)
+		return SBI_EINVAL;
+
 	spin_lock(&fifo->qlock);
 	ret = __sbi_fifo_is_full(fifo);
 	spin_unlock(&fifo->qlock);
@@ -75,10 +78,13 @@ static inline bool __sbi_fifo_is_empty(struct sbi_fifo *fifo)
 	return (fifo->avail == 0) ? TRUE : FALSE;
 }
 
-bool sbi_fifo_is_empty(struct sbi_fifo *fifo)
+int sbi_fifo_is_empty(struct sbi_fifo *fifo)
 {
 	bool ret;
 
+	if (!fifo)
+		return SBI_EINVAL;
+
 	spin_lock(&fifo->qlock);
 	ret = __sbi_fifo_is_empty(fifo);
 	spin_unlock(&fifo->qlock);
@@ -118,7 +124,7 @@ bool sbi_fifo_reset(struct sbi_fifo *fifo)
 int sbi_fifo_inplace_update(struct sbi_fifo *fifo, void *in,
 			    int (*fptr)(void *in, void *data))
 {
-	int i, index = 0;
+	int i, index;
 	int ret = SBI_FIFO_UNCHANGED;
 	void *entry;
 
@@ -135,7 +141,7 @@ int sbi_fifo_inplace_update(struct sbi_fifo *fifo, void *in,
 	for (i = 0; i < fifo->avail; i++) {
 		index = fifo->tail + i;
 		if (index >= fifo->num_entries)
-			index = index - fifo->num_entries;
+			index -= fifo->num_entries;
 		entry = (void *)fifo->queue + (u32)index * fifo->entry_size;
 		ret = fptr(in, entry);
 

lib/sbi/sbi_hart.c

@@ -33,6 +33,7 @@ struct hart_features {
 	unsigned int pmp_addr_bits;
 	unsigned long pmp_gran;
 	unsigned int mhpm_count;
+	unsigned int mhpm_bits;
 };
 static unsigned long hart_features_offset;
 
@@ -50,13 +51,23 @@ static void mstatus_init(struct sbi_scratch *scratch)
 
 	csr_write(CSR_MSTATUS, mstatus_val);
 
-	/* Enable user/supervisor use of perf counters */
+	/* Disable user mode usage of all perf counters except default ones (CY, TM, IR) */
 	if (misa_extension('S') &&
 	    sbi_hart_has_feature(scratch, SBI_HART_HAS_SCOUNTEREN))
-		csr_write(CSR_SCOUNTEREN, -1);
+		csr_write(CSR_SCOUNTEREN, 7);
+
+	/**
+	 * OpenSBI doesn't use any PMU counters in M-mode.
+	 * Supervisor mode usage for all counters are enabled by default
+	 * But counters will not run until mcountinhibit is set.
+	 */
 	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTEREN))
 		csr_write(CSR_MCOUNTEREN, -1);
 
+	/* All programmable counters will start running at runtime after S-mode request */
+	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTINHIBIT))
+		csr_write(CSR_MCOUNTINHIBIT, 0xFFFFFFF8);
+
 	/* Disable all interrupts */
 	csr_write(CSR_MIE, 0);
 
@@ -97,6 +108,9 @@ static int delegate_traps(struct sbi_scratch *scratch)
 
 	/* Send M-mode interrupts and most exceptions to S-mode */
 	interrupts = MIP_SSIP | MIP_STIP | MIP_SEIP;
+	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_SSCOFPMF))
+		interrupts |= MIP_LCOFIP;
+
 	exceptions = (1U << CAUSE_MISALIGNED_FETCH) | (1U << CAUSE_BREAKPOINT) |
 		     (1U << CAUSE_USER_ECALL);
 	if (sbi_platform_has_mfaults_delegation(plat))
@@ -132,17 +146,10 @@ void sbi_hart_delegation_dump(struct sbi_scratch *scratch,
 		/* No delegation possible as mideleg does not exist*/
 		return;
 
-#if __riscv_xlen == 32
-	sbi_printf("%sMIDELEG%s: 0x%08lx\n",
+	sbi_printf("%sMIDELEG%s: 0x%" PRILX "\n",
 		   prefix, suffix, csr_read(CSR_MIDELEG));
-	sbi_printf("%sMEDELEG%s: 0x%08lx\n",
+	sbi_printf("%sMEDELEG%s: 0x%" PRILX "\n",
 		   prefix, suffix, csr_read(CSR_MEDELEG));
-#else
-	sbi_printf("%sMIDELEG%s: 0x%016lx\n",
-		   prefix, suffix, csr_read(CSR_MIDELEG));
-	sbi_printf("%sMEDELEG%s: 0x%016lx\n",
-		   prefix, suffix, csr_read(CSR_MEDELEG));
-#endif
 }
 
 unsigned int sbi_hart_mhpm_count(struct sbi_scratch *scratch)
@@ -177,6 +184,14 @@ unsigned int sbi_hart_pmp_addrbits(struct sbi_scratch *scratch)
 	return hfeatures->pmp_addr_bits;
 }
 
+unsigned int sbi_hart_mhpm_bits(struct sbi_scratch *scratch)
+{
+	struct hart_features *hfeatures =
+			sbi_scratch_offset_ptr(scratch, hart_features_offset);
+
+	return hfeatures->mhpm_bits;
+}
+
 int sbi_hart_pmp_configure(struct sbi_scratch *scratch)
 {
 	struct sbi_domain_memregion *reg;
@@ -259,6 +274,12 @@ static inline char *sbi_hart_feature_id2string(unsigned long feature)
 	case SBI_HART_HAS_MCOUNTEREN:
 		fstr = "mcounteren";
 		break;
+	case SBI_HART_HAS_MCOUNTINHIBIT:
+		fstr = "mcountinhibit";
+		break;
+	case SBI_HART_HAS_SSCOFPMF:
+		fstr = "sscofpmf";
+		break;
 	case SBI_HART_HAS_TIME:
 		fstr = "time";
 		break;
@@ -317,7 +338,11 @@ static unsigned long hart_pmp_get_allowed_addr(void)
 	unsigned long val = 0;
 	struct sbi_trap_info trap = {0};
 
-	csr_write_allowed(CSR_PMPADDR0, (ulong)&trap, PMP_ADDR_MASK);			\
+	csr_write_allowed(CSR_PMPCFG0, (ulong)&trap, 0);
+	if (trap.cause)
+		return 0;
+
+	csr_write_allowed(CSR_PMPADDR0, (ulong)&trap, PMP_ADDR_MASK);
 	if (!trap.cause) {
 		val = csr_read_allowed(CSR_PMPADDR0, (ulong)&trap);
 		if (trap.cause)
@@ -327,6 +352,37 @@ static unsigned long hart_pmp_get_allowed_addr(void)
 	return val;
 }
 
+static int hart_pmu_get_allowed_bits(void)
+{
+	unsigned long val = ~(0UL);
+	struct sbi_trap_info trap = {0};
+	int num_bits = 0;
+
+	/**
+	 * It is assumed that platforms will implement same number of bits for
+	 * all the performance counters including mcycle/minstret.
+	 */
+	csr_write_allowed(CSR_MHPMCOUNTER3, (ulong)&trap, val);
+	if (!trap.cause) {
+		val = csr_read_allowed(CSR_MHPMCOUNTER3, (ulong)&trap);
+		if (trap.cause)
+			return 0;
+	}
+	num_bits = __fls(val) + 1;
+#if __riscv_xlen == 32
+	csr_write_allowed(CSR_MHPMCOUNTER3H, (ulong)&trap, val);
+	if (!trap.cause) {
+		val = csr_read_allowed(CSR_MHPMCOUNTER3H, (ulong)&trap);
+		if (trap.cause)
+			return num_bits;
+	}
+	num_bits += __fls(val) + 1;
+
+#endif
+
+	return num_bits;
+}
+
 static void hart_detect_features(struct sbi_scratch *scratch)
 {
 	struct sbi_trap_info trap = {0};
@@ -392,9 +448,17 @@ __pmp_skip:
 
 	/* Detect number of MHPM counters */
 	__check_csr(CSR_MHPMCOUNTER3, 0, 1UL, mhpm_count, __mhpm_skip);
+	hfeatures->mhpm_bits = hart_pmu_get_allowed_bits();
+
 	__check_csr_4(CSR_MHPMCOUNTER4, 0, 1UL, mhpm_count, __mhpm_skip);
 	__check_csr_8(CSR_MHPMCOUNTER8, 0, 1UL, mhpm_count, __mhpm_skip);
 	__check_csr_16(CSR_MHPMCOUNTER16, 0, 1UL, mhpm_count, __mhpm_skip);
+
+	/**
+	 * No need to check for MHPMCOUNTERH for RV32 as they are expected to be
+	 * implemented if MHPMCOUNTER is implemented.
+	 */
+
 __mhpm_skip:
 
 #undef __check_csr_64
@@ -406,7 +470,6 @@ __mhpm_skip:
 #undef __check_csr
 
 	/* Detect if hart supports SCOUNTEREN feature */
-	trap.cause = 0;
 	val = csr_read_allowed(CSR_SCOUNTEREN, (unsigned long)&trap);
 	if (!trap.cause) {
 		csr_write_allowed(CSR_SCOUNTEREN, (unsigned long)&trap, val);
@@ -415,7 +478,6 @@ __mhpm_skip:
 	}
 
 	/* Detect if hart supports MCOUNTEREN feature */
-	trap.cause = 0;
 	val = csr_read_allowed(CSR_MCOUNTEREN, (unsigned long)&trap);
 	if (!trap.cause) {
 		csr_write_allowed(CSR_MCOUNTEREN, (unsigned long)&trap, val);
@@ -423,30 +485,33 @@ __mhpm_skip:
 			hfeatures->features |= SBI_HART_HAS_MCOUNTEREN;
 	}
 
+	/* Detect if hart supports MCOUNTINHIBIT feature */
+	val = csr_read_allowed(CSR_MCOUNTINHIBIT, (unsigned long)&trap);
+	if (!trap.cause) {
+		csr_write_allowed(CSR_MCOUNTINHIBIT, (unsigned long)&trap, val);
+		if (!trap.cause)
+			hfeatures->features |= SBI_HART_HAS_MCOUNTINHIBIT;
+	}
+
+	/* Counter overflow/filtering is not useful without mcounter/inhibit */
+	if (hfeatures->features & SBI_HART_HAS_MCOUNTINHIBIT &&
+	    hfeatures->features & SBI_HART_HAS_MCOUNTEREN) {
+		/* Detect if hart supports sscofpmf */
+		csr_read_allowed(CSR_SCOUNTOVF, (unsigned long)&trap);
+		if (!trap.cause)
+			hfeatures->features |= SBI_HART_HAS_SSCOFPMF;
+	}
+
 	/* Detect if hart supports time CSR */
-	trap.cause = 0;
 	csr_read_allowed(CSR_TIME, (unsigned long)&trap);
 	if (!trap.cause)
 		hfeatures->features |= SBI_HART_HAS_TIME;
 }
 
-int sbi_hart_init(struct sbi_scratch *scratch, bool cold_boot)
+int sbi_hart_reinit(struct sbi_scratch *scratch)
 {
 	int rc;
 
-	if (cold_boot) {
-		if (misa_extension('H'))
-			sbi_hart_expected_trap = &__sbi_expected_trap_hext;
-
-		hart_features_offset = sbi_scratch_alloc_offset(
-						sizeof(struct hart_features),
-						"HART_FEATURES");
-		if (!hart_features_offset)
-			return SBI_ENOMEM;
-	}
-
-	hart_detect_features(scratch);
-
 	mstatus_init(scratch);
 
 	rc = fp_init(scratch);
@@ -460,6 +525,23 @@ int sbi_hart_init(struct sbi_scratch *scratch, bool cold_boot)
 	return 0;
 }
 
+int sbi_hart_init(struct sbi_scratch *scratch, bool cold_boot)
+{
+	if (cold_boot) {
+		if (misa_extension('H'))
+			sbi_hart_expected_trap = &__sbi_expected_trap_hext;
+
+		hart_features_offset = sbi_scratch_alloc_offset(
+						sizeof(struct hart_features));
+		if (!hart_features_offset)
+			return SBI_ENOMEM;
+	}
+
+	hart_detect_features(scratch);
+
+	return sbi_hart_reinit(scratch);
+}
+
 void __attribute__((noreturn)) sbi_hart_hang(void)
 {
 	while (1)

lib/sbi/sbi_hfence.S

@@ -27,7 +27,7 @@
 	.global __sbi_hfence_gvma_vmid_gpa
 __sbi_hfence_gvma_vmid_gpa:
 	/*
-	 * rs1 = a0 (GPA)
+	 * rs1 = a0 (GPA >> 2)
 	 * rs2 = a1 (VMID)
 	 * HFENCE.GVMA a0, a1
 	 * 0110001 01011 01010 000 00000 1110011
@@ -51,7 +51,7 @@ __sbi_hfence_gvma_vmid:
 	.global __sbi_hfence_gvma_gpa
 __sbi_hfence_gvma_gpa:
 	/*
-	 * rs1 = a0 (GPA)
+	 * rs1 = a0 (GPA >> 2)
 	 * rs2 = zero
 	 * HFENCE.GVMA a0
 	 * 0110001 00000 01010 000 00000 1110011

lib/sbi/sbi_hsm.c

@@ -21,42 +21,22 @@
 #include <sbi/sbi_hsm.h>
 #include <sbi/sbi_init.h>
 #include <sbi/sbi_ipi.h>
-#include <sbi/sbi_platform.h>
+#include <sbi/sbi_scratch.h>
 #include <sbi/sbi_system.h>
 #include <sbi/sbi_timer.h>
 #include <sbi/sbi_console.h>
 
+static const struct sbi_hsm_device *hsm_dev = NULL;
 static unsigned long hart_data_offset;
 
 /** Per hart specific data to manage state transition **/
 struct sbi_hsm_data {
 	atomic_t state;
+	unsigned long suspend_type;
+	unsigned long saved_mie;
+	unsigned long saved_mip;
 };
 
-int sbi_hsm_hart_state_to_status(int state)
-{
-	int ret;
-
-	switch (state) {
-	case SBI_HART_STOPPED:
-		ret = SBI_HSM_HART_STATUS_STOPPED;
-		break;
-	case SBI_HART_STOPPING:
-		ret = SBI_HSM_HART_STATUS_STOP_PENDING;
-		break;
-	case SBI_HART_STARTING:
-		ret = SBI_HSM_HART_STATUS_START_PENDING;
-		break;
-	case SBI_HART_STARTED:
-		ret = SBI_HSM_HART_STATUS_STARTED;
-		break;
-	default:
-		ret = SBI_EINVAL;
-	}
-
-	return ret;
-}
-
 static inline int __sbi_hsm_hart_get_state(u32 hartid)
 {
 	struct sbi_hsm_data *hdata;
@@ -64,7 +44,7 @@ static inline int __sbi_hsm_hart_get_state(u32 hartid)
 
 	scratch = sbi_hartid_to_scratch(hartid);
 	if (!scratch)
-		return SBI_HART_UNKNOWN;
+		return SBI_EINVAL;
 
 	hdata = sbi_scratch_offset_ptr(scratch, hart_data_offset);
 	return atomic_read(&hdata->state);
@@ -73,19 +53,11 @@ static inline int __sbi_hsm_hart_get_state(u32 hartid)
 int sbi_hsm_hart_get_state(const struct sbi_domain *dom, u32 hartid)
 {
 	if (!sbi_domain_is_assigned_hart(dom, hartid))
-		return SBI_HART_UNKNOWN;
+		return SBI_EINVAL;
 
 	return __sbi_hsm_hart_get_state(hartid);
 }
 
-static bool sbi_hsm_hart_started(const struct sbi_domain *dom, u32 hartid)
-{
-	if (sbi_hsm_hart_get_state(dom, hartid) == SBI_HART_STARTED)
-		return TRUE;
-	else
-		return FALSE;
-}
-
 /**
  * Get ulong HART mask for given HART base ID
  * @param dom the domain to be used for output HART mask
@@ -94,9 +66,10 @@ static bool sbi_hsm_hart_started(const struct sbi_domain *dom, u32 hartid)
  * @return 0 on success and SBI_Exxx (< 0) on failure
  * Note: the output HART mask will be set to zero on failure as well.
  */
-int sbi_hsm_hart_started_mask(const struct sbi_domain *dom,
+int sbi_hsm_hart_interruptible_mask(const struct sbi_domain *dom,
 				    ulong hbase, ulong *out_hmask)
 {
+	int hstate;
 	ulong i, hmask, dmask;
 	ulong hend = sbi_scratch_last_hartid() + 1;
 
@@ -109,10 +82,13 @@ int sbi_hsm_hart_started_mask(const struct sbi_domain *dom,
 	dmask = sbi_domain_get_assigned_hartmask(dom, hbase);
 	for (i = hbase; i < hend; i++) {
 		hmask = 1UL << (i - hbase);
-		if ((dmask & hmask) &&
-		    (__sbi_hsm_hart_get_state(i) == SBI_HART_STARTED))
+		if (dmask & hmask) {
+			hstate = __sbi_hsm_hart_get_state(i);
+			if (hstate == SBI_HSM_STATE_STARTED ||
+			    hstate == SBI_HSM_STATE_SUSPENDED)
 				*out_hmask |= hmask;
 		}
+	}
 
 	return 0;
 }
@@ -123,16 +99,15 @@ void sbi_hsm_prepare_next_jump(struct sbi_scratch *scratch, u32 hartid)
 	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
 							    hart_data_offset);
 
-	oldstate = atomic_cmpxchg(&hdata->state, SBI_HART_STARTING,
-				  SBI_HART_STARTED);
-	if (oldstate != SBI_HART_STARTING)
+	oldstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_START_PENDING,
+				  SBI_HSM_STATE_STARTED);
+	if (oldstate != SBI_HSM_STATE_START_PENDING)
 		sbi_hart_hang();
 }
 
 static void sbi_hsm_hart_wait(struct sbi_scratch *scratch, u32 hartid)
 {
 	unsigned long saved_mie;
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
 							    hart_data_offset);
 	/* Save MIE CSR */
@@ -142,15 +117,65 @@ static void sbi_hsm_hart_wait(struct sbi_scratch *scratch, u32 hartid)
 	csr_set(CSR_MIE, MIP_MSIP);
 
 	/* Wait for hart_add call*/
-	while (atomic_read(&hdata->state) != SBI_HART_STARTING) {
+	while (atomic_read(&hdata->state) != SBI_HSM_STATE_START_PENDING) {
 		wfi();
 	};
 
 	/* Restore MIE CSR */
 	csr_write(CSR_MIE, saved_mie);
 
-	/* Clear current HART IPI */
-	sbi_platform_ipi_clear(plat, hartid);
+	/*
+	 * No need to clear IPI here because the sbi_ipi_init() will
+	 * clear it for current HART via sbi_platform_ipi_init().
+	 */
+}
+
+const struct sbi_hsm_device *sbi_hsm_get_device(void)
+{
+	return hsm_dev;
+}
+
+void sbi_hsm_set_device(const struct sbi_hsm_device *dev)
+{
+	if (!dev || hsm_dev)
+		return;
+
+	hsm_dev = dev;
+}
+
+static bool hsm_device_has_hart_hotplug(void)
+{
+	if (hsm_dev && hsm_dev->hart_start && hsm_dev->hart_stop)
+		return true;
+	return false;
+}
+
+static bool hsm_device_has_hart_secondary_boot(void)
+{
+	if (hsm_dev && hsm_dev->hart_start && !hsm_dev->hart_stop)
+		return true;
+	return false;
+}
+
+static int hsm_device_hart_start(u32 hartid, ulong saddr)
+{
+	if (hsm_dev && hsm_dev->hart_start)
+		return hsm_dev->hart_start(hartid, saddr);
+	return SBI_ENOTSUPP;
+}
+
+static int hsm_device_hart_stop(void)
+{
+	if (hsm_dev && hsm_dev->hart_stop)
+		return hsm_dev->hart_stop();
+	return SBI_ENOTSUPP;
+}
+
+static int hsm_device_hart_suspend(u32 suspend_type, ulong raddr)
+{
+	if (hsm_dev && hsm_dev->hart_suspend)
+		return hsm_dev->hart_suspend(suspend_type, raddr);
+	return SBI_ENOTSUPP;
 }
 
 int sbi_hsm_init(struct sbi_scratch *scratch, u32 hartid, bool cold_boot)
@@ -160,8 +185,7 @@ int sbi_hsm_init(struct sbi_scratch *scratch, u32 hartid, bool cold_boot)
 	struct sbi_hsm_data *hdata;
 
 	if (cold_boot) {
-		hart_data_offset = sbi_scratch_alloc_offset(sizeof(*hdata),
-							    "HART_DATA");
+		hart_data_offset = sbi_scratch_alloc_offset(sizeof(*hdata));
 		if (!hart_data_offset)
 			return SBI_ENOMEM;
 
@@ -174,7 +198,9 @@ int sbi_hsm_init(struct sbi_scratch *scratch, u32 hartid, bool cold_boot)
 			hdata = sbi_scratch_offset_ptr(rscratch,
 						       hart_data_offset);
 			ATOMIC_INIT(&hdata->state,
-			(i == hartid) ? SBI_HART_STARTING : SBI_HART_STOPPED);
+				    (i == hartid) ?
+				    SBI_HSM_STATE_START_PENDING :
+				    SBI_HSM_STATE_STOPPED);
 		}
 	} else {
 		sbi_hsm_hart_wait(scratch, hartid);
@@ -186,18 +212,17 @@ int sbi_hsm_init(struct sbi_scratch *scratch, u32 hartid, bool cold_boot)
 void __noreturn sbi_hsm_exit(struct sbi_scratch *scratch)
 {
 	u32 hstate;
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
 							    hart_data_offset);
 	void (*jump_warmboot)(void) = (void (*)(void))scratch->warmboot_addr;
 
-	hstate = atomic_cmpxchg(&hdata->state, SBI_HART_STOPPING,
-				SBI_HART_STOPPED);
-	if (hstate != SBI_HART_STOPPING)
+	hstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_STOP_PENDING,
+				SBI_HSM_STATE_STOPPED);
+	if (hstate != SBI_HSM_STATE_STOP_PENDING)
 		goto fail_exit;
 
-	if (sbi_platform_has_hart_hotplug(plat)) {
-		sbi_platform_hart_stop(plat);
+	if (hsm_device_has_hart_hotplug()) {
+		hsm_device_hart_stop();
 		/* It should never reach here */
 		goto fail_exit;
 	}
@@ -223,7 +248,6 @@ int sbi_hsm_hart_start(struct sbi_scratch *scratch,
 	unsigned int hstate;
 	struct sbi_scratch *rscratch;
 	struct sbi_hsm_data *hdata;
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 
 	/* For now, we only allow start mode to be S-mode or U-mode. */
 	if (smode != PRV_S && smode != PRV_U)
@@ -232,22 +256,22 @@ int sbi_hsm_hart_start(struct sbi_scratch *scratch,
 		return SBI_EINVAL;
 	if (dom && !sbi_domain_check_addr(dom, saddr, smode,
 					  SBI_DOMAIN_EXECUTE))
-		return SBI_EINVAL;
+		return SBI_EINVALID_ADDR;
 
 	rscratch = sbi_hartid_to_scratch(hartid);
 	if (!rscratch)
 		return SBI_EINVAL;
 	hdata = sbi_scratch_offset_ptr(rscratch, hart_data_offset);
-	hstate = atomic_cmpxchg(&hdata->state, SBI_HART_STOPPED,
-				SBI_HART_STARTING);
-	if (hstate == SBI_HART_STARTED)
+	hstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_STOPPED,
+				SBI_HSM_STATE_START_PENDING);
+	if (hstate == SBI_HSM_STATE_STARTED)
 		return SBI_EALREADY;
 
 	/**
 	 * if a hart is already transition to start or stop, another start call
 	 * is considered as invalid request.
 	 */
-	if (hstate != SBI_HART_STOPPED)
+	if (hstate != SBI_HSM_STATE_STOPPED)
 		return SBI_EINVAL;
 
 	init_count = sbi_init_count(hartid);
@@ -255,12 +279,11 @@ int sbi_hsm_hart_start(struct sbi_scratch *scratch,
 	rscratch->next_addr = saddr;
 	rscratch->next_mode = smode;
 
-	if (sbi_platform_has_hart_hotplug(plat) ||
-	   (sbi_platform_has_hart_secondary_boot(plat) && !init_count)) {
-		return sbi_platform_hart_start(plat, hartid,
-					       scratch->warmboot_addr);
+	if (hsm_device_has_hart_hotplug() ||
+	   (hsm_device_has_hart_secondary_boot() && !init_count)) {
+		return hsm_device_hart_start(hartid, scratch->warmboot_addr);
 	} else {
-		sbi_platform_ipi_send(plat, hartid);
+		sbi_ipi_raw_send(hartid);
 	}
 
 	return 0;
@@ -269,19 +292,19 @@ int sbi_hsm_hart_start(struct sbi_scratch *scratch,
 int sbi_hsm_hart_stop(struct sbi_scratch *scratch, bool exitnow)
 {
 	int oldstate;
-	u32 hartid = current_hartid();
+	const struct sbi_domain *dom = sbi_domain_thishart_ptr();
 	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
 							    hart_data_offset);
 
-	if (!sbi_hsm_hart_started(sbi_domain_thishart_ptr(), hartid))
-		return SBI_EINVAL;
+	if (!dom)
+		return SBI_EFAIL;
 
-	oldstate = atomic_cmpxchg(&hdata->state, SBI_HART_STARTED,
-				  SBI_HART_STOPPING);
-	if (oldstate != SBI_HART_STARTED) {
+	oldstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_STARTED,
+				  SBI_HSM_STATE_STOP_PENDING);
+	if (oldstate != SBI_HSM_STATE_STARTED) {
 		sbi_printf("%s: ERR: The hart is in invalid state [%u]\n",
 			   __func__, oldstate);
-		return SBI_EDENIED;
+		return SBI_EFAIL;
 	}
 
 	if (exitnow)
@@ -289,3 +312,182 @@ int sbi_hsm_hart_stop(struct sbi_scratch *scratch, bool exitnow)
 
 	return 0;
 }
+
+static int __sbi_hsm_suspend_ret_default(struct sbi_scratch *scratch)
+{
+	/* Wait for interrupt */
+	wfi();
+
+	return 0;
+}
+
+static void __sbi_hsm_suspend_non_ret_save(struct sbi_scratch *scratch)
+{
+	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
+							    hart_data_offset);
+
+	/*
+	 * We will be resuming in warm-boot path so the MIE and MIP CSRs
+	 * will be back to initial state. It is possible that HART has
+	 * configured timer event before going to suspend state so we
+	 * should save MIE and MIP CSRs and restore it after resuming.
+	 *
+	 * Further, the M-mode bits in MIP CSR are read-only and set by
+	 * external devices (such as interrupt controller) whereas all
+	 * VS-mode bits in MIP are read-only alias of bits in HVIP CSR.
+	 *
+	 * This means we should only save/restore S-mode bits of MIP CSR
+	 * such as MIP.SSIP and MIP.STIP.
+	 */
+
+	hdata->saved_mie = csr_read(CSR_MIE);
+	hdata->saved_mip = csr_read(CSR_MIP) & (MIP_SSIP | MIP_STIP);
+}
+
+static void __sbi_hsm_suspend_non_ret_restore(struct sbi_scratch *scratch)
+{
+	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
+							    hart_data_offset);
+
+	csr_write(CSR_MIE, hdata->saved_mie);
+	csr_write(CSR_MIP, (hdata->saved_mip & (MIP_SSIP | MIP_STIP)));
+}
+
+static int __sbi_hsm_suspend_non_ret_default(struct sbi_scratch *scratch,
+					     ulong raddr)
+{
+	void (*jump_warmboot)(void) = (void (*)(void))scratch->warmboot_addr;
+
+	/* Wait for interrupt */
+	wfi();
+
+	/*
+	 * Directly jump to warm reboot to simulate resume from a
+	 * non-retentive suspend.
+	 */
+	jump_warmboot();
+
+	return 0;
+}
+
+void sbi_hsm_hart_resume_start(struct sbi_scratch *scratch)
+{
+	int oldstate;
+	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
+							    hart_data_offset);
+
+	/* If current HART was SUSPENDED then set RESUME_PENDING state */
+	oldstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_SUSPENDED,
+			SBI_HSM_STATE_RESUME_PENDING);
+	if (oldstate != SBI_HSM_STATE_SUSPENDED) {
+		sbi_printf("%s: ERR: The hart is in invalid state [%u]\n",
+			   __func__, oldstate);
+		sbi_hart_hang();
+	}
+}
+
+void sbi_hsm_hart_resume_finish(struct sbi_scratch *scratch)
+{
+	u32 oldstate;
+	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
+							    hart_data_offset);
+
+	/* If current HART was RESUME_PENDING then set STARTED state */
+	oldstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_RESUME_PENDING,
+				  SBI_HSM_STATE_STARTED);
+	if (oldstate != SBI_HSM_STATE_RESUME_PENDING) {
+		sbi_printf("%s: ERR: The hart is in invalid state [%u]\n",
+			   __func__, oldstate);
+		sbi_hart_hang();
+	}
+
+	/*
+	 * Restore some of the M-mode CSRs which we are re-configured by
+	 * the warm-boot sequence.
+	 */
+	__sbi_hsm_suspend_non_ret_restore(scratch);
+}
+
+int sbi_hsm_hart_suspend(struct sbi_scratch *scratch, u32 suspend_type,
+			 ulong raddr, ulong rmode, ulong priv)
+{
+	int oldstate, ret;
+	const struct sbi_domain *dom = sbi_domain_thishart_ptr();
+	struct sbi_hsm_data *hdata = sbi_scratch_offset_ptr(scratch,
+							    hart_data_offset);
+
+	/* For now, we only allow suspend from S-mode or U-mode. */
+
+	/* Sanity check on domain assigned to current HART */
+	if (!dom)
+		return SBI_EINVAL;
+
+	/* Sanity check on suspend type */
+	if (SBI_HSM_SUSPEND_RET_DEFAULT < suspend_type &&
+	    suspend_type < SBI_HSM_SUSPEND_RET_PLATFORM)
+		return SBI_EINVAL;
+	if (SBI_HSM_SUSPEND_NON_RET_DEFAULT < suspend_type &&
+	    suspend_type < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
+		return SBI_EINVAL;
+
+	/* Additional sanity check for non-retentive suspend */
+	if (suspend_type & SBI_HSM_SUSP_NON_RET_BIT) {
+		if (rmode != PRV_S && rmode != PRV_U)
+			return SBI_EINVAL;
+		if (dom && !sbi_domain_check_addr(dom, raddr, rmode,
+						  SBI_DOMAIN_EXECUTE))
+			return SBI_EINVALID_ADDR;
+	}
+
+	/* Save the resume address and resume mode */
+	scratch->next_arg1 = priv;
+	scratch->next_addr = raddr;
+	scratch->next_mode = rmode;
+
+	/* Directly move from STARTED to SUSPENDED state */
+	oldstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_STARTED,
+				  SBI_HSM_STATE_SUSPENDED);
+	if (oldstate != SBI_HSM_STATE_STARTED) {
+		sbi_printf("%s: ERR: The hart is in invalid state [%u]\n",
+			   __func__, oldstate);
+		ret = SBI_EDENIED;
+		goto fail_restore_state;
+	}
+
+	/* Save the suspend type */
+	hdata->suspend_type = suspend_type;
+
+	/*
+	 * Save context which will be restored after resuming from
+	 * non-retentive suspend.
+	 */
+	if (suspend_type & SBI_HSM_SUSP_NON_RET_BIT)
+		__sbi_hsm_suspend_non_ret_save(scratch);
+
+	/* Try platform specific suspend */
+	ret = hsm_device_hart_suspend(suspend_type, scratch->warmboot_addr);
+	if (ret == SBI_ENOTSUPP) {
+		/* Try generic implementation of default suspend types */
+		if (suspend_type == SBI_HSM_SUSPEND_RET_DEFAULT) {
+			ret = __sbi_hsm_suspend_ret_default(scratch);
+		} else if (suspend_type == SBI_HSM_SUSPEND_NON_RET_DEFAULT) {
+			ret = __sbi_hsm_suspend_non_ret_default(scratch,
+						scratch->warmboot_addr);
+		}
+	}
+
+fail_restore_state:
+	/*
+	 * We might have successfully resumed from retentive suspend
+	 * or suspend failed. In both cases, we restore state of hart.
+	 */
+	oldstate = atomic_cmpxchg(&hdata->state, SBI_HSM_STATE_SUSPENDED,
+				  SBI_HSM_STATE_STARTED);
+	if (oldstate != SBI_HSM_STATE_SUSPENDED) {
+		sbi_printf("%s: ERR: The hart is in invalid state [%u]\n",
+			   __func__, oldstate);
+		sbi_hart_hang();
+	}
+
+	return ret;
+}

lib/sbi/sbi_illegal_insn.c

@@ -13,6 +13,7 @@
 #include <sbi/sbi_emulate_csr.h>
 #include <sbi/sbi_error.h>
 #include <sbi/sbi_illegal_insn.h>
+#include <sbi/sbi_pmu.h>
 #include <sbi/sbi_trap.h>
 #include <sbi/sbi_unpriv.h>
 
@@ -129,6 +130,7 @@ int sbi_illegal_insn_handler(ulong insn, struct sbi_trap_regs *regs)
 	 * instruction trap.
 	 */
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_ILLEGAL_INSN);
 	if (unlikely((insn & 3) != 3)) {
 		insn = sbi_get_insn(regs->mepc, &uptrap);
 		if (uptrap.cause) {

lib/sbi/sbi_init.c

@@ -19,6 +19,7 @@
 #include <sbi/sbi_hsm.h>
 #include <sbi/sbi_ipi.h>
 #include <sbi/sbi_platform.h>
+#include <sbi/sbi_pmu.h>
 #include <sbi/sbi_system.h>
 #include <sbi/sbi_string.h>
 #include <sbi/sbi_timer.h>
@@ -47,12 +48,25 @@ static void sbi_boot_print_banner(struct sbi_scratch *scratch)
 		   OPENSBI_VERSION_MINOR);
 #endif
 
+#ifdef OPENSBI_BUILD_TIME_STAMP
+	sbi_printf("Build time: %s\n", OPENSBI_BUILD_TIME_STAMP);
+#endif
+
+#ifdef OPENSBI_BUILD_COMPILER_VERSION
+	sbi_printf("Build compiler: %s\n", OPENSBI_BUILD_COMPILER_VERSION);
+#endif
+
 	sbi_printf(BANNER);
 }
 
 static void sbi_boot_print_general(struct sbi_scratch *scratch)
 {
 	char str[128];
+	const struct sbi_hsm_device *hdev;
+	const struct sbi_ipi_device *idev;
+	const struct sbi_timer_device *tdev;
+	const struct sbi_console_device *cdev;
+	const struct sbi_system_reset_device *srdev;
 	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 
 	if (scratch->options & SBI_SCRATCH_NO_BOOT_PRINTS)
@@ -65,6 +79,25 @@ static void sbi_boot_print_general(struct sbi_scratch *scratch)
 	sbi_printf("Platform Features         : %s\n", str);
 	sbi_printf("Platform HART Count       : %u\n",
 		   sbi_platform_hart_count(plat));
+	idev = sbi_ipi_get_device();
+	sbi_printf("Platform IPI Device       : %s\n",
+		   (idev) ? idev->name : "---");
+	tdev = sbi_timer_get_device();
+	sbi_printf("Platform Timer Device     : %s @ %luHz\n",
+		   (tdev) ? tdev->name : "---",
+		   (tdev) ? tdev->timer_freq : 0);
+	cdev = sbi_console_get_device();
+	sbi_printf("Platform Console Device   : %s\n",
+		   (cdev) ? cdev->name : "---");
+	hdev = sbi_hsm_get_device();
+	sbi_printf("Platform HSM Device       : %s\n",
+		   (hdev) ? hdev->name : "---");
+	srdev = sbi_system_reset_get_device(SBI_SRST_RESET_TYPE_COLD_REBOOT, 0);
+	sbi_printf("Platform Reboot Device    : %s\n",
+		   (srdev) ? srdev->name : "---");
+	srdev = sbi_system_reset_get_device(SBI_SRST_RESET_TYPE_SHUTDOWN, 0);
+	sbi_printf("Platform Shutdown Device  : %s\n",
+		   (srdev) ? srdev->name : "---");
 
 	/* Firmware details */
 	sbi_printf("Firmware Base             : 0x%lx\n", scratch->fw_start);
@@ -117,8 +150,6 @@ static void sbi_boot_print_hart(struct sbi_scratch *scratch, u32 hartid)
 		   sbi_hart_pmp_addrbits(scratch));
 	sbi_printf("Boot HART MHPM Count      : %d\n",
 		   sbi_hart_mhpm_count(scratch));
-	sbi_printf("Boot HART MHPM Count      : %d\n",
-		   sbi_hart_mhpm_count(scratch));
 	sbi_hart_delegation_dump(scratch, "Boot HART ", "         ");
 }
 
@@ -130,7 +161,6 @@ static unsigned long coldboot_done;
 static void wait_for_coldboot(struct sbi_scratch *scratch, u32 hartid)
 {
 	unsigned long saved_mie, cmip;
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 
 	/* Save MIE CSR */
 	saved_mie = csr_read(CSR_MIE);
@@ -167,14 +197,18 @@ static void wait_for_coldboot(struct sbi_scratch *scratch, u32 hartid)
 	/* Restore MIE CSR */
 	csr_write(CSR_MIE, saved_mie);
 
-	/* Clear current HART IPI */
-	sbi_platform_ipi_clear(plat, hartid);
+	/*
+	 * The wait for coldboot is common for both warm startup and
+	 * warm resume path so clearing IPI here would result in losing
+	 * an IPI in warm resume path.
+	 *
+	 * Also, the sbi_platform_ipi_init() called from sbi_ipi_init()
+	 * will automatically clear IPI for current HART.
+	 */
 }
 
 static void wake_coldboot_harts(struct sbi_scratch *scratch, u32 hartid)
 {
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
-
 	/* Mark coldboot done */
 	__smp_store_release(&coldboot_done, 1);
 
@@ -182,10 +216,10 @@ static void wake_coldboot_harts(struct sbi_scratch *scratch, u32 hartid)
 	spin_lock(&coldboot_lock);
 
 	/* Send an IPI to all HARTs waiting for coldboot */
-	for (int i = 0; i <= sbi_scratch_last_hartid(); i++) {
+	for (u32 i = 0; i <= sbi_scratch_last_hartid(); i++) {
 		if ((i != hartid) &&
 		    sbi_hartmask_test_hart(i, &coldboot_wait_hmask))
-			sbi_platform_ipi_send(plat, i);
+			sbi_ipi_raw_send(i);
 	}
 
 	/* Release coldboot lock */
@@ -210,8 +244,7 @@ static void __noreturn init_coldboot(struct sbi_scratch *scratch, u32 hartid)
 	if (rc)
 		sbi_hart_hang();
 
-	init_count_offset = sbi_scratch_alloc_offset(__SIZEOF_POINTER__,
-						     "INIT_COUNT");
+	init_count_offset = sbi_scratch_alloc_offset(__SIZEOF_POINTER__);
 	if (!init_count_offset)
 		sbi_hart_hang();
 
@@ -231,6 +264,10 @@ static void __noreturn init_coldboot(struct sbi_scratch *scratch, u32 hartid)
 	if (rc)
 		sbi_hart_hang();
 
+	rc = sbi_pmu_init(scratch, TRUE);
+	if (rc)
+		sbi_hart_hang();
+
 	sbi_boot_print_banner(scratch);
 
 	rc = sbi_platform_irqchip_init(plat, TRUE);
@@ -264,8 +301,6 @@ static void __noreturn init_coldboot(struct sbi_scratch *scratch, u32 hartid)
 		sbi_hart_hang();
 	}
 
-	sbi_boot_print_general(scratch);
-
 	/*
 	 * Note: Finalize domains after HSM initialization so that we
 	 * can startup non-root domains.
@@ -279,8 +314,6 @@ static void __noreturn init_coldboot(struct sbi_scratch *scratch, u32 hartid)
 		sbi_hart_hang();
 	}
 
-	sbi_boot_print_domains(scratch);
-
 	rc = sbi_hart_pmp_configure(scratch);
 	if (rc) {
 		sbi_printf("%s: PMP configure failed (error %d)\n",
@@ -299,6 +332,10 @@ static void __noreturn init_coldboot(struct sbi_scratch *scratch, u32 hartid)
 		sbi_hart_hang();
 	}
 
+	sbi_boot_print_general(scratch);
+
+	sbi_boot_print_domains(scratch);
+
 	sbi_boot_print_hart(scratch, hartid);
 
 	wake_coldboot_harts(scratch, hartid);
@@ -311,14 +348,12 @@ static void __noreturn init_coldboot(struct sbi_scratch *scratch, u32 hartid)
 			     scratch->next_mode, FALSE);
 }
 
-static void __noreturn init_warmboot(struct sbi_scratch *scratch, u32 hartid)
+static void init_warm_startup(struct sbi_scratch *scratch, u32 hartid)
 {
 	int rc;
 	unsigned long *init_count;
 	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 
-	wait_for_coldboot(scratch, hartid);
-
 	if (!init_count_offset)
 		sbi_hart_hang();
 
@@ -334,6 +369,10 @@ static void __noreturn init_warmboot(struct sbi_scratch *scratch, u32 hartid)
 	if (rc)
 		sbi_hart_hang();
 
+	rc = sbi_pmu_init(scratch, FALSE);
+	if (rc)
+		sbi_hart_hang();
+
 	rc = sbi_platform_irqchip_init(plat, FALSE);
 	if (rc)
 		sbi_hart_hang();
@@ -362,6 +401,40 @@ static void __noreturn init_warmboot(struct sbi_scratch *scratch, u32 hartid)
 	(*init_count)++;
 
 	sbi_hsm_prepare_next_jump(scratch, hartid);
+}
+
+static void init_warm_resume(struct sbi_scratch *scratch)
+{
+	int rc;
+
+	sbi_hsm_hart_resume_start(scratch);
+
+	rc = sbi_hart_reinit(scratch);
+	if (rc)
+		sbi_hart_hang();
+
+	rc = sbi_hart_pmp_configure(scratch);
+	if (rc)
+		sbi_hart_hang();
+
+	sbi_hsm_hart_resume_finish(scratch);
+}
+
+static void __noreturn init_warmboot(struct sbi_scratch *scratch, u32 hartid)
+{
+	int hstate;
+
+	wait_for_coldboot(scratch, hartid);
+
+	hstate = sbi_hsm_hart_get_state(sbi_domain_thishart_ptr(), hartid);
+	if (hstate < 0)
+		sbi_hart_hang();
+
+	if (hstate == SBI_HSM_STATE_SUSPENDED)
+		init_warm_resume(scratch);
+	else
+		init_warm_startup(scratch, hartid);
+
 	sbi_hart_switch_mode(hartid, scratch->next_arg1,
 			     scratch->next_addr,
 			     scratch->next_mode, FALSE);
@@ -463,6 +536,8 @@ void __noreturn sbi_exit(struct sbi_scratch *scratch)
 
 	sbi_platform_early_exit(plat);
 
+	sbi_pmu_exit(scratch);
+
 	sbi_timer_exit(scratch);
 
 	sbi_ipi_exit(scratch);

lib/sbi/sbi_ipi.c

@@ -19,13 +19,16 @@
 #include <sbi/sbi_init.h>
 #include <sbi/sbi_ipi.h>
 #include <sbi/sbi_platform.h>
+#include <sbi/sbi_pmu.h>
+#include <sbi/sbi_string.h>
+#include <sbi/sbi_tlb.h>
 
 struct sbi_ipi_data {
 	unsigned long ipi_type;
 };
 
 static unsigned long ipi_data_off;
-
+static const struct sbi_ipi_device *ipi_dev = NULL;
 static const struct sbi_ipi_event_ops *ipi_ops_array[SBI_IPI_EVENT_MAX];
 
 static int sbi_ipi_send(struct sbi_scratch *scratch, u32 remote_hartid,
@@ -33,7 +36,6 @@ static int sbi_ipi_send(struct sbi_scratch *scratch, u32 remote_hartid,
 {
 	int ret;
 	struct sbi_scratch *remote_scratch = NULL;
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 	struct sbi_ipi_data *ipi_data;
 	const struct sbi_ipi_event_ops *ipi_ops;
 
@@ -61,7 +63,11 @@ static int sbi_ipi_send(struct sbi_scratch *scratch, u32 remote_hartid,
 	 */
 	atomic_raw_set_bit(event, &ipi_data->ipi_type);
 	smp_wmb();
-	sbi_platform_ipi_send(plat, remote_hartid);
+
+	if (ipi_dev && ipi_dev->ipi_send)
+		ipi_dev->ipi_send(remote_hartid);
+
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_IPI_SENT);
 
 	if (ipi_ops->sync)
 		ipi_ops->sync(scratch);
@@ -82,7 +88,7 @@ int sbi_ipi_send_many(ulong hmask, ulong hbase, u32 event, void *data)
 	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
 
 	if (hbase != -1UL) {
-		rc = sbi_hsm_hart_started_mask(dom, hbase, &m);
+		rc = sbi_hsm_hart_interruptible_mask(dom, hbase, &m);
 		if (rc)
 			return rc;
 		m &= hmask;
@@ -94,7 +100,7 @@ int sbi_ipi_send_many(ulong hmask, ulong hbase, u32 event, void *data)
 		}
 	} else {
 		hbase = 0;
-		while (!sbi_hsm_hart_started_mask(dom, hbase, &m)) {
+		while (!sbi_hsm_hart_interruptible_mask(dom, hbase, &m)) {
 			/* Send IPIs */
 			for (i = hbase; m; i++, m >>= 1) {
 				if (m & 1UL)
@@ -178,12 +184,13 @@ void sbi_ipi_process(void)
 	unsigned int ipi_event;
 	const struct sbi_ipi_event_ops *ipi_ops;
 	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
-	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 	struct sbi_ipi_data *ipi_data =
 			sbi_scratch_offset_ptr(scratch, ipi_data_off);
-
 	u32 hartid = current_hartid();
-	sbi_platform_ipi_clear(plat, hartid);
+
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_IPI_RECVD);
+	if (ipi_dev && ipi_dev->ipi_clear)
+		ipi_dev->ipi_clear(hartid);
 
 	ipi_type = atomic_raw_xchg_ulong(&ipi_data->ipi_type, 0);
 	ipi_event = 0;
@@ -201,14 +208,32 @@ skip:
 	};
 }
 
+void sbi_ipi_raw_send(u32 target_hart)
+{
+	if (ipi_dev && ipi_dev->ipi_send)
+		ipi_dev->ipi_send(target_hart);
+}
+
+const struct sbi_ipi_device *sbi_ipi_get_device(void)
+{
+	return ipi_dev;
+}
+
+void sbi_ipi_set_device(const struct sbi_ipi_device *dev)
+{
+	if (!dev || ipi_dev)
+		return;
+
+	ipi_dev = dev;
+}
+
 int sbi_ipi_init(struct sbi_scratch *scratch, bool cold_boot)
 {
 	int ret;
 	struct sbi_ipi_data *ipi_data;
 
 	if (cold_boot) {
-		ipi_data_off = sbi_scratch_alloc_offset(sizeof(*ipi_data),
-							"IPI_DATA");
+		ipi_data_off = sbi_scratch_alloc_offset(sizeof(*ipi_data));
 		if (!ipi_data_off)
 			return SBI_ENOMEM;
 		ret = sbi_ipi_event_create(&ipi_smode_ops);
@@ -230,7 +255,10 @@ int sbi_ipi_init(struct sbi_scratch *scratch, bool cold_boot)
 	ipi_data = sbi_scratch_offset_ptr(scratch, ipi_data_off);
 	ipi_data->ipi_type = 0x00;
 
-	/* Platform init */
+	/*
+	 * Initialize platform IPI support. This will also clear any
+	 * pending IPIs for current/calling HART.
+	 */
 	ret = sbi_platform_ipi_init(sbi_platform_ptr(scratch), cold_boot);
 	if (ret)
 		return ret;

lib/sbi/sbi_misaligned_ldst.c

@@ -12,6 +12,7 @@
 #include <sbi/riscv_fp.h>
 #include <sbi/sbi_error.h>
 #include <sbi/sbi_misaligned_ldst.h>
+#include <sbi/sbi_pmu.h>
 #include <sbi/sbi_trap.h>
 #include <sbi/sbi_unpriv.h>
 
@@ -29,6 +30,8 @@ int sbi_misaligned_load_handler(ulong addr, ulong tval2, ulong tinst,
 	struct sbi_trap_info uptrap;
 	int i, fp = 0, shift = 0, len = 0;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_MISALIGNED_LOAD);
+
 	if (tinst & 0x1) {
 		/*
 		 * Bit[0] == 1 implies trapped instruction value is
@@ -149,6 +152,8 @@ int sbi_misaligned_store_handler(ulong addr, ulong tval2, ulong tinst,
 	struct sbi_trap_info uptrap;
 	int i, len = 0;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_MISALIGNED_STORE);
+
 	if (tinst & 0x1) {
 		/*
 		 * Bit[0] == 1 implies trapped instruction value is

lib/sbi/sbi_platform.c

@@ -19,17 +19,8 @@ static inline char *sbi_platform_feature_id2string(unsigned long feature)
 		return NULL;
 
 	switch (feature) {
-	case SBI_PLATFORM_HAS_TIMER_VALUE:
-		fstr = "timer";
-		break;
-	case SBI_PLATFORM_HAS_HART_HOTPLUG:
-		fstr = "hotplug";
-		break;
 	case SBI_PLATFORM_HAS_MFAULTS_DELEGATION:
-		fstr = "mfdeleg";
-		break;
-	case SBI_PLATFORM_HAS_HART_SECONDARY_BOOT:
-		fstr = "sec_boot";
+		fstr = "medeleg";
 		break;
 	default:
 		break;
@@ -57,9 +48,18 @@ void sbi_platform_get_features_str(const struct sbi_platform *plat,
 		if (features & feat) {
 			temp = sbi_platform_feature_id2string(feat);
 			if (temp) {
-				sbi_snprintf(features_str + offset, nfstr,
+				int len = sbi_snprintf(features_str + offset,
+						       nfstr - offset,
 						       "%s,", temp);
-				offset = offset + sbi_strlen(temp) + 1;
+				if (len < 0)
+					break;
+
+				if (offset + len >= nfstr) {
+					/* No more space for features */
+					offset = nfstr;
+					break;
+				} else
+					offset = offset + len;
 			}
 		}
 		feat = feat << 1;

lib/sbi/sbi_pmu.c

@@ -0,0 +1,753 @@
+/*
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ *
+ * Authors:
+ *   Atish Patra <atish.patra@wdc.com>
+ */
+
+#include <sbi/riscv_asm.h>
+#include <sbi/sbi_bitops.h>
+#include <sbi/sbi_console.h>
+#include <sbi/sbi_hart.h>
+#include <sbi/sbi_platform.h>
+#include <sbi/sbi_pmu.h>
+#include <sbi/sbi_scratch.h>
+#include <sbi/sbi_string.h>
+
+/** Information about hardware counters */
+struct sbi_pmu_hw_event {
+	uint32_t counters;
+	uint32_t start_idx;
+	uint32_t end_idx;
+	/* Event selector value used only for raw events. The event select value
+	 * can be a even id or a selector value for set of events encoded in few
+	 * bits. In case latter, the bits used for encoding of the events should
+	 * be zeroed out in the select value.
+	 */
+	uint64_t select;
+	 /**
+	  * The select_mask indicates which bits are encoded for the event(s).
+	  */
+	uint64_t select_mask;
+};
+
+/** Representation of a firmware event */
+struct sbi_pmu_fw_event {
+	/* Event associated with the particular counter */
+	uint32_t event_idx;
+
+	/* Current value of the counter */
+	unsigned long curr_count;
+
+	/* A flag indicating pmu event monitoring is started */
+	bool bStarted;
+};
+
+/* Information about PMU counters as per SBI specification */
+union sbi_pmu_ctr_info {
+	unsigned long value;
+	struct {
+		unsigned long csr:12;
+		unsigned long width:6;
+#if __riscv_xlen == 32
+		unsigned long reserved:13;
+#else
+		unsigned long reserved:45;
+#endif
+		unsigned long type:1;
+	};
+};
+
+/* Mapping between event range and possible counters  */
+static struct sbi_pmu_hw_event hw_event_map[SBI_PMU_HW_EVENT_MAX] = {0};
+
+/* counter to enabled event mapping */
+static uint32_t active_events[SBI_HARTMASK_MAX_BITS][SBI_PMU_HW_CTR_MAX + SBI_PMU_FW_CTR_MAX];
+
+/* Contains all the information about firmwares events */
+static struct sbi_pmu_fw_event fw_event_map[SBI_HARTMASK_MAX_BITS][SBI_PMU_FW_EVENT_MAX] = {0};
+
+/* Maximum number of hardware events available */
+static uint32_t num_hw_events;
+/* Maximum number of hardware counters available */
+static uint32_t num_hw_ctrs;
+
+/* Maximum number of counters available */
+static uint32_t total_ctrs;
+
+/* Helper macros to retrieve event idx and code type */
+#define get_cidx_type(x) ((x & SBI_PMU_EVENT_IDX_TYPE_MASK) >> 16)
+#define get_cidx_code(x) (x & SBI_PMU_EVENT_IDX_CODE_MASK)
+
+/**
+ * Perform a sanity check on event & counter mappings with event range overlap check
+ * @param evtA Pointer to the existing hw event structure
+ * @param evtB Pointer to the new hw event structure
+ *
+ * Return FALSE if the range doesn't overlap, TRUE otherwise
+ */
+static bool pmu_event_range_overlap(struct sbi_pmu_hw_event *evtA,
+				    struct sbi_pmu_hw_event *evtB)
+{
+	/* check if the range of events overlap with a previous entry */
+	if (((evtA->end_idx < evtB->start_idx) && (evtA->end_idx < evtB->end_idx)) ||
+	   ((evtA->start_idx > evtB->start_idx) && (evtA->start_idx > evtB->end_idx)))
+		return FALSE;
+	return TRUE;
+}
+
+static bool pmu_event_select_overlap(struct sbi_pmu_hw_event *evt,
+				     uint64_t select_val, uint64_t select_mask)
+{
+	if ((evt->select == select_val) && (evt->select_mask == select_mask))
+		return TRUE;
+
+	return FALSE;
+}
+
+static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code)
+{
+	uint32_t event_idx_val;
+	uint32_t event_idx_type;
+	u32 hartid = current_hartid();
+
+	event_idx_val = active_events[hartid][cidx];
+
+	if (cidx >= total_ctrs || (event_idx_val == SBI_PMU_EVENT_IDX_INVALID))
+		return SBI_EINVAL;
+
+	event_idx_type = get_cidx_type(event_idx_val);
+	if (event_idx_type >= SBI_PMU_EVENT_TYPE_MAX)
+		return SBI_EINVAL;
+
+	*event_idx_code = get_cidx_code(event_idx_val);
+
+	return event_idx_type;
+}
+
+static int pmu_ctr_read_fw(uint32_t cidx, unsigned long *cval,
+			       uint32_t fw_evt_code)
+{
+	u32 hartid = current_hartid();
+	struct sbi_pmu_fw_event fevent;
+
+	fevent = fw_event_map[hartid][fw_evt_code];
+	*cval = fevent.curr_count;
+
+	return 0;
+}
+
+/* Add a hardware counter read for completeness for future purpose */
+static int pmu_ctr_read_hw(uint32_t cidx, uint64_t *cval)
+{
+	/* Check for invalid hw counter read requests */
+	if (unlikely(cidx == 1))
+		return SBI_EINVAL;
+#if __riscv_xlen == 32
+	uint32_t temp, temph = 0;
+
+	temp = csr_read_num(CSR_MCYCLE + cidx);
+	temph = csr_read_num(CSR_MCYCLEH + cidx);
+	*cval = ((uint64_t)temph << 32) | temp;
+#else
+	*cval = csr_read_num(CSR_MCYCLE + cidx);
+#endif
+
+	return 0;
+}
+
+int sbi_pmu_ctr_read(uint32_t cidx, unsigned long *cval)
+{
+	int event_idx_type;
+	uint32_t event_code;
+	uint64_t cval64;
+
+	event_idx_type = pmu_ctr_validate(cidx, &event_code);
+	if (event_idx_type < 0)
+		return SBI_EINVAL;
+	else if (event_idx_type == SBI_PMU_EVENT_TYPE_FW)
+		pmu_ctr_read_fw(cidx, cval, event_code);
+	else
+		pmu_ctr_read_hw(cidx, &cval64);
+
+	return 0;
+}
+
+static int pmu_add_hw_event_map(u32 eidx_start, u32 eidx_end, u32 cmap,
+				uint64_t select, uint64_t select_mask)
+{
+	int i = 0;
+	bool is_overlap;
+	struct sbi_pmu_hw_event *event = &hw_event_map[num_hw_events];
+
+	/* The first two counters are reserved by priv spec */
+	if (eidx_start > SBI_PMU_HW_INSTRUCTIONS && (cmap & SBI_PMU_FIXED_CTR_MASK))
+		return SBI_EDENIED;
+
+	if (num_hw_events >= SBI_PMU_HW_EVENT_MAX - 1) {
+		sbi_printf("Can not handle more than %d perf events\n",
+			    SBI_PMU_HW_EVENT_MAX);
+		return SBI_EFAIL;
+	}
+
+	event->start_idx = eidx_start;
+	event->end_idx = eidx_end;
+
+	/* Sanity check */
+	for (i = 0; i < num_hw_events; i++) {
+		if (eidx_start == SBI_PMU_EVENT_RAW_IDX)
+		/* All raw events have same event idx. Just do sanity check on select */
+			is_overlap = pmu_event_select_overlap(&hw_event_map[i],
+							      select, select_mask);
+		else
+			is_overlap = pmu_event_range_overlap(&hw_event_map[i], event);
+		if (is_overlap)
+			goto reset_event;
+	}
+
+	event->select_mask = select_mask;
+	event->counters = cmap;
+	event->select = select;
+	num_hw_events++;
+
+	return 0;
+
+reset_event:
+	event->start_idx = 0;
+	event->end_idx = 0;
+	return SBI_EINVAL;
+}
+
+/**
+ * Logical counter ids are assigned to hardware counters are assigned consecutively.
+ * E.g. counter0 must count MCYCLE where counter2 must count minstret. Similarly,
+ * counterX will mhpmcounterX.
+ */
+int sbi_pmu_add_hw_event_counter_map(u32 eidx_start, u32 eidx_end, u32 cmap)
+{
+	if ((eidx_start > eidx_end) || eidx_start == SBI_PMU_EVENT_RAW_IDX ||
+	     eidx_end == SBI_PMU_EVENT_RAW_IDX)
+		return SBI_EINVAL;
+
+	return pmu_add_hw_event_map(eidx_start, eidx_end, cmap, 0, 0);
+}
+
+int sbi_pmu_add_raw_event_counter_map(uint64_t select, uint64_t select_mask, u32 cmap)
+{
+	return pmu_add_hw_event_map(SBI_PMU_EVENT_RAW_IDX,
+				    SBI_PMU_EVENT_RAW_IDX, cmap, select, select_mask);
+}
+
+static int pmu_ctr_enable_irq_hw(int ctr_idx)
+{
+	unsigned long mhpmevent_csr;
+	unsigned long mhpmevent_curr;
+	unsigned long mip_val;
+	unsigned long of_mask;
+
+	if (ctr_idx < 3 || ctr_idx >= SBI_PMU_HW_CTR_MAX)
+		return SBI_EFAIL;
+
+#if __riscv_xlen == 32
+	mhpmevent_csr = CSR_MHPMEVENT3H  + ctr_idx - 3;
+	of_mask = ~MHPMEVENTH_OF;
+#else
+	mhpmevent_csr = CSR_MHPMEVENT3 + ctr_idx - 3;
+	of_mask = ~MHPMEVENT_OF;
+#endif
+
+	mhpmevent_curr = csr_read_num(mhpmevent_csr);
+	mip_val = csr_read(CSR_MIP);
+	/**
+	 * Clear out the OF bit so that next interrupt can be enabled.
+	 * This should be done only when the corresponding overflow interrupt
+	 * bit is cleared. That indicates that software has already handled the
+	 * previous interrupts or the hardware yet to set an overflow interrupt.
+	 * Otherwise, there will be race conditions where we may clear the bit
+	 * the software is yet to handle the interrupt.
+	 */
+	if (!(mip_val & MIP_LCOFIP)) {
+		mhpmevent_curr &= of_mask;
+		csr_write_num(mhpmevent_csr, mhpmevent_curr);
+	}
+
+	return 0;
+}
+
+static void pmu_ctr_write_hw(uint32_t cidx, uint64_t ival)
+{
+#if __riscv_xlen == 32
+	csr_write_num(CSR_MCYCLE + cidx, 0);
+	csr_write_num(CSR_MCYCLE + cidx, ival & 0xFFFFFFFF);
+	csr_write_num(CSR_MCYCLEH + cidx, ival >> BITS_PER_LONG);
+#else
+	csr_write_num(CSR_MCYCLE + cidx, ival);
+#endif
+}
+
+static int pmu_ctr_start_hw(uint32_t cidx, uint64_t ival, bool ival_update)
+{
+	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
+	unsigned long mctr_inhbt;
+
+	/* Make sure the counter index lies within the range and is not TM bit */
+	if (cidx > num_hw_ctrs || cidx == 1)
+		return SBI_EINVAL;
+
+	if (!sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTINHIBIT))
+		goto skip_inhibit_update;
+
+	/*
+	 * Some of the hardware may not support mcountinhibit but perf stat
+	 * still can work if supervisor mode programs the initial value.
+	 */
+	mctr_inhbt = csr_read(CSR_MCOUNTINHIBIT);
+	if (!__test_bit(cidx, &mctr_inhbt))
+		return SBI_EALREADY_STARTED;
+
+	__clear_bit(cidx, &mctr_inhbt);
+
+	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_SSCOFPMF))
+		pmu_ctr_enable_irq_hw(cidx);
+	csr_write(CSR_MCOUNTINHIBIT, mctr_inhbt);
+
+skip_inhibit_update:
+	if (ival_update)
+		pmu_ctr_write_hw(cidx, ival);
+
+	return 0;
+}
+
+static int pmu_ctr_start_fw(uint32_t cidx, uint32_t fw_evt_code,
+			    uint64_t ival, bool ival_update)
+{
+	u32 hartid = current_hartid();
+	struct sbi_pmu_fw_event *fevent;
+
+	fevent = &fw_event_map[hartid][fw_evt_code];
+	if (ival_update)
+		fevent->curr_count = ival;
+	fevent->bStarted = TRUE;
+
+	return 0;
+}
+
+int sbi_pmu_ctr_start(unsigned long cbase, unsigned long cmask,
+		      unsigned long flags, uint64_t ival)
+{
+	int event_idx_type;
+	uint32_t event_code;
+	unsigned long ctr_mask = cmask << cbase;
+	int ret = SBI_EINVAL;
+	bool bUpdate = FALSE;
+
+	if (__fls(ctr_mask) >= total_ctrs)
+		return ret;
+
+	if (flags & SBI_PMU_START_FLAG_SET_INIT_VALUE)
+		bUpdate = TRUE;
+
+	for_each_set_bit_from(cbase, &ctr_mask, total_ctrs) {
+		event_idx_type = pmu_ctr_validate(cbase, &event_code);
+		if (event_idx_type < 0)
+			/* Continue the start operation for other counters */
+			continue;
+		else if (event_idx_type == SBI_PMU_EVENT_TYPE_FW)
+			ret = pmu_ctr_start_fw(cbase, event_code, ival, bUpdate);
+		else
+			ret = pmu_ctr_start_hw(cbase, ival, bUpdate);
+	}
+
+	return ret;
+}
+
+static int pmu_ctr_stop_hw(uint32_t cidx)
+{
+	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
+	unsigned long mctr_inhbt;
+
+	if (!sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTINHIBIT))
+		return 0;
+
+	mctr_inhbt = csr_read(CSR_MCOUNTINHIBIT);
+
+	/* Make sure the counter index lies within the range and is not TM bit */
+	if (cidx > num_hw_ctrs || cidx == 1)
+		return SBI_EINVAL;
+
+	if (!__test_bit(cidx, &mctr_inhbt)) {
+		__set_bit(cidx, &mctr_inhbt);
+		csr_write(CSR_MCOUNTINHIBIT, mctr_inhbt);
+		return 0;
+	} else
+		return SBI_EALREADY_STOPPED;
+}
+
+static int pmu_ctr_stop_fw(uint32_t cidx, uint32_t fw_evt_code)
+{
+	u32 hartid = current_hartid();
+
+	fw_event_map[hartid][fw_evt_code].bStarted = FALSE;
+
+	return 0;
+}
+
+static int pmu_reset_hw_mhpmevent(int ctr_idx)
+{
+	if (ctr_idx < 3 || ctr_idx >= SBI_PMU_HW_CTR_MAX)
+		return SBI_EFAIL;
+#if __riscv_xlen == 32
+	csr_write_num(CSR_MHPMEVENT3 + ctr_idx - 3, 0);
+	csr_write_num(CSR_MHPMEVENT3H + ctr_idx - 3, 0);
+#else
+	csr_write_num(CSR_MHPMEVENT3 + ctr_idx - 3, 0);
+#endif
+
+	return 0;
+}
+
+int sbi_pmu_ctr_stop(unsigned long cbase, unsigned long cmask,
+		     unsigned long flag)
+{
+	u32 hartid = current_hartid();
+	int ret = SBI_EINVAL;
+	int event_idx_type;
+	uint32_t event_code;
+	unsigned long ctr_mask = cmask << cbase;
+
+	if (__fls(ctr_mask) >= total_ctrs)
+		return SBI_EINVAL;
+
+	for_each_set_bit_from(cbase, &ctr_mask, total_ctrs) {
+		event_idx_type = pmu_ctr_validate(cbase, &event_code);
+		if (event_idx_type < 0)
+			/* Continue the stop operation for other counters */
+			continue;
+
+		else if (event_idx_type == SBI_PMU_EVENT_TYPE_FW)
+			ret = pmu_ctr_stop_fw(cbase, event_code);
+		else
+			ret = pmu_ctr_stop_hw(cbase);
+
+		if (flag & SBI_PMU_STOP_FLAG_RESET) {
+			active_events[hartid][cbase] = SBI_PMU_EVENT_IDX_INVALID;
+			pmu_reset_hw_mhpmevent(cbase);
+		}
+	}
+
+	return ret;
+}
+
+static void pmu_update_inhibit_flags(unsigned long flags, uint64_t *mhpmevent_val)
+{
+	if (flags & SBI_PMU_CFG_FLAG_SET_VUINH)
+		*mhpmevent_val |= MHPMEVENT_VUINH;
+	if (flags & SBI_PMU_CFG_FLAG_SET_VSINH)
+		*mhpmevent_val |= MHPMEVENT_VSINH;
+	if (flags & SBI_PMU_CFG_FLAG_SET_UINH)
+		*mhpmevent_val |= MHPMEVENT_UINH;
+	if (flags & SBI_PMU_CFG_FLAG_SET_SINH)
+		*mhpmevent_val |= MHPMEVENT_SINH;
+}
+
+static int pmu_update_hw_mhpmevent(struct sbi_pmu_hw_event *hw_evt, int ctr_idx,
+				   unsigned long flags, unsigned long eindex,
+				   uint64_t data)
+{
+	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
+	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
+	uint64_t mhpmevent_val;
+
+	/* Get the final mhpmevent value to be written from platform */
+	mhpmevent_val = sbi_platform_pmu_xlate_to_mhpmevent(plat, eindex, data);
+
+	if (!mhpmevent_val || ctr_idx < 3 || ctr_idx >= SBI_PMU_HW_CTR_MAX)
+		return SBI_EFAIL;
+
+	/* Always clear the OVF bit and inhibit countin of events in M-mode */
+	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_SSCOFPMF))
+		mhpmevent_val = (mhpmevent_val & ~MHPMEVENT_SSCOF_MASK) | MHPMEVENT_MINH;
+
+	/* Update the inhibit flags based on inhibit flags received from supervisor */
+	pmu_update_inhibit_flags(flags, &mhpmevent_val);
+
+#if __riscv_xlen == 32
+	csr_write_num(CSR_MHPMEVENT3 + ctr_idx - 3, mhpmevent_val & 0xFFFFFFFF);
+	csr_write_num(CSR_MHPMEVENT3H + ctr_idx - 3, mhpmevent_val >> BITS_PER_LONG);
+#else
+	csr_write_num(CSR_MHPMEVENT3 + ctr_idx - 3, mhpmevent_val);
+#endif
+
+	return 0;
+}
+
+static int pmu_ctr_find_fixed_fw(unsigned long evt_idx_code)
+{
+	/* Non-programmables counters are enabled always. No need to do lookup */
+	if (evt_idx_code == SBI_PMU_HW_CPU_CYCLES)
+		return 0;
+	else if (evt_idx_code == SBI_PMU_HW_INSTRUCTIONS)
+		return 2;
+	else
+		return SBI_EINVAL;
+}
+
+static int pmu_ctr_find_hw(unsigned long cbase, unsigned long cmask, unsigned long flags,
+			   unsigned long event_idx, uint64_t data)
+{
+	unsigned long ctr_mask;
+	int i, ret = 0, fixed_ctr, ctr_idx = SBI_ENOTSUPP;
+	struct sbi_pmu_hw_event *temp;
+	unsigned long mctr_inhbt = 0;
+	u32 hartid = current_hartid();
+	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
+
+	if (cbase > num_hw_ctrs)
+		return SBI_EINVAL;
+
+	/**
+	 * If Sscof is present try to find the programmable counter for
+	 * cycle/instret as well.
+	 */
+	fixed_ctr = pmu_ctr_find_fixed_fw(event_idx);
+	if (fixed_ctr >= 0 &&
+	    !sbi_hart_has_feature(scratch, SBI_HART_HAS_SSCOFPMF))
+		return fixed_ctr;
+
+	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTINHIBIT))
+		mctr_inhbt = csr_read(CSR_MCOUNTINHIBIT);
+	for (i = 0; i < num_hw_events; i++) {
+		temp = &hw_event_map[i];
+		if ((temp->start_idx > event_idx && event_idx < temp->end_idx) ||
+		    (temp->start_idx < event_idx && event_idx > temp->end_idx))
+			continue;
+
+		/* For raw events, event data is used as the select value */
+		if (event_idx == SBI_PMU_EVENT_RAW_IDX) {
+			uint64_t select_mask = temp->select_mask;
+
+			/* The non-event map bits of data should match the selector */
+			if (temp->select != (data & select_mask))
+				continue;
+		}
+		/* Fixed counters should not be part of the search */
+		ctr_mask = temp->counters & (cmask << cbase) &
+			   (~SBI_PMU_FIXED_CTR_MASK);
+		for_each_set_bit_from(cbase, &ctr_mask, SBI_PMU_HW_CTR_MAX) {
+			/**
+			 * Some of the platform may not support mcountinhibit.
+			 * Checking the active_events is enough for them
+			 */
+			if (active_events[hartid][cbase] != SBI_PMU_EVENT_IDX_INVALID)
+				continue;
+			/* If mcountinhibit is supported, the bit must be enabled */
+			if ((sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTINHIBIT)) &&
+			    !__test_bit(cbase, &mctr_inhbt))
+				continue;
+			/* We found a valid counter that is not started yet */
+			ctr_idx = cbase;
+		}
+	}
+
+	if (ctr_idx == SBI_ENOTSUPP) {
+		/**
+		 * We can't find any programmable counters for cycle/instret.
+		 * Return the fixed counter as they are mandatory anyways.
+		 */
+		if (fixed_ctr >= 0)
+			return fixed_ctr;
+		else
+			return SBI_EFAIL;
+	}
+	ret = pmu_update_hw_mhpmevent(temp, ctr_idx, flags, event_idx, data);
+
+	if (!ret)
+		ret = ctr_idx;
+
+	return ret;
+}
+
+
+/**
+ * Any firmware counter can map to any firmware event.
+ * Thus, select the first available fw counter after sanity
+ * check.
+ */
+static int pmu_ctr_find_fw(unsigned long cbase, unsigned long cmask, u32 hartid)
+{
+	int i = 0;
+	int fw_base;
+	unsigned long ctr_mask = cmask << cbase;
+
+	if (cbase <= num_hw_ctrs)
+		fw_base = num_hw_ctrs + 1;
+	else
+		fw_base = cbase;
+
+	for (i = fw_base; i < total_ctrs; i++)
+		if ((active_events[hartid][i] == SBI_PMU_EVENT_IDX_INVALID) &&
+		    ((1UL << i) & ctr_mask))
+			return i;
+
+	return SBI_ENOTSUPP;
+}
+
+int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,
+			  unsigned long flags, unsigned long event_idx,
+			  uint64_t event_data)
+{
+	int ctr_idx = SBI_ENOTSUPP;
+	u32 hartid = current_hartid();
+	int event_type = get_cidx_type(event_idx);
+	struct sbi_pmu_fw_event *fevent;
+	uint32_t fw_evt_code;
+	unsigned long tmp = cidx_mask << cidx_base;
+
+	/* Do a basic sanity check of counter base & mask */
+	if (__fls(tmp) >= total_ctrs || event_type >= SBI_PMU_EVENT_TYPE_MAX)
+		return SBI_EINVAL;
+
+	if (flags & SBI_PMU_CFG_FLAG_SKIP_MATCH) {
+		/* The caller wants to skip the match because it already knows the
+		 * counter idx for the given event. Verify that the counter idx
+		 * is still valid.
+		 */
+		if (active_events[hartid][cidx_base] == SBI_PMU_EVENT_IDX_INVALID)
+			return SBI_EINVAL;
+		ctr_idx = cidx_base;
+		goto skip_match;
+	}
+
+	if (event_type == SBI_PMU_EVENT_TYPE_FW) {
+		/* Any firmware counter can be used track any firmware event */
+		ctr_idx = pmu_ctr_find_fw(cidx_base, cidx_mask, hartid);
+	} else {
+		ctr_idx = pmu_ctr_find_hw(cidx_base, cidx_mask, flags, event_idx,
+					  event_data);
+	}
+
+	if (ctr_idx < 0)
+		return SBI_ENOTSUPP;
+
+	active_events[hartid][ctr_idx] = event_idx;
+skip_match:
+	if (event_type == SBI_PMU_EVENT_TYPE_HW) {
+		if (flags & SBI_PMU_CFG_FLAG_CLEAR_VALUE)
+			pmu_ctr_write_hw(ctr_idx, 0);
+		if (flags & SBI_PMU_CFG_FLAG_AUTO_START)
+			pmu_ctr_start_hw(ctr_idx, 0, false);
+	} else if (event_type == SBI_PMU_EVENT_TYPE_FW) {
+		fw_evt_code = get_cidx_code(event_idx);
+		fevent = &fw_event_map[hartid][fw_evt_code];
+		if (flags & SBI_PMU_CFG_FLAG_CLEAR_VALUE)
+			fevent->curr_count = 0;
+		if (flags & SBI_PMU_CFG_FLAG_AUTO_START)
+			fevent->bStarted = TRUE;
+	}
+
+	return ctr_idx;
+}
+
+inline int sbi_pmu_ctr_incr_fw(enum sbi_pmu_fw_event_code_id fw_id)
+{
+	u32 hartid = current_hartid();
+	struct sbi_pmu_fw_event *fevent;
+
+	if (unlikely(fw_id >= SBI_PMU_FW_MAX))
+		return SBI_EINVAL;
+
+	fevent = &fw_event_map[hartid][fw_id];
+
+	/* PMU counters will be only enabled during performance debugging */
+	if (unlikely(fevent->bStarted))
+		fevent->curr_count++;
+
+	return 0;
+}
+
+unsigned long sbi_pmu_num_ctr(void)
+{
+	return (num_hw_ctrs + SBI_PMU_FW_CTR_MAX);
+}
+
+int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)
+{
+	union sbi_pmu_ctr_info cinfo = {0};
+	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
+
+	/* Sanity check. Counter1 is not mapped at all */
+	if (cidx >= total_ctrs || cidx == 1)
+		return SBI_EINVAL;
+
+	/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */
+	if (cidx <= num_hw_ctrs) {
+		cinfo.type = SBI_PMU_CTR_TYPE_HW;
+		cinfo.csr = CSR_CYCLE + cidx;
+		/* mcycle & minstret are always 64 bit */
+		if (cidx == 0 || cidx == 2)
+			cinfo.width = 63;
+		else
+			cinfo.width = sbi_hart_mhpm_bits(scratch) - 1;
+	} else {
+		/* it's a firmware counter */
+		cinfo.type = SBI_PMU_CTR_TYPE_FW;
+		/* Firmware counters are XLEN bits wide */
+		cinfo.width = BITS_PER_LONG - 1;
+	}
+
+	*ctr_info = cinfo.value;
+
+	return 0;
+}
+
+static void pmu_reset_event_map(u32 hartid)
+{
+	int j;
+
+	/* Initialize the counter to event mapping table */
+	for (j = 3; j < total_ctrs; j++)
+		active_events[hartid][j] = SBI_PMU_EVENT_IDX_INVALID;
+	for (j = 0; j < SBI_PMU_FW_CTR_MAX; j++)
+		sbi_memset(&fw_event_map[hartid][j], 0,
+			   sizeof(struct sbi_pmu_fw_event));
+}
+
+void sbi_pmu_exit(struct sbi_scratch *scratch)
+{
+	u32 hartid = current_hartid();
+
+	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_MCOUNTINHIBIT))
+		csr_write(CSR_MCOUNTINHIBIT, 0xFFFFFFF8);
+
+	csr_write(CSR_MCOUNTEREN, -1);
+	pmu_reset_event_map(hartid);
+}
+
+int sbi_pmu_init(struct sbi_scratch *scratch, bool cold_boot)
+{
+	const struct sbi_platform *plat;
+	u32 hartid = current_hartid();
+
+	if (cold_boot) {
+		plat = sbi_platform_ptr(scratch);
+		/* Initialize hw pmu events */
+		sbi_platform_pmu_init(plat);
+
+		/* mcycle & minstret is available always */
+		num_hw_ctrs = sbi_hart_mhpm_count(scratch) + 2;
+		total_ctrs = num_hw_ctrs + SBI_PMU_FW_CTR_MAX;
+	}
+
+	pmu_reset_event_map(hartid);
+
+	/* First three counters are fixed by the priv spec and we enable it by default */
+	active_events[hartid][0] = SBI_PMU_EVENT_TYPE_HW << SBI_PMU_EVENT_IDX_OFFSET |
+				   SBI_PMU_HW_CPU_CYCLES;
+	active_events[hartid][1] = SBI_PMU_EVENT_IDX_INVALID;
+	active_events[hartid][2] = SBI_PMU_EVENT_TYPE_HW << SBI_PMU_EVENT_IDX_OFFSET |
+				   SBI_PMU_HW_INSTRUCTIONS;
+
+	return 0;
+}

lib/sbi/sbi_scratch.c

@@ -14,7 +14,7 @@
 #include <sbi/sbi_scratch.h>
 #include <sbi/sbi_string.h>
 
-u32 last_hartid_having_scratch = SBI_HARTMASK_MAX_BITS;
+u32 last_hartid_having_scratch = SBI_HARTMASK_MAX_BITS - 1;
 struct sbi_scratch *hartid_to_scratch_table[SBI_HARTMASK_MAX_BITS] = { 0 };
 
 static spinlock_t extra_lock = SPIN_LOCK_INITIALIZER;
@@ -40,7 +40,7 @@ int sbi_scratch_init(struct sbi_scratch *scratch)
 	return 0;
 }
 
-unsigned long sbi_scratch_alloc_offset(unsigned long size, const char *owner)
+unsigned long sbi_scratch_alloc_offset(unsigned long size)
 {
 	u32 i;
 	void *ptr;
@@ -74,7 +74,7 @@ done:
 	spin_unlock(&extra_lock);
 
 	if (ret) {
-		for (i = 0; i < sbi_scratch_last_hartid(); i++) {
+		for (i = 0; i <= sbi_scratch_last_hartid(); i++) {
 			rscratch = sbi_hartid_to_scratch(i);
 			if (!rscratch)
 				continue;

lib/sbi/sbi_string.c

@@ -33,6 +33,10 @@ int sbi_strncmp(const char *a, const char *b, size_t count)
 	for (; count > 0 && *a == *b && *a != '\0'; a++, b++, count--)
 		;
 
+	/* No difference till the end */
+	if (!count)
+		return 0;
+
 	return *a - *b;
 }
 

lib/sbi/sbi_system.c

@@ -18,13 +18,48 @@
 #include <sbi/sbi_ipi.h>
 #include <sbi/sbi_init.h>
 
+static SBI_LIST_HEAD(reset_devices_list);
+
+const struct sbi_system_reset_device *sbi_system_reset_get_device(
+					u32 reset_type, u32 reset_reason)
+{
+	struct sbi_system_reset_device *reset_dev = NULL;
+	struct sbi_dlist *pos;
+	/** lowest priority - any non zero is our candidate */
+	int priority = 0;
+
+	/* Check each reset device registered for supported reset type */
+	sbi_list_for_each(pos, &(reset_devices_list)) {
+		struct sbi_system_reset_device *dev =
+			to_system_reset_device(pos);
+		if (dev->system_reset_check) {
+			int status = dev->system_reset_check(reset_type,
+							     reset_reason);
+			/** reset_type not supported */
+			if (status == 0)
+				continue;
+
+			if (status > priority) {
+				reset_dev = dev;
+				priority = status;
+			}
+		}
+	}
+
+	return reset_dev;
+}
+
+void sbi_system_reset_add_device(struct sbi_system_reset_device *dev)
+{
+	if (!dev || !dev->system_reset_check)
+		return;
+
+	sbi_list_add(&(dev->node), &(reset_devices_list));
+}
+
 bool sbi_system_reset_supported(u32 reset_type, u32 reset_reason)
 {
-	if (sbi_platform_system_reset_check(sbi_platform_thishart_ptr(),
-					    reset_type, reset_reason))
-		return TRUE;
-
-	return FALSE;
+	return !!sbi_system_reset_get_device(reset_type, reset_reason);
 }
 
 void __noreturn sbi_system_reset(u32 reset_type, u32 reset_reason)
@@ -35,7 +70,7 @@ void __noreturn sbi_system_reset(u32 reset_type, u32 reset_reason)
 	struct sbi_scratch *scratch = sbi_scratch_thishart_ptr();
 
 	/* Send HALT IPI to every hart other than the current hart */
-	while (!sbi_hsm_hart_started_mask(dom, hbase, &hmask)) {
+	while (!sbi_hsm_hart_interruptible_mask(dom, hbase, &hmask)) {
 		if (hbase <= cur_hartid)
 			hmask &= ~(1UL << (cur_hartid - hbase));
 		if (hmask)
@@ -47,9 +82,12 @@ void __noreturn sbi_system_reset(u32 reset_type, u32 reset_reason)
 	sbi_hsm_hart_stop(scratch, FALSE);
 
 	/* Platform specific reset if domain allowed system reset */
-	if (dom->system_reset_allowed)
-		sbi_platform_system_reset(sbi_platform_ptr(scratch),
-					  reset_type, reset_reason);
+	if (dom->system_reset_allowed) {
+		const struct sbi_system_reset_device *dev =
+			sbi_system_reset_get_device(reset_type, reset_reason);
+		if (dev)
+			dev->system_reset(reset_type, reset_reason);
+	}
 
 	/* If platform specific reset did not work then do sbi_exit() */
 	sbi_exit(scratch);

lib/sbi/sbi_timer.c

@@ -8,18 +8,22 @@
  */
 
 #include <sbi/riscv_asm.h>
+#include <sbi/riscv_barrier.h>
 #include <sbi/riscv_encoding.h>
+#include <sbi/sbi_console.h>
 #include <sbi/sbi_error.h>
 #include <sbi/sbi_hart.h>
 #include <sbi/sbi_platform.h>
+#include <sbi/sbi_pmu.h>
 #include <sbi/sbi_scratch.h>
 #include <sbi/sbi_timer.h>
 
 static unsigned long time_delta_off;
-static u64 (*get_time_val)(const struct sbi_platform *plat);
+static u64 (*get_time_val)(void);
+static const struct sbi_timer_device *timer_dev = NULL;
 
 #if __riscv_xlen == 32
-static u64 get_ticks(const struct sbi_platform *plat)
+static u64 get_ticks(void)
 {
 	u32 lo, hi, tmp;
 	__asm__ __volatile__("1:\n"
@@ -31,7 +35,7 @@ static u64 get_ticks(const struct sbi_platform *plat)
 	return ((u64)hi << 32) | lo;
 }
 #else
-static u64 get_ticks(const struct sbi_platform *plat)
+static u64 get_ticks(void)
 {
 	unsigned long n;
 
@@ -40,9 +44,48 @@ static u64 get_ticks(const struct sbi_platform *plat)
 }
 #endif
 
+static u64 get_platform_ticks(void)
+{
+	return timer_dev->timer_value();
+}
+
+static void nop_delay_fn(void *opaque)
+{
+	cpu_relax();
+}
+
+void sbi_timer_delay_loop(ulong units, u64 unit_freq,
+			  void (*delay_fn)(void *), void *opaque)
+{
+	u64 start_val, delta;
+
+	/* Do nothing if we don't have timer device */
+	if (!timer_dev || !get_time_val) {
+		sbi_printf("%s: called without timer device\n", __func__);
+		return;
+	}
+
+	/* Save starting timer value */
+	start_val = get_time_val();
+
+	/* Compute desired timer value delta */
+	delta = ((u64)timer_dev->timer_freq * (u64)units);
+	delta = delta / unit_freq;
+
+	/* Use NOP delay function if delay function not available */
+	if (!delay_fn)
+		delay_fn = nop_delay_fn;
+
+	/* Busy loop until desired timer value delta reached */
+	while ((get_time_val() - start_val) < delta)
+		delay_fn(opaque);
+}
+
 u64 sbi_timer_value(void)
 {
-	return get_time_val(sbi_platform_thishart_ptr());
+	if (get_time_val)
+		return get_time_val();
+	return 0;
 }
 
 u64 sbi_timer_virt_value(void)
@@ -80,7 +123,9 @@ void sbi_timer_set_delta_upper(ulong delta_upper)
 
 void sbi_timer_event_start(u64 next_event)
 {
-	sbi_platform_timer_event_start(sbi_platform_thishart_ptr(), next_event);
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_SET_TIMER);
+	if (timer_dev && timer_dev->timer_event_start)
+		timer_dev->timer_event_start(next_event);
 	csr_clear(CSR_MIP, MIP_STIP);
 	csr_set(CSR_MIE, MIP_MTIP);
 }
@@ -91,17 +136,33 @@ void sbi_timer_process(void)
 	csr_set(CSR_MIP, MIP_STIP);
 }
 
+const struct sbi_timer_device *sbi_timer_get_device(void)
+{
+	return timer_dev;
+}
+
+void sbi_timer_set_device(const struct sbi_timer_device *dev)
+{
+	if (!dev || timer_dev)
+		return;
+
+	timer_dev = dev;
+	if (!get_time_val && timer_dev->timer_value)
+		get_time_val = get_platform_ticks;
+}
+
 int sbi_timer_init(struct sbi_scratch *scratch, bool cold_boot)
 {
 	u64 *time_delta;
 	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
-	int ret;
 
 	if (cold_boot) {
-		time_delta_off = sbi_scratch_alloc_offset(sizeof(*time_delta),
-							  "TIME_DELTA");
+		time_delta_off = sbi_scratch_alloc_offset(sizeof(*time_delta));
 		if (!time_delta_off)
 			return SBI_ENOMEM;
+
+		if (sbi_hart_has_feature(scratch, SBI_HART_HAS_TIME))
+			get_time_val = get_ticks;
 	} else {
 		if (!time_delta_off)
 			return SBI_ENOMEM;
@@ -110,24 +171,13 @@ int sbi_timer_init(struct sbi_scratch *scratch, bool cold_boot)
 	time_delta = sbi_scratch_offset_ptr(scratch, time_delta_off);
 	*time_delta = 0;
 
-	ret = sbi_platform_timer_init(plat, cold_boot);
-	if (ret)
-		return ret;
-
-	if (sbi_hart_has_feature(scratch, SBI_HART_HAS_TIME))
-		get_time_val = get_ticks;
-	else if (sbi_platform_has_timer_value(plat))
-		get_time_val = sbi_platform_timer_value;
-	else
-		/* There is no method to provide timer value */
-		return SBI_ENODEV;
-
-	return 0;
+	return sbi_platform_timer_init(plat, cold_boot);
 }
 
 void sbi_timer_exit(struct sbi_scratch *scratch)
 {
-	sbi_platform_timer_event_stop(sbi_platform_ptr(scratch));
+	if (timer_dev && timer_dev->timer_event_stop)
+		timer_dev->timer_event_stop();
 
 	csr_clear(CSR_MIP, MIP_STIP);
 	csr_clear(CSR_MIE, MIP_MTIP);

lib/sbi/sbi_tlb.c

@@ -21,13 +21,14 @@
 #include <sbi/sbi_string.h>
 #include <sbi/sbi_console.h>
 #include <sbi/sbi_platform.h>
+#include <sbi/sbi_pmu.h>
 
 static unsigned long tlb_sync_off;
 static unsigned long tlb_fifo_off;
 static unsigned long tlb_fifo_mem_off;
 static unsigned long tlb_range_flush_limit;
 
-static void sbi_tlb_flush_all(void)
+static void tlb_flush_all(void)
 {
 	__asm__ __volatile("sfence.vma");
 }
@@ -39,6 +40,8 @@ void sbi_tlb_local_hfence_vvma(struct sbi_tlb_info *tinfo)
 	unsigned long vmid  = tinfo->vmid;
 	unsigned long i, hgatp;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_VVMA_RCVD);
+
 	hgatp = csr_swap(CSR_HGATP,
 			 (vmid << HGATP_VMID_SHIFT) & HGATP_VMID_MASK);
 
@@ -61,13 +64,15 @@ void sbi_tlb_local_hfence_gvma(struct sbi_tlb_info *tinfo)
 	unsigned long size  = tinfo->size;
 	unsigned long i;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_GVMA_RCVD);
+
 	if ((start == 0 && size == 0) || (size == SBI_TLB_FLUSH_ALL)) {
 		__sbi_hfence_gvma_all();
 		return;
 	}
 
 	for (i = 0; i < size; i += PAGE_SIZE) {
-		__sbi_hfence_gvma_gpa(start+i);
+		__sbi_hfence_gvma_gpa((start + i) >> 2);
 	}
 }
 
@@ -77,8 +82,10 @@ void sbi_tlb_local_sfence_vma(struct sbi_tlb_info *tinfo)
 	unsigned long size  = tinfo->size;
 	unsigned long i;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_SFENCE_VMA_RCVD);
+
 	if ((start == 0 && size == 0) || (size == SBI_TLB_FLUSH_ALL)) {
-		sbi_tlb_flush_all();
+		tlb_flush_all();
 		return;
 	}
 
@@ -98,6 +105,8 @@ void sbi_tlb_local_hfence_vvma_asid(struct sbi_tlb_info *tinfo)
 	unsigned long vmid  = tinfo->vmid;
 	unsigned long i, hgatp;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_VVMA_ASID_RCVD);
+
 	hgatp = csr_swap(CSR_HGATP,
 			 (vmid << HGATP_VMID_SHIFT) & HGATP_VMID_MASK);
 
@@ -126,6 +135,8 @@ void sbi_tlb_local_hfence_gvma_vmid(struct sbi_tlb_info *tinfo)
 	unsigned long vmid  = tinfo->vmid;
 	unsigned long i;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_GVMA_VMID_RCVD);
+
 	if (start == 0 && size == 0) {
 		__sbi_hfence_gvma_all();
 		return;
@@ -137,7 +148,7 @@ void sbi_tlb_local_hfence_gvma_vmid(struct sbi_tlb_info *tinfo)
 	}
 
 	for (i = 0; i < size; i += PAGE_SIZE) {
-		__sbi_hfence_gvma_vmid_gpa(start + i, vmid);
+		__sbi_hfence_gvma_vmid_gpa((start + i) >> 2, vmid);
 	}
 }
 
@@ -148,8 +159,10 @@ void sbi_tlb_local_sfence_vma_asid(struct sbi_tlb_info *tinfo)
 	unsigned long asid  = tinfo->asid;
 	unsigned long i;
 
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_SFENCE_VMA_ASID_RCVD);
+
 	if (start == 0 && size == 0) {
-		sbi_tlb_flush_all();
+		tlb_flush_all();
 		return;
 	}
 
@@ -172,10 +185,33 @@ void sbi_tlb_local_sfence_vma_asid(struct sbi_tlb_info *tinfo)
 
 void sbi_tlb_local_fence_i(struct sbi_tlb_info *tinfo)
 {
+	sbi_pmu_ctr_incr_fw(SBI_PMU_FW_FENCE_I_RECVD);
+
 	__asm__ __volatile("fence.i");
 }
 
-static void sbi_tlb_entry_process(struct sbi_tlb_info *tinfo)
+static void tlb_pmu_incr_fw_ctr(struct sbi_tlb_info *data)
+{
+	if (unlikely(!data))
+		return;
+
+	if (data->local_fn == sbi_tlb_local_fence_i)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_FENCE_I_SENT);
+	else if (data->local_fn == sbi_tlb_local_sfence_vma)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_SFENCE_VMA_SENT);
+	else if (data->local_fn == sbi_tlb_local_sfence_vma_asid)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_SFENCE_VMA_ASID_SENT);
+	else if (data->local_fn == sbi_tlb_local_hfence_gvma)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_GVMA_SENT);
+	else if (data->local_fn == sbi_tlb_local_hfence_gvma_vmid)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_GVMA_VMID_SENT);
+	else if (data->local_fn == sbi_tlb_local_hfence_vvma)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_VVMA_SENT);
+	else if (data->local_fn == sbi_tlb_local_hfence_vvma_asid)
+		sbi_pmu_ctr_incr_fw(SBI_PMU_FW_HFENCE_VVMA_ASID_SENT);
+}
+
+static void tlb_entry_process(struct sbi_tlb_info *tinfo)
 {
 	u32 rhartid;
 	struct sbi_scratch *rscratch = NULL;
@@ -193,15 +229,15 @@ static void sbi_tlb_entry_process(struct sbi_tlb_info *tinfo)
 	}
 }
 
-static void sbi_tlb_process_count(struct sbi_scratch *scratch, int count)
+static void tlb_process_count(struct sbi_scratch *scratch, int count)
 {
 	struct sbi_tlb_info tinfo;
-	u32 deq_count = 0;
+	unsigned int deq_count = 0;
 	struct sbi_fifo *tlb_fifo =
 			sbi_scratch_offset_ptr(scratch, tlb_fifo_off);
 
 	while (!sbi_fifo_dequeue(tlb_fifo, &tinfo)) {
-		sbi_tlb_entry_process(&tinfo);
+		tlb_entry_process(&tinfo);
 		deq_count++;
 		if (deq_count > count)
 			break;
@@ -209,17 +245,17 @@ static void sbi_tlb_process_count(struct sbi_scratch *scratch, int count)
 	}
 }
 
-static void sbi_tlb_process(struct sbi_scratch *scratch)
+static void tlb_process(struct sbi_scratch *scratch)
 {
 	struct sbi_tlb_info tinfo;
 	struct sbi_fifo *tlb_fifo =
 			sbi_scratch_offset_ptr(scratch, tlb_fifo_off);
 
 	while (!sbi_fifo_dequeue(tlb_fifo, &tinfo))
-		sbi_tlb_entry_process(&tinfo);
+		tlb_entry_process(&tinfo);
 }
 
-static void sbi_tlb_sync(struct sbi_scratch *scratch)
+static void tlb_sync(struct sbi_scratch *scratch)
 {
 	unsigned long *tlb_sync =
 			sbi_scratch_offset_ptr(scratch, tlb_sync_off);
@@ -229,13 +265,13 @@ static void sbi_tlb_sync(struct sbi_scratch *scratch)
 		 * While we are waiting for remote hart to set the sync,
 		 * consume fifo requests to avoid deadlock.
 		 */
-		sbi_tlb_process_count(scratch, 1);
+		tlb_process_count(scratch, 1);
 	}
 
 	return;
 }
 
-static inline int __sbi_tlb_range_check(struct sbi_tlb_info *curr,
+static inline int tlb_range_check(struct sbi_tlb_info *curr,
 					struct sbi_tlb_info *next)
 {
 	unsigned long curr_end;
@@ -278,7 +314,7 @@ static inline int __sbi_tlb_range_check(struct sbi_tlb_info *curr,
  *	before continuing the while loop. This method is preferred over wfi/ipi because
  *	of MMIO cost involved in later method.
  */
-static int sbi_tlb_update_cb(void *in, void *data)
+static int tlb_update_cb(void *in, void *data)
 {
 	struct sbi_tlb_info *curr;
 	struct sbi_tlb_info *next;
@@ -293,16 +329,16 @@ static int sbi_tlb_update_cb(void *in, void *data)
 	if (next->local_fn == sbi_tlb_local_sfence_vma_asid &&
 	    curr->local_fn == sbi_tlb_local_sfence_vma_asid) {
 		if (next->asid == curr->asid)
-			ret = __sbi_tlb_range_check(curr, next);
+			ret = tlb_range_check(curr, next);
 	} else if (next->local_fn == sbi_tlb_local_sfence_vma &&
 		   curr->local_fn == sbi_tlb_local_sfence_vma) {
-		ret = __sbi_tlb_range_check(curr, next);
+		ret = tlb_range_check(curr, next);
 	}
 
 	return ret;
 }
 
-static int sbi_tlb_update(struct sbi_scratch *scratch,
+static int tlb_update(struct sbi_scratch *scratch,
 			  struct sbi_scratch *remote_scratch,
 			  u32 remote_hartid, void *data)
 {
@@ -332,7 +368,7 @@ static int sbi_tlb_update(struct sbi_scratch *scratch,
 
 	tlb_fifo_r = sbi_scratch_offset_ptr(remote_scratch, tlb_fifo_off);
 
-	ret = sbi_fifo_inplace_update(tlb_fifo_r, data, sbi_tlb_update_cb);
+	ret = sbi_fifo_inplace_update(tlb_fifo_r, data, tlb_update_cb);
 	if (ret != SBI_FIFO_UNCHANGED) {
 		return 1;
 	}
@@ -346,7 +382,7 @@ static int sbi_tlb_update(struct sbi_scratch *scratch,
 		 * TODO: Introduce a wait/wakeup event mechanism to handle
 		 * this properly.
 		 */
-		sbi_tlb_process_count(scratch, 1);
+		tlb_process_count(scratch, 1);
 		sbi_dprintf("hart%d: hart%d tlb fifo full\n",
 			    curr_hartid, remote_hartid);
 	}
@@ -356,9 +392,9 @@ static int sbi_tlb_update(struct sbi_scratch *scratch,
 
 static struct sbi_ipi_event_ops tlb_ops = {
 	.name = "IPI_TLB",
-	.update = sbi_tlb_update,
-	.sync = sbi_tlb_sync,
-	.process = sbi_tlb_process,
+	.update = tlb_update,
+	.sync = tlb_sync,
+	.process = tlb_process,
 };
 
 static u32 tlb_event = SBI_IPI_EVENT_MAX;
@@ -368,6 +404,8 @@ int sbi_tlb_request(ulong hmask, ulong hbase, struct sbi_tlb_info *tinfo)
 	if (!tinfo->local_fn)
 		return SBI_EINVAL;
 
+	tlb_pmu_incr_fw_ctr(tinfo);
+
 	return sbi_ipi_send_many(hmask, hbase, tlb_event, tinfo);
 }
 
@@ -380,19 +418,16 @@ int sbi_tlb_init(struct sbi_scratch *scratch, bool cold_boot)
 	const struct sbi_platform *plat = sbi_platform_ptr(scratch);
 
 	if (cold_boot) {
-		tlb_sync_off = sbi_scratch_alloc_offset(sizeof(*tlb_sync),
-							"IPI_TLB_SYNC");
+		tlb_sync_off = sbi_scratch_alloc_offset(sizeof(*tlb_sync));
 		if (!tlb_sync_off)
 			return SBI_ENOMEM;
-		tlb_fifo_off = sbi_scratch_alloc_offset(sizeof(*tlb_q),
-							"IPI_TLB_FIFO");
+		tlb_fifo_off = sbi_scratch_alloc_offset(sizeof(*tlb_q));
 		if (!tlb_fifo_off) {
 			sbi_scratch_free_offset(tlb_sync_off);
 			return SBI_ENOMEM;
 		}
 		tlb_fifo_mem_off = sbi_scratch_alloc_offset(
-				SBI_TLB_FIFO_NUM_ENTRIES * SBI_TLB_INFO_SIZE,
-				"IPI_TLB_FIFO_MEM");
+				SBI_TLB_FIFO_NUM_ENTRIES * SBI_TLB_INFO_SIZE);
 		if (!tlb_fifo_mem_off) {
 			sbi_scratch_free_offset(tlb_fifo_off);
 			sbi_scratch_free_offset(tlb_sync_off);

lib/sbi/sbi_trap.c

@@ -16,6 +16,7 @@
 #include <sbi/sbi_illegal_insn.h>
 #include <sbi/sbi_ipi.h>
 #include <sbi/sbi_misaligned_ldst.h>
+#include <sbi/sbi_pmu.h>
 #include <sbi/sbi_scratch.h>
 #include <sbi/sbi_timer.h>
 #include <sbi/sbi_trap.h>
@@ -210,7 +211,7 @@ int sbi_trap_redirect(struct sbi_trap_regs *regs,
  *
  * @param regs pointer to register state
  */
-void sbi_trap_handler(struct sbi_trap_regs *regs)
+struct sbi_trap_regs *sbi_trap_handler(struct sbi_trap_regs *regs)
 {
 	int rc = SBI_ENOTSUPP;
 	const char *msg = "trap handler failed";
@@ -236,7 +237,7 @@ void sbi_trap_handler(struct sbi_trap_regs *regs)
 			msg = "unhandled external interrupt";
 			goto trap_error;
 		};
-		return;
+		return regs;
 	}
 
 	switch (mcause) {
@@ -257,6 +258,11 @@ void sbi_trap_handler(struct sbi_trap_regs *regs)
 		rc  = sbi_ecall_handler(regs);
 		msg = "ecall handler failed";
 		break;
+	case CAUSE_LOAD_ACCESS:
+	case CAUSE_STORE_ACCESS:
+		sbi_pmu_ctr_incr_fw(mcause == CAUSE_LOAD_ACCESS ?
+			SBI_PMU_FW_ACCESS_LOAD : SBI_PMU_FW_ACCESS_STORE);
+		/* fallthrough */
 	default:
 		/* If the trap came from S or U mode, redirect it there */
 		trap.epc = regs->mepc;
@@ -271,6 +277,7 @@ void sbi_trap_handler(struct sbi_trap_regs *regs)
 trap_error:
 	if (rc)
 		sbi_trap_error(msg, rc, mcause, mtval, mtval2, mtinst, regs);
+	return regs;
 }
 
 typedef void (*trap_exit_t)(const struct sbi_trap_regs *regs);

lib/utils/fdt/fdt_domain.c

@@ -9,6 +9,7 @@
  */
 
 #include <libfdt.h>
+#include <libfdt_env.h>
 #include <sbi/sbi_domain.h>
 #include <sbi/sbi_error.h>
 #include <sbi/sbi_hartmask.h>
@@ -95,7 +96,7 @@ static int __fixup_find_domain_offset(void *fdt, int doff, void *p)
 {
 	struct __fixup_find_domain_offset_info *fdo = p;
 
-	if (!sbi_strcmp(fdo->name, fdt_get_name(fdt, doff, NULL)))
+	if (!strncmp(fdo->name, fdt_get_name(fdt, doff, NULL), strlen(fdo->name)))
 		*fdo->doffset = doff;
 
 	return 0;
@@ -222,7 +223,7 @@ static u32 fdt_domains_count;
 static struct sbi_domain fdt_domains[FDT_DOMAIN_MAX_COUNT];
 static struct sbi_hartmask fdt_masks[FDT_DOMAIN_MAX_COUNT];
 static struct sbi_domain_memregion
-	fdt_regions[FDT_DOMAIN_MAX_COUNT][FDT_DOMAIN_REGION_MAX_COUNT + 2];
+	fdt_regions[FDT_DOMAIN_MAX_COUNT][FDT_DOMAIN_REGION_MAX_COUNT + 1];
 
 static int __fdt_parse_region(void *fdt, int domain_offset,
 			      int region_offset, u32 region_access,
@@ -276,7 +277,7 @@ static int __fdt_parse_domain(void *fdt, int domain_offset, void *opaque)
 	struct sbi_hartmask *mask;
 	struct sbi_hartmask assign_mask;
 	int *cold_domain_offset = opaque;
-	struct sbi_domain_memregion *regions;
+	struct sbi_domain_memregion *reg, *regions;
 	int i, err, len, cpus_offset, cpu_offset, doffset;
 
 	/* Sanity check on maximum domains we can handle */
@@ -287,7 +288,7 @@ static int __fdt_parse_domain(void *fdt, int domain_offset, void *opaque)
 	regions = &fdt_regions[fdt_domains_count][0];
 
 	/* Read DT node name */
-	sbi_strncpy(dom->name, fdt_get_name(fdt, domain_offset, NULL),
+	strncpy(dom->name, fdt_get_name(fdt, domain_offset, NULL),
 		    sizeof(dom->name));
 	dom->name[sizeof(dom->name) - 1] = '\0';
 
@@ -313,14 +314,32 @@ static int __fdt_parse_domain(void *fdt, int domain_offset, void *opaque)
 
 	/* Setup memregions from DT */
 	val32 = 0;
-	sbi_memset(regions, 0,
-		   sizeof(*regions) * (FDT_DOMAIN_REGION_MAX_COUNT + 2));
+	memset(regions, 0,
+		   sizeof(*regions) * (FDT_DOMAIN_REGION_MAX_COUNT + 1));
 	dom->regions = regions;
 	err = fdt_iterate_each_memregion(fdt, domain_offset, &val32,
 					 __fdt_parse_region);
 	if (err)
 		return err;
-	sbi_domain_memregion_initfw(&regions[val32]);
+
+	/*
+	 * Copy over root domain memregions which don't allow
+	 * read, write and execute from lower privilege modes.
+	 *
+	 * These root domain memregions without read, write,
+	 * and execute permissions include:
+	 * 1) firmware region protecting the firmware memory
+	 * 2) mmio regions protecting M-mode only mmio devices
+	 */
+	sbi_domain_for_each_memregion(&root, reg) {
+		if ((reg->flags & SBI_DOMAIN_MEMREGION_READABLE) ||
+		    (reg->flags & SBI_DOMAIN_MEMREGION_WRITEABLE) ||
+		    (reg->flags & SBI_DOMAIN_MEMREGION_EXECUTABLE))
+			continue;
+		if (FDT_DOMAIN_REGION_MAX_COUNT <= val32)
+			return SBI_EINVAL;
+		memcpy(&regions[val32++], reg, sizeof(*reg));
+	}
 
 	/* Read "boot-hart" DT property */
 	val32 = -1U;

lib/utils/fdt/fdt_fixup.c

@@ -15,6 +15,7 @@
 #include <sbi/sbi_scratch.h>
 #include <sbi/sbi_string.h>
 #include <sbi_utils/fdt/fdt_fixup.h>
+#include <sbi_utils/fdt/fdt_pmu.h>
 #include <sbi_utils/fdt/fdt_helper.h>
 
 void fdt_cpu_fixup(void *fdt)
@@ -51,15 +52,18 @@ void fdt_cpu_fixup(void *fdt)
 	}
 }
 
-void fdt_plic_fixup(void *fdt, const char *compat)
+void fdt_plic_fixup(void *fdt)
 {
 	u32 *cells;
 	int i, cells_count;
 	int plic_off;
 
-	plic_off = fdt_node_offset_by_compatible(fdt, 0, compat);
+	plic_off = fdt_node_offset_by_compatible(fdt, 0, "sifive,plic-1.0.0");
+	if (plic_off < 0) {
+		plic_off = fdt_node_offset_by_compatible(fdt, 0, "riscv,plic0");
 		if (plic_off < 0)
 			return;
+	}
 
 	cells = (u32 *)fdt_getprop(fdt, plic_off,
 				   "interrupts-extended", &cells_count);
@@ -257,9 +261,10 @@ int fdt_reserved_memory_nomap_fixup(void *fdt)
 
 void fdt_fixups(void *fdt)
 {
-	fdt_plic_fixup(fdt, "riscv,plic0");
+	fdt_plic_fixup(fdt);
 
 	fdt_reserved_memory_fixup(fdt);
+	fdt_pmu_fixup(fdt);
 }