RISC-V Patches for the 6.9 Merge Window

* Support for various vector-accelerated crypto routines.
 * Hibernation is now enabled for portable kernel builds.
 * mmap_rnd_bits_max is larger on systems with larger VAs.
 * Support for fast GUP.
 * Support for membarrier-based instruction cache synchronization.
 * Support for the Andes hart-level interrupt controller and PMU.
 * Some cleanups around unaligned access speed probing and Kconfig
   settings.
 * Support for ACPI LPI and CPPC.
 * Various cleanus related to barriers.
 * A handful of fixes.
 -----BEGIN PGP SIGNATURE-----
 
 iQJHBAABCAAxFiEEKzw3R0RoQ7JKlDp6LhMZ81+7GIkFAmX9icgTHHBhbG1lckBk
 YWJiZWx0LmNvbQAKCRAuExnzX7sYib+UD/4xyL6UMixx6A06BVBL9UT4vOrxRvNr
 JIihG5y5QNMjes9DHWL35mZTMqFtQ0tq94ViWFLmJWloV/8KRVM2C9R9KX7vplf3
 M/OwvP106spxgvNHoeQbycgs42RU1t2mpqT7N1iK2hCjqieP3vLn6hsSLXWTAG0L
 3gQbQw6XCLC3hPyLq+nbFY2i4faeCmpXWmixoy/IvQ5calZQrRU0LNlP6lcMBhVo
 uocjG0uGAhrahw2s81jxcMZcxa3AvUCiplapdD5H5v9rBM85SkYJj2Q9SqdSorkb
 xzuimRnKPI5s47yM3pTfZY0qnQUYHV7PXXuw4WujpCQVQdhaG+Ggq63UUZA61J9t
 IzZK2zdcfHqICrGTtXImUzRT3dcc3oq+IFq4tTY+rEJm29hrXkAtx+qBm5xtMvax
 fJz5feJ/iT0u7MDj4Oq24n+Kpl+Olm+MJaZX3m5Ovi/9V6a9iK9HXqxg9/Fs0fMO
 +J/0kTgd8Vu9CYH7KNWz3uztcO9eMAH3VyzuXuab4BGj1i1Y/9EjpALQi7rDN73S
 OsYQX6NnzMkBV4dvElJVLXiPlvNlMHZZwdak5CqPb48jaJu6iiIZAuvOrG6/naGP
 wnQSLVA2WWWoOkl3AJhxfpa11CLhbMl9E2gYm1VtNvASXoSFIxlAq1Yv3sG8yjty
 4ZT0rYFJOstYiQ==
 =3dL5
 -----END PGP SIGNATURE-----

Merge tag 'riscv-for-linus-6.9-mw2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux

Pull RISC-V updates from Palmer Dabbelt:

 - Support for various vector-accelerated crypto routines

 - Hibernation is now enabled for portable kernel builds

 - mmap_rnd_bits_max is larger on systems with larger VAs

 - Support for fast GUP

 - Support for membarrier-based instruction cache synchronization

 - Support for the Andes hart-level interrupt controller and PMU

 - Some cleanups around unaligned access speed probing and Kconfig
   settings

 - Support for ACPI LPI and CPPC

 - Various cleanus related to barriers

 - A handful of fixes

* tag 'riscv-for-linus-6.9-mw2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: (66 commits)
  riscv: Fix syscall wrapper for >word-size arguments
  crypto: riscv - add vector crypto accelerated AES-CBC-CTS
  crypto: riscv - parallelize AES-CBC decryption
  riscv: Only flush the mm icache when setting an exec pte
  riscv: Use kcalloc() instead of kzalloc()
  riscv/barrier: Add missing space after ','
  riscv/barrier: Consolidate fence definitions
  riscv/barrier: Define RISCV_FULL_BARRIER
  riscv/barrier: Define __{mb,rmb,wmb}
  RISC-V: defconfig: Enable CONFIG_ACPI_CPPC_CPUFREQ
  cpufreq: Move CPPC configs to common Kconfig and add RISC-V
  ACPI: RISC-V: Add CPPC driver
  ACPI: Enable ACPI_PROCESSOR for RISC-V
  ACPI: RISC-V: Add LPI driver
  cpuidle: RISC-V: Move few functions to arch/riscv
  riscv: Introduce set_compat_task() in asm/compat.h
  riscv: Introduce is_compat_thread() into compat.h
  riscv: add compile-time test into is_compat_task()
  riscv: Replace direct thread flag check with is_compat_task()
  riscv: Improve arch_get_mmap_end() macro
  ...

Documentation/arch/riscv/vm-layout.rst

@@ -144,14 +144,8 @@ passing 0 into the hint address parameter of mmap. On CPUs with an address space
 smaller than sv48, the CPU maximum supported address space will be the default.
 
 Software can "opt-in" to receiving VAs from another VA space by providing
-a hint address to mmap. A hint address passed to mmap will cause the largest
-address space that fits entirely into the hint to be used, unless there is no
-space left in the address space. If there is no space available in the requested
-address space, an address in the next smallest available address space will be
-returned.
-
-For example, in order to obtain 48-bit VA space, a hint address greater than
-:code:`1 << 47` must be provided. Note that this is 47 due to sv48 userspace
-ending at :code:`1 << 47` and the addresses beyond this are reserved for the
-kernel. Similarly, to obtain 57-bit VA space addresses, a hint address greater
-than or equal to :code:`1 << 56` must be provided.
+a hint address to mmap. When a hint address is passed to mmap, the returned
+address will never use more bits than the hint address. For example, if a hint
+address of `1 << 40` is passed to mmap, a valid returned address will never use
+bits 41 through 63. If no mappable addresses are available in that range, mmap
+will return `MAP_FAILED`.

Documentation/devicetree/bindings/riscv/cpus.yaml

@@ -110,7 +110,11 @@ properties:
         const: 1
 
       compatible:
-        const: riscv,cpu-intc
+        oneOf:
+          - items:
+              - const: andestech,cpu-intc
+              - const: riscv,cpu-intc
+          - const: riscv,cpu-intc
 
       interrupt-controller: true
 

Documentation/devicetree/bindings/riscv/extensions.yaml

@@ -477,5 +477,12 @@ properties:
             latency, as ratified in commit 56ed795 ("Update
             riscv-crypto-spec-vector.adoc") of riscv-crypto.
 
+        - const: xandespmu
+          description:
+            The Andes Technology performance monitor extension for counter overflow
+            and privilege mode filtering. For more details, see Counter Related
+            Registers in the AX45MP datasheet.
+            https://www.andestech.com/wp-content/uploads/AX45MP-1C-Rev.-5.0.0-Datasheet.pdf
+
 additionalProperties: true
 ...

Documentation/features/sched/membarrier-sync-core/arch-support.txt

@@ -10,6 +10,22 @@
 # Rely on implicit context synchronization as a result of exception return
 # when returning from IPI handler, and when returning to user-space.
 #
+# * riscv
+#
+# riscv uses xRET as return from interrupt and to return to user-space.
+#
+# Given that xRET is not core serializing, we rely on FENCE.I for providing
+# core serialization:
+#
+#  - by calling sync_core_before_usermode() on return from interrupt (cf.
+#    ipi_sync_core()),
+#
+#  - via switch_mm() and sync_core_before_usermode() (respectively, for
+#    uthread->uthread and kthread->uthread transitions) before returning
+#    to user-space.
+#
+#  The serialization in switch_mm() is activated by prepare_sync_core_cmd().
+#
 # * x86
 #
 # x86-32 uses IRET as return from interrupt, which takes care of the IPI.
@@ -43,7 +59,7 @@
     |    openrisc: | TODO |
     |      parisc: | TODO |
     |     powerpc: |  ok  |
-    |       riscv: | TODO |
+    |       riscv: |  ok  |
     |        s390: |  ok  |
     |          sh: | TODO |
     |       sparc: | TODO |

Documentation/scheduler/index.rst

@@ -7,6 +7,7 @@ Scheduler
 
 
     completion
+    membarrier
     sched-arch
     sched-bwc
     sched-deadline

Documentation/scheduler/membarrier.rst

@@ -0,0 +1,39 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+========================
+membarrier() System Call
+========================
+
+MEMBARRIER_CMD_{PRIVATE,GLOBAL}_EXPEDITED - Architecture requirements
+=====================================================================
+
+Memory barriers before updating rq->curr
+----------------------------------------
+
+The commands MEMBARRIER_CMD_PRIVATE_EXPEDITED and MEMBARRIER_CMD_GLOBAL_EXPEDITED
+require each architecture to have a full memory barrier after coming from
+user-space, before updating rq->curr.  This barrier is implied by the sequence
+rq_lock(); smp_mb__after_spinlock() in __schedule().  The barrier matches a full
+barrier in the proximity of the membarrier system call exit, cf.
+membarrier_{private,global}_expedited().
+
+Memory barriers after updating rq->curr
+---------------------------------------
+
+The commands MEMBARRIER_CMD_PRIVATE_EXPEDITED and MEMBARRIER_CMD_GLOBAL_EXPEDITED
+require each architecture to have a full memory barrier after updating rq->curr,
+before returning to user-space.  The schemes providing this barrier on the various
+architectures are as follows.
+
+ - alpha, arc, arm, hexagon, mips rely on the full barrier implied by
+   spin_unlock() in finish_lock_switch().
+
+ - arm64 relies on the full barrier implied by switch_to().
+
+ - powerpc, riscv, s390, sparc, x86 rely on the full barrier implied by
+   switch_mm(), if mm is not NULL; they rely on the full barrier implied
+   by mmdrop(), otherwise.  On powerpc and riscv, switch_mm() relies on
+   membarrier_arch_switch_mm().
+
+The barrier matches a full barrier in the proximity of the membarrier system call
+entry, cf. membarrier_{private,global}_expedited().

MAINTAINERS

@@ -14134,7 +14134,9 @@ M:	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 M:	"Paul E. McKenney" <paulmck@kernel.org>
 L:	linux-kernel@vger.kernel.org
 S:	Supported
-F:	arch/powerpc/include/asm/membarrier.h
+F:	Documentation/scheduler/membarrier.rst
+F:	arch/*/include/asm/membarrier.h
+F:	arch/*/include/asm/sync_core.h
 F:	include/uapi/linux/membarrier.h
 F:	kernel/sched/membarrier.c
 

arch/riscv/Kbuild

@@ -2,6 +2,7 @@
 
 obj-y += kernel/ mm/ net/
 obj-$(CONFIG_BUILTIN_DTB) += boot/dts/
+obj-$(CONFIG_CRYPTO) += crypto/
 obj-y += errata/
 obj-$(CONFIG_KVM) += kvm/
 

arch/riscv/Kconfig

@@ -27,14 +27,18 @@ config RISCV
 	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_HAS_GIGANTIC_PAGE
 	select ARCH_HAS_KCOV
+	select ARCH_HAS_MEMBARRIER_CALLBACKS
+	select ARCH_HAS_MEMBARRIER_SYNC_CORE
 	select ARCH_HAS_MMIOWB
 	select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	select ARCH_HAS_PMEM_API
+	select ARCH_HAS_PREPARE_SYNC_CORE_CMD
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_SET_DIRECT_MAP if MMU
 	select ARCH_HAS_SET_MEMORY if MMU
 	select ARCH_HAS_STRICT_KERNEL_RWX if MMU && !XIP_KERNEL
 	select ARCH_HAS_STRICT_MODULE_RWX if MMU && !XIP_KERNEL
+	select ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
 	select ARCH_HAS_SYSCALL_WRAPPER
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HAS_UBSAN
@@ -47,6 +51,9 @@ config RISCV
 	select ARCH_SUPPORTS_CFI_CLANG
 	select ARCH_SUPPORTS_DEBUG_PAGEALLOC if MMU
 	select ARCH_SUPPORTS_HUGETLBFS if MMU
+	# LLD >= 14: https://github.com/llvm/llvm-project/issues/50505
+	select ARCH_SUPPORTS_LTO_CLANG if LLD_VERSION >= 140000
+	select ARCH_SUPPORTS_LTO_CLANG_THIN if LLD_VERSION >= 140000
 	select ARCH_SUPPORTS_PAGE_TABLE_CHECK if MMU
 	select ARCH_SUPPORTS_PER_VMA_LOCK if MMU
 	select ARCH_SUPPORTS_SHADOW_CALL_STACK if HAVE_SHADOW_CALL_STACK
@@ -106,6 +113,7 @@ config RISCV
 	select HAVE_ARCH_KGDB_QXFER_PKT
 	select HAVE_ARCH_MMAP_RND_BITS if MMU
 	select HAVE_ARCH_MMAP_RND_COMPAT_BITS if COMPAT
+	select HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
 	select HAVE_ARCH_SECCOMP_FILTER
 	select HAVE_ARCH_THREAD_STRUCT_WHITELIST
 	select HAVE_ARCH_TRACEHOOK
@@ -124,6 +132,7 @@ config RISCV
 	select HAVE_FUNCTION_GRAPH_RETVAL if HAVE_FUNCTION_GRAPH_TRACER
 	select HAVE_FUNCTION_TRACER if !XIP_KERNEL && !PREEMPTION
 	select HAVE_EBPF_JIT if MMU
+	select HAVE_FAST_GUP if MMU
 	select HAVE_FUNCTION_ARG_ACCESS_API
 	select HAVE_FUNCTION_ERROR_INJECTION
 	select HAVE_GCC_PLUGINS
@@ -155,6 +164,7 @@ config RISCV
 	select IRQ_FORCED_THREADING
 	select KASAN_VMALLOC if KASAN
 	select LOCK_MM_AND_FIND_VMA
+	select MMU_GATHER_RCU_TABLE_FREE if SMP && MMU
 	select MODULES_USE_ELF_RELA if MODULES
 	select MODULE_SECTIONS if MODULES
 	select OF
@@ -576,6 +586,13 @@ config TOOLCHAIN_HAS_ZBB
 	depends on LLD_VERSION >= 150000 || LD_VERSION >= 23900
 	depends on AS_HAS_OPTION_ARCH
 
+# This symbol indicates that the toolchain supports all v1.0 vector crypto
+# extensions, including Zvk*, Zvbb, and Zvbc.  LLVM added all of these at once.
+# binutils added all except Zvkb, then added Zvkb.  So we just check for Zvkb.
+config TOOLCHAIN_HAS_VECTOR_CRYPTO
+	def_bool $(as-instr, .option arch$(comma) +v$(comma) +zvkb)
+	depends on AS_HAS_OPTION_ARCH
+
 config RISCV_ISA_ZBB
 	bool "Zbb extension support for bit manipulation instructions"
 	depends on TOOLCHAIN_HAS_ZBB
@@ -686,27 +703,61 @@ config THREAD_SIZE_ORDER
 	  affects irq stack size, which is equal to thread stack size.
 
 config RISCV_MISALIGNED
-	bool "Support misaligned load/store traps for kernel and userspace"
+	bool
 	select SYSCTL_ARCH_UNALIGN_ALLOW
-	default y
 	help
-	  Say Y here if you want the kernel to embed support for misaligned
-	  load/store for both kernel and userspace. When disable, misaligned
-	  accesses will generate SIGBUS in userspace and panic in kernel.
+	  Embed support for emulating misaligned loads and stores.
+
+choice
+	prompt "Unaligned Accesses Support"
+	default RISCV_PROBE_UNALIGNED_ACCESS
+	help
+	  This determines the level of support for unaligned accesses. This
+	  information is used by the kernel to perform optimizations. It is also
+	  exposed to user space via the hwprobe syscall. The hardware will be
+	  probed at boot by default.
+
+config RISCV_PROBE_UNALIGNED_ACCESS
+	bool "Probe for hardware unaligned access support"
+	select RISCV_MISALIGNED
+	help
+	  During boot, the kernel will run a series of tests to determine the
+	  speed of unaligned accesses. This probing will dynamically determine
+	  the speed of unaligned accesses on the underlying system. If unaligned
+	  memory accesses trap into the kernel as they are not supported by the
+	  system, the kernel will emulate the unaligned accesses to preserve the
+	  UABI.
+
+config RISCV_EMULATED_UNALIGNED_ACCESS
+	bool "Emulate unaligned access where system support is missing"
+	select RISCV_MISALIGNED
+	help
+	  If unaligned memory accesses trap into the kernel as they are not
+	  supported by the system, the kernel will emulate the unaligned
+	  accesses to preserve the UABI. When the underlying system does support
+	  unaligned accesses, the unaligned accesses are assumed to be slow.
+
+config RISCV_SLOW_UNALIGNED_ACCESS
+	bool "Assume the system supports slow unaligned memory accesses"
+	depends on NONPORTABLE
+	help
+	  Assume that the system supports slow unaligned memory accesses. The
+	  kernel and userspace programs may not be able to run at all on systems
+	  that do not support unaligned memory accesses.
 
 config RISCV_EFFICIENT_UNALIGNED_ACCESS
-	bool "Assume the CPU supports fast unaligned memory accesses"
+	bool "Assume the system supports fast unaligned memory accesses"
 	depends on NONPORTABLE
 	select DCACHE_WORD_ACCESS if MMU
 	select HAVE_EFFICIENT_UNALIGNED_ACCESS
 	help
-	  Say Y here if you want the kernel to assume that the CPU supports
-	  efficient unaligned memory accesses.  When enabled, this option
-	  improves the performance of the kernel on such CPUs.  However, the
-	  kernel will run much more slowly, or will not be able to run at all,
-	  on CPUs that do not support efficient unaligned memory accesses.
+	  Assume that the system supports fast unaligned memory accesses. When
+	  enabled, this option improves the performance of the kernel on such
+	  systems. However, the kernel and userspace programs will run much more
+	  slowly, or will not be able to run at all, on systems that do not
+	  support efficient unaligned memory accesses.
 
-	  If unsure what to do here, say N.
+endchoice
 
 endmenu # "Platform type"
 
@@ -1011,11 +1062,8 @@ menu "Power management options"
 
 source "kernel/power/Kconfig"
 
-# Hibernation is only possible on systems where the SBI implementation has
-# marked its reserved memory as not accessible from, or does not run
-# from the same memory as, Linux
 config ARCH_HIBERNATION_POSSIBLE
-	def_bool NONPORTABLE
+	def_bool y
 
 config ARCH_HIBERNATION_HEADER
 	def_bool HIBERNATION

arch/riscv/Makefile

@@ -50,6 +50,11 @@ ifndef CONFIG_AS_IS_LLVM
 	KBUILD_CFLAGS += -Wa,-mno-relax
 	KBUILD_AFLAGS += -Wa,-mno-relax
 endif
+# LLVM has an issue with target-features and LTO: https://github.com/llvm/llvm-project/issues/59350
+# Ensure it is aware of linker relaxation with LTO, otherwise relocations may
+# be incorrect: https://github.com/llvm/llvm-project/issues/65090
+else ifeq ($(CONFIG_LTO_CLANG),y)
+	KBUILD_LDFLAGS += -mllvm -mattr=+c -mllvm -mattr=+relax
 endif
 
 ifeq ($(CONFIG_SHADOW_CALL_STACK),y)

arch/riscv/boot/dts/renesas/r9a07g043f.dtsi

@@ -27,7 +27,7 @@ cpu0: cpu@0 {
 			riscv,isa-base = "rv64i";
 			riscv,isa-extensions = "i", "m", "a", "f", "d", "c",
 					       "zicntr", "zicsr", "zifencei",
-					       "zihpm";
+					       "zihpm", "xandespmu";
 			mmu-type = "riscv,sv39";
 			i-cache-size = <0x8000>;
 			i-cache-line-size = <0x40>;
@@ -39,7 +39,7 @@ cpu0: cpu@0 {
 
 			cpu0_intc: interrupt-controller {
 				#interrupt-cells = <1>;
-				compatible = "riscv,cpu-intc";
+				compatible = "andestech,cpu-intc", "riscv,cpu-intc";
 				interrupt-controller;
 			};
 		};

arch/riscv/configs/defconfig

@@ -44,6 +44,7 @@ CONFIG_CPU_FREQ_GOV_USERSPACE=y
 CONFIG_CPU_FREQ_GOV_ONDEMAND=y
 CONFIG_CPU_FREQ_GOV_CONSERVATIVE=m
 CONFIG_CPUFREQ_DT=y
+CONFIG_ACPI_CPPC_CPUFREQ=m
 CONFIG_VIRTUALIZATION=y
 CONFIG_KVM=m
 CONFIG_ACPI=y
@@ -215,6 +216,7 @@ CONFIG_MMC=y
 CONFIG_MMC_SDHCI=y
 CONFIG_MMC_SDHCI_PLTFM=y
 CONFIG_MMC_SDHCI_CADENCE=y
+CONFIG_MMC_SDHCI_OF_DWCMSHC=y
 CONFIG_MMC_SPI=y
 CONFIG_MMC_DW=y
 CONFIG_MMC_DW_STARFIVE=y
@@ -224,6 +226,7 @@ CONFIG_RTC_CLASS=y
 CONFIG_RTC_DRV_SUN6I=y
 CONFIG_DMADEVICES=y
 CONFIG_DMA_SUN6I=m
+CONFIG_DW_AXI_DMAC=y
 CONFIG_RZ_DMAC=y
 CONFIG_VIRTIO_PCI=y
 CONFIG_VIRTIO_BALLOON=y

arch/riscv/crypto/Kconfig

@@ -0,0 +1,93 @@
+# SPDX-License-Identifier: GPL-2.0
+
+menu "Accelerated Cryptographic Algorithms for CPU (riscv)"
+
+config CRYPTO_AES_RISCV64
+	tristate "Ciphers: AES, modes: ECB, CBC, CTS, CTR, XTS"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_ALGAPI
+	select CRYPTO_LIB_AES
+	select CRYPTO_SKCIPHER
+	help
+	  Block cipher: AES cipher algorithms
+	  Length-preserving ciphers: AES with ECB, CBC, CTS, CTR, XTS
+
+	  Architecture: riscv64 using:
+	  - Zvkned vector crypto extension
+	  - Zvbb vector extension (XTS)
+	  - Zvkb vector crypto extension (CTR)
+	  - Zvkg vector crypto extension (XTS)
+
+config CRYPTO_CHACHA_RISCV64
+	tristate "Ciphers: ChaCha"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_SKCIPHER
+	select CRYPTO_LIB_CHACHA_GENERIC
+	help
+	  Length-preserving ciphers: ChaCha20 stream cipher algorithm
+
+	  Architecture: riscv64 using:
+	  - Zvkb vector crypto extension
+
+config CRYPTO_GHASH_RISCV64
+	tristate "Hash functions: GHASH"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_GCM
+	help
+	  GCM GHASH function (NIST SP 800-38D)
+
+	  Architecture: riscv64 using:
+	  - Zvkg vector crypto extension
+
+config CRYPTO_SHA256_RISCV64
+	tristate "Hash functions: SHA-224 and SHA-256"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_SHA256
+	help
+	  SHA-224 and SHA-256 secure hash algorithm (FIPS 180)
+
+	  Architecture: riscv64 using:
+	  - Zvknha or Zvknhb vector crypto extensions
+	  - Zvkb vector crypto extension
+
+config CRYPTO_SHA512_RISCV64
+	tristate "Hash functions: SHA-384 and SHA-512"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_SHA512
+	help
+	  SHA-384 and SHA-512 secure hash algorithm (FIPS 180)
+
+	  Architecture: riscv64 using:
+	  - Zvknhb vector crypto extension
+	  - Zvkb vector crypto extension
+
+config CRYPTO_SM3_RISCV64
+	tristate "Hash functions: SM3 (ShangMi 3)"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_HASH
+	select CRYPTO_SM3
+	help
+	  SM3 (ShangMi 3) secure hash function (OSCCA GM/T 0004-2012)
+
+	  Architecture: riscv64 using:
+	  - Zvksh vector crypto extension
+	  - Zvkb vector crypto extension
+
+config CRYPTO_SM4_RISCV64
+	tristate "Ciphers: SM4 (ShangMi 4)"
+	depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+	select CRYPTO_ALGAPI
+	select CRYPTO_SM4
+	help
+	  SM4 block cipher algorithm (OSCCA GB/T 32907-2016,
+	  ISO/IEC 18033-3:2010/Amd 1:2021)
+
+	  SM4 (GBT.32907-2016) is a cryptographic standard issued by the
+	  Organization of State Commercial Administration of China (OSCCA)
+	  as an authorized cryptographic algorithm for use within China.
+
+	  Architecture: riscv64 using:
+	  - Zvksed vector crypto extension
+	  - Zvkb vector crypto extension
+
+endmenu

arch/riscv/crypto/Makefile

@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-$(CONFIG_CRYPTO_AES_RISCV64) += aes-riscv64.o
+aes-riscv64-y := aes-riscv64-glue.o aes-riscv64-zvkned.o \
+		 aes-riscv64-zvkned-zvbb-zvkg.o aes-riscv64-zvkned-zvkb.o
+
+obj-$(CONFIG_CRYPTO_CHACHA_RISCV64) += chacha-riscv64.o
+chacha-riscv64-y := chacha-riscv64-glue.o chacha-riscv64-zvkb.o
+
+obj-$(CONFIG_CRYPTO_GHASH_RISCV64) += ghash-riscv64.o
+ghash-riscv64-y := ghash-riscv64-glue.o ghash-riscv64-zvkg.o
+
+obj-$(CONFIG_CRYPTO_SHA256_RISCV64) += sha256-riscv64.o
+sha256-riscv64-y := sha256-riscv64-glue.o sha256-riscv64-zvknha_or_zvknhb-zvkb.o
+
+obj-$(CONFIG_CRYPTO_SHA512_RISCV64) += sha512-riscv64.o
+sha512-riscv64-y := sha512-riscv64-glue.o sha512-riscv64-zvknhb-zvkb.o
+
+obj-$(CONFIG_CRYPTO_SM3_RISCV64) += sm3-riscv64.o
+sm3-riscv64-y := sm3-riscv64-glue.o sm3-riscv64-zvksh-zvkb.o
+
+obj-$(CONFIG_CRYPTO_SM4_RISCV64) += sm4-riscv64.o
+sm4-riscv64-y := sm4-riscv64-glue.o sm4-riscv64-zvksed-zvkb.o

arch/riscv/crypto/aes-macros.S

@@ -0,0 +1,156 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com>
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// This file contains macros that are shared by the other aes-*.S files.  The
+// generated code of these macros depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector AES block cipher extension ('Zvkned')
+
+// Loads the AES round keys from \keyp into vector registers and jumps to code
+// specific to the length of the key.  Specifically:
+//   - If AES-128, loads round keys into v1-v11 and jumps to \label128.
+//   - If AES-192, loads round keys into v1-v13 and jumps to \label192.
+//   - If AES-256, loads round keys into v1-v15 and continues onwards.
+//
+// Also sets vl=4 and vtype=e32,m1,ta,ma.  Clobbers t0 and t1.
+.macro	aes_begin	keyp, label128, label192
+	lwu		t0, 480(\keyp)	// t0 = key length in bytes
+	li		t1, 24		// t1 = key length for AES-192
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vle32.v		v1, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v2, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v3, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v4, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v5, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v6, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v7, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v8, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v9, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v10, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v11, (\keyp)
+	blt		t0, t1, \label128	// If AES-128, goto label128.
+	addi		\keyp, \keyp, 16
+	vle32.v		v12, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v13, (\keyp)
+	beq		t0, t1, \label192	// If AES-192, goto label192.
+	// Else, it's AES-256.
+	addi		\keyp, \keyp, 16
+	vle32.v		v14, (\keyp)
+	addi		\keyp, \keyp, 16
+	vle32.v		v15, (\keyp)
+.endm
+
+// Encrypts \data using zvkned instructions, using the round keys loaded into
+// v1-v11 (for AES-128), v1-v13 (for AES-192), or v1-v15 (for AES-256).  \keylen
+// is the AES key length in bits.  vl and vtype must already be set
+// appropriately.  Note that if vl > 4, multiple blocks are encrypted.
+.macro	aes_encrypt	data, keylen
+	vaesz.vs	\data, v1
+	vaesem.vs	\data, v2
+	vaesem.vs	\data, v3
+	vaesem.vs	\data, v4
+	vaesem.vs	\data, v5
+	vaesem.vs	\data, v6
+	vaesem.vs	\data, v7
+	vaesem.vs	\data, v8
+	vaesem.vs	\data, v9
+	vaesem.vs	\data, v10
+.if \keylen == 128
+	vaesef.vs	\data, v11
+.elseif \keylen == 192
+	vaesem.vs	\data, v11
+	vaesem.vs	\data, v12
+	vaesef.vs	\data, v13
+.else
+	vaesem.vs	\data, v11
+	vaesem.vs	\data, v12
+	vaesem.vs	\data, v13
+	vaesem.vs	\data, v14
+	vaesef.vs	\data, v15
+.endif
+.endm
+
+// Same as aes_encrypt, but decrypts instead of encrypts.
+.macro	aes_decrypt	data, keylen
+.if \keylen == 128
+	vaesz.vs	\data, v11
+.elseif \keylen == 192
+	vaesz.vs	\data, v13
+	vaesdm.vs	\data, v12
+	vaesdm.vs	\data, v11
+.else
+	vaesz.vs	\data, v15
+	vaesdm.vs	\data, v14
+	vaesdm.vs	\data, v13
+	vaesdm.vs	\data, v12
+	vaesdm.vs	\data, v11
+.endif
+	vaesdm.vs	\data, v10
+	vaesdm.vs	\data, v9
+	vaesdm.vs	\data, v8
+	vaesdm.vs	\data, v7
+	vaesdm.vs	\data, v6
+	vaesdm.vs	\data, v5
+	vaesdm.vs	\data, v4
+	vaesdm.vs	\data, v3
+	vaesdm.vs	\data, v2
+	vaesdf.vs	\data, v1
+.endm
+
+// Expands to aes_encrypt or aes_decrypt according to \enc, which is 1 or 0.
+.macro	aes_crypt	data, enc, keylen
+.if \enc
+	aes_encrypt	\data, \keylen
+.else
+	aes_decrypt	\data, \keylen
+.endif
+.endm

arch/riscv/crypto/aes-riscv64-glue.c

@@ -0,0 +1,637 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * AES using the RISC-V vector crypto extensions.  Includes the bare block
+ * cipher and the ECB, CBC, CBC-CTS, CTR, and XTS modes.
+ *
+ * Copyright (C) 2023 VRULL GmbH
+ * Author: Heiko Stuebner <heiko.stuebner@vrull.eu>
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ *
+ * Copyright 2024 Google LLC
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/aes.h>
+#include <crypto/internal/cipher.h>
+#include <crypto/internal/simd.h>
+#include <crypto/internal/skcipher.h>
+#include <crypto/scatterwalk.h>
+#include <crypto/xts.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+asmlinkage void aes_encrypt_zvkned(const struct crypto_aes_ctx *key,
+				   const u8 in[AES_BLOCK_SIZE],
+				   u8 out[AES_BLOCK_SIZE]);
+asmlinkage void aes_decrypt_zvkned(const struct crypto_aes_ctx *key,
+				   const u8 in[AES_BLOCK_SIZE],
+				   u8 out[AES_BLOCK_SIZE]);
+
+asmlinkage void aes_ecb_encrypt_zvkned(const struct crypto_aes_ctx *key,
+				       const u8 *in, u8 *out, size_t len);
+asmlinkage void aes_ecb_decrypt_zvkned(const struct crypto_aes_ctx *key,
+				       const u8 *in, u8 *out, size_t len);
+
+asmlinkage void aes_cbc_encrypt_zvkned(const struct crypto_aes_ctx *key,
+				       const u8 *in, u8 *out, size_t len,
+				       u8 iv[AES_BLOCK_SIZE]);
+asmlinkage void aes_cbc_decrypt_zvkned(const struct crypto_aes_ctx *key,
+				       const u8 *in, u8 *out, size_t len,
+				       u8 iv[AES_BLOCK_SIZE]);
+
+asmlinkage void aes_cbc_cts_crypt_zvkned(const struct crypto_aes_ctx *key,
+					 const u8 *in, u8 *out, size_t len,
+					 const u8 iv[AES_BLOCK_SIZE], bool enc);
+
+asmlinkage void aes_ctr32_crypt_zvkned_zvkb(const struct crypto_aes_ctx *key,
+					    const u8 *in, u8 *out, size_t len,
+					    u8 iv[AES_BLOCK_SIZE]);
+
+asmlinkage void aes_xts_encrypt_zvkned_zvbb_zvkg(
+			const struct crypto_aes_ctx *key,
+			const u8 *in, u8 *out, size_t len,
+			u8 tweak[AES_BLOCK_SIZE]);
+
+asmlinkage void aes_xts_decrypt_zvkned_zvbb_zvkg(
+			const struct crypto_aes_ctx *key,
+			const u8 *in, u8 *out, size_t len,
+			u8 tweak[AES_BLOCK_SIZE]);
+
+static int riscv64_aes_setkey(struct crypto_aes_ctx *ctx,
+			      const u8 *key, unsigned int keylen)
+{
+	/*
+	 * For now we just use the generic key expansion, for these reasons:
+	 *
+	 * - zvkned's key expansion instructions don't support AES-192.
+	 *   So, non-zvkned fallback code would be needed anyway.
+	 *
+	 * - Users of AES in Linux usually don't change keys frequently.
+	 *   So, key expansion isn't performance-critical.
+	 *
+	 * - For single-block AES exposed as a "cipher" algorithm, it's
+	 *   necessary to use struct crypto_aes_ctx and initialize its 'key_dec'
+	 *   field with the round keys for the Equivalent Inverse Cipher.  This
+	 *   is because with "cipher", decryption can be requested from a
+	 *   context where the vector unit isn't usable, necessitating a
+	 *   fallback to aes_decrypt().  But, zvkned can only generate and use
+	 *   the normal round keys.  Of course, it's preferable to not have
+	 *   special code just for "cipher", as e.g. XTS also uses a
+	 *   single-block AES encryption.  It's simplest to just use
+	 *   struct crypto_aes_ctx and aes_expandkey() everywhere.
+	 */
+	return aes_expandkey(ctx, key, keylen);
+}
+
+static int riscv64_aes_setkey_cipher(struct crypto_tfm *tfm,
+				     const u8 *key, unsigned int keylen)
+{
+	struct crypto_aes_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	return riscv64_aes_setkey(ctx, key, keylen);
+}
+
+static int riscv64_aes_setkey_skcipher(struct crypto_skcipher *tfm,
+				       const u8 *key, unsigned int keylen)
+{
+	struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm);
+
+	return riscv64_aes_setkey(ctx, key, keylen);
+}
+
+/* Bare AES, without a mode of operation */
+
+static void riscv64_aes_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	const struct crypto_aes_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		aes_encrypt_zvkned(ctx, src, dst);
+		kernel_vector_end();
+	} else {
+		aes_encrypt(ctx, dst, src);
+	}
+}
+
+static void riscv64_aes_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	const struct crypto_aes_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		aes_decrypt_zvkned(ctx, src, dst);
+		kernel_vector_end();
+	} else {
+		aes_decrypt(ctx, dst, src);
+	}
+}
+
+/* AES-ECB */
+
+static inline int riscv64_aes_ecb_crypt(struct skcipher_request *req, bool enc)
+{
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct skcipher_walk walk;
+	unsigned int nbytes;
+	int err;
+
+	err = skcipher_walk_virt(&walk, req, false);
+	while ((nbytes = walk.nbytes) != 0) {
+		kernel_vector_begin();
+		if (enc)
+			aes_ecb_encrypt_zvkned(ctx, walk.src.virt.addr,
+					       walk.dst.virt.addr,
+					       nbytes & ~(AES_BLOCK_SIZE - 1));
+		else
+			aes_ecb_decrypt_zvkned(ctx, walk.src.virt.addr,
+					       walk.dst.virt.addr,
+					       nbytes & ~(AES_BLOCK_SIZE - 1));
+		kernel_vector_end();
+		err = skcipher_walk_done(&walk, nbytes & (AES_BLOCK_SIZE - 1));
+	}
+
+	return err;
+}
+
+static int riscv64_aes_ecb_encrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_ecb_crypt(req, true);
+}
+
+static int riscv64_aes_ecb_decrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_ecb_crypt(req, false);
+}
+
+/* AES-CBC */
+
+static int riscv64_aes_cbc_crypt(struct skcipher_request *req, bool enc)
+{
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct skcipher_walk walk;
+	unsigned int nbytes;
+	int err;
+
+	err = skcipher_walk_virt(&walk, req, false);
+	while ((nbytes = walk.nbytes) != 0) {
+		kernel_vector_begin();
+		if (enc)
+			aes_cbc_encrypt_zvkned(ctx, walk.src.virt.addr,
+					       walk.dst.virt.addr,
+					       nbytes & ~(AES_BLOCK_SIZE - 1),
+					       walk.iv);
+		else
+			aes_cbc_decrypt_zvkned(ctx, walk.src.virt.addr,
+					       walk.dst.virt.addr,
+					       nbytes & ~(AES_BLOCK_SIZE - 1),
+					       walk.iv);
+		kernel_vector_end();
+		err = skcipher_walk_done(&walk, nbytes & (AES_BLOCK_SIZE - 1));
+	}
+
+	return err;
+}
+
+static int riscv64_aes_cbc_encrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_cbc_crypt(req, true);
+}
+
+static int riscv64_aes_cbc_decrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_cbc_crypt(req, false);
+}
+
+/* AES-CBC-CTS */
+
+static int riscv64_aes_cbc_cts_crypt(struct skcipher_request *req, bool enc)
+{
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct scatterlist sg_src[2], sg_dst[2];
+	struct skcipher_request subreq;
+	struct scatterlist *src, *dst;
+	struct skcipher_walk walk;
+	unsigned int cbc_len;
+	int err;
+
+	if (req->cryptlen < AES_BLOCK_SIZE)
+		return -EINVAL;
+
+	err = skcipher_walk_virt(&walk, req, false);
+	if (err)
+		return err;
+	/*
+	 * If the full message is available in one step, decrypt it in one call
+	 * to the CBC-CTS assembly function.  This reduces overhead, especially
+	 * on short messages.  Otherwise, fall back to doing CBC up to the last
+	 * two blocks, then invoke CTS just for the ciphertext stealing.
+	 */
+	if (unlikely(walk.nbytes != req->cryptlen)) {
+		cbc_len = round_down(req->cryptlen - AES_BLOCK_SIZE - 1,
+				     AES_BLOCK_SIZE);
+		skcipher_walk_abort(&walk);
+		skcipher_request_set_tfm(&subreq, tfm);
+		skcipher_request_set_callback(&subreq,
+					      skcipher_request_flags(req),
+					      NULL, NULL);
+		skcipher_request_set_crypt(&subreq, req->src, req->dst,
+					   cbc_len, req->iv);
+		err = riscv64_aes_cbc_crypt(&subreq, enc);
+		if (err)
+			return err;
+		dst = src = scatterwalk_ffwd(sg_src, req->src, cbc_len);
+		if (req->dst != req->src)
+			dst = scatterwalk_ffwd(sg_dst, req->dst, cbc_len);
+		skcipher_request_set_crypt(&subreq, src, dst,
+					   req->cryptlen - cbc_len, req->iv);
+		err = skcipher_walk_virt(&walk, &subreq, false);
+		if (err)
+			return err;
+	}
+	kernel_vector_begin();
+	aes_cbc_cts_crypt_zvkned(ctx, walk.src.virt.addr, walk.dst.virt.addr,
+				 walk.nbytes, req->iv, enc);
+	kernel_vector_end();
+	return skcipher_walk_done(&walk, 0);
+}
+
+static int riscv64_aes_cbc_cts_encrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_cbc_cts_crypt(req, true);
+}
+
+static int riscv64_aes_cbc_cts_decrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_cbc_cts_crypt(req, false);
+}
+
+/* AES-CTR */
+
+static int riscv64_aes_ctr_crypt(struct skcipher_request *req)
+{
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm);
+	unsigned int nbytes, p1_nbytes;
+	struct skcipher_walk walk;
+	u32 ctr32, nblocks;
+	int err;
+
+	/* Get the low 32-bit word of the 128-bit big endian counter. */
+	ctr32 = get_unaligned_be32(req->iv + 12);
+
+	err = skcipher_walk_virt(&walk, req, false);
+	while ((nbytes = walk.nbytes) != 0) {
+		if (nbytes < walk.total) {
+			/* Not the end yet, so keep the length block-aligned. */
+			nbytes = round_down(nbytes, AES_BLOCK_SIZE);
+			nblocks = nbytes / AES_BLOCK_SIZE;
+		} else {
+			/* It's the end, so include any final partial block. */
+			nblocks = DIV_ROUND_UP(nbytes, AES_BLOCK_SIZE);
+		}
+		ctr32 += nblocks;
+
+		kernel_vector_begin();
+		if (ctr32 >= nblocks) {
+			/* The low 32-bit word of the counter won't overflow. */
+			aes_ctr32_crypt_zvkned_zvkb(ctx, walk.src.virt.addr,
+						    walk.dst.virt.addr, nbytes,
+						    req->iv);
+		} else {
+			/*
+			 * The low 32-bit word of the counter will overflow.
+			 * The assembly doesn't handle this case, so split the
+			 * operation into two at the point where the overflow
+			 * will occur.  After the first part, add the carry bit.
+			 */
+			p1_nbytes = min_t(unsigned int, nbytes,
+					  (nblocks - ctr32) * AES_BLOCK_SIZE);
+			aes_ctr32_crypt_zvkned_zvkb(ctx, walk.src.virt.addr,
+						    walk.dst.virt.addr,
+						    p1_nbytes, req->iv);
+			crypto_inc(req->iv, 12);
+
+			if (ctr32) {
+				aes_ctr32_crypt_zvkned_zvkb(
+					ctx,
+					walk.src.virt.addr + p1_nbytes,
+					walk.dst.virt.addr + p1_nbytes,
+					nbytes - p1_nbytes, req->iv);
+			}
+		}
+		kernel_vector_end();
+
+		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
+	}
+
+	return err;
+}
+
+/* AES-XTS */
+
+struct riscv64_aes_xts_ctx {
+	struct crypto_aes_ctx ctx1;
+	struct crypto_aes_ctx ctx2;
+};
+
+static int riscv64_aes_xts_setkey(struct crypto_skcipher *tfm, const u8 *key,
+				  unsigned int keylen)
+{
+	struct riscv64_aes_xts_ctx *ctx = crypto_skcipher_ctx(tfm);
+
+	return xts_verify_key(tfm, key, keylen) ?:
+	       riscv64_aes_setkey(&ctx->ctx1, key, keylen / 2) ?:
+	       riscv64_aes_setkey(&ctx->ctx2, key + keylen / 2, keylen / 2);
+}
+
+static int riscv64_aes_xts_crypt(struct skcipher_request *req, bool enc)
+{
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct riscv64_aes_xts_ctx *ctx = crypto_skcipher_ctx(tfm);
+	int tail = req->cryptlen % AES_BLOCK_SIZE;
+	struct scatterlist sg_src[2], sg_dst[2];
+	struct skcipher_request subreq;
+	struct scatterlist *src, *dst;
+	struct skcipher_walk walk;
+	int err;
+
+	if (req->cryptlen < AES_BLOCK_SIZE)
+		return -EINVAL;
+
+	/* Encrypt the IV with the tweak key to get the first tweak. */
+	kernel_vector_begin();
+	aes_encrypt_zvkned(&ctx->ctx2, req->iv, req->iv);
+	kernel_vector_end();
+
+	err = skcipher_walk_virt(&walk, req, false);
+
+	/*
+	 * If the message length isn't divisible by the AES block size and the
+	 * full message isn't available in one step of the scatterlist walk,
+	 * then separate off the last full block and the partial block.  This
+	 * ensures that they are processed in the same call to the assembly
+	 * function, which is required for ciphertext stealing.
+	 */
+	if (unlikely(tail > 0 && walk.nbytes < walk.total)) {
+		skcipher_walk_abort(&walk);
+
+		skcipher_request_set_tfm(&subreq, tfm);
+		skcipher_request_set_callback(&subreq,
+					      skcipher_request_flags(req),
+					      NULL, NULL);
+		skcipher_request_set_crypt(&subreq, req->src, req->dst,
+					   req->cryptlen - tail - AES_BLOCK_SIZE,
+					   req->iv);
+		req = &subreq;
+		err = skcipher_walk_virt(&walk, req, false);
+	} else {
+		tail = 0;
+	}
+
+	while (walk.nbytes) {
+		unsigned int nbytes = walk.nbytes;
+
+		if (nbytes < walk.total)
+			nbytes = round_down(nbytes, AES_BLOCK_SIZE);
+
+		kernel_vector_begin();
+		if (enc)
+			aes_xts_encrypt_zvkned_zvbb_zvkg(
+				&ctx->ctx1, walk.src.virt.addr,
+				walk.dst.virt.addr, nbytes, req->iv);
+		else
+			aes_xts_decrypt_zvkned_zvbb_zvkg(
+				&ctx->ctx1, walk.src.virt.addr,
+				walk.dst.virt.addr, nbytes, req->iv);
+		kernel_vector_end();
+		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
+	}
+
+	if (err || likely(!tail))
+		return err;
+
+	/* Do ciphertext stealing with the last full block and partial block. */
+
+	dst = src = scatterwalk_ffwd(sg_src, req->src, req->cryptlen);
+	if (req->dst != req->src)
+		dst = scatterwalk_ffwd(sg_dst, req->dst, req->cryptlen);
+
+	skcipher_request_set_crypt(req, src, dst, AES_BLOCK_SIZE + tail,
+				   req->iv);
+
+	err = skcipher_walk_virt(&walk, req, false);
+	if (err)
+		return err;
+
+	kernel_vector_begin();
+	if (enc)
+		aes_xts_encrypt_zvkned_zvbb_zvkg(
+			&ctx->ctx1, walk.src.virt.addr,
+			walk.dst.virt.addr, walk.nbytes, req->iv);
+	else
+		aes_xts_decrypt_zvkned_zvbb_zvkg(
+			&ctx->ctx1, walk.src.virt.addr,
+			walk.dst.virt.addr, walk.nbytes, req->iv);
+	kernel_vector_end();
+
+	return skcipher_walk_done(&walk, 0);
+}
+
+static int riscv64_aes_xts_encrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_xts_crypt(req, true);
+}
+
+static int riscv64_aes_xts_decrypt(struct skcipher_request *req)
+{
+	return riscv64_aes_xts_crypt(req, false);
+}
+
+/* Algorithm definitions */
+
+static struct crypto_alg riscv64_zvkned_aes_cipher_alg = {
+	.cra_flags = CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize = AES_BLOCK_SIZE,
+	.cra_ctxsize = sizeof(struct crypto_aes_ctx),
+	.cra_priority = 300,
+	.cra_name = "aes",
+	.cra_driver_name = "aes-riscv64-zvkned",
+	.cra_cipher = {
+		.cia_min_keysize = AES_MIN_KEY_SIZE,
+		.cia_max_keysize = AES_MAX_KEY_SIZE,
+		.cia_setkey = riscv64_aes_setkey_cipher,
+		.cia_encrypt = riscv64_aes_encrypt,
+		.cia_decrypt = riscv64_aes_decrypt,
+	},
+	.cra_module = THIS_MODULE,
+};
+
+static struct skcipher_alg riscv64_zvkned_aes_skcipher_algs[] = {
+	{
+		.setkey = riscv64_aes_setkey_skcipher,
+		.encrypt = riscv64_aes_ecb_encrypt,
+		.decrypt = riscv64_aes_ecb_decrypt,
+		.min_keysize = AES_MIN_KEY_SIZE,
+		.max_keysize = AES_MAX_KEY_SIZE,
+		.walksize = 8 * AES_BLOCK_SIZE, /* matches LMUL=8 */
+		.base = {
+			.cra_blocksize = AES_BLOCK_SIZE,
+			.cra_ctxsize = sizeof(struct crypto_aes_ctx),
+			.cra_priority = 300,
+			.cra_name = "ecb(aes)",
+			.cra_driver_name = "ecb-aes-riscv64-zvkned",
+			.cra_module = THIS_MODULE,
+		},
+	}, {
+		.setkey = riscv64_aes_setkey_skcipher,
+		.encrypt = riscv64_aes_cbc_encrypt,
+		.decrypt = riscv64_aes_cbc_decrypt,
+		.min_keysize = AES_MIN_KEY_SIZE,
+		.max_keysize = AES_MAX_KEY_SIZE,
+		.ivsize = AES_BLOCK_SIZE,
+		.base = {
+			.cra_blocksize = AES_BLOCK_SIZE,
+			.cra_ctxsize = sizeof(struct crypto_aes_ctx),
+			.cra_priority = 300,
+			.cra_name = "cbc(aes)",
+			.cra_driver_name = "cbc-aes-riscv64-zvkned",
+			.cra_module = THIS_MODULE,
+		},
+	}, {
+		.setkey = riscv64_aes_setkey_skcipher,
+		.encrypt = riscv64_aes_cbc_cts_encrypt,
+		.decrypt = riscv64_aes_cbc_cts_decrypt,
+		.min_keysize = AES_MIN_KEY_SIZE,
+		.max_keysize = AES_MAX_KEY_SIZE,
+		.ivsize = AES_BLOCK_SIZE,
+		.walksize = 4 * AES_BLOCK_SIZE, /* matches LMUL=4 */
+		.base = {
+			.cra_blocksize = AES_BLOCK_SIZE,
+			.cra_ctxsize = sizeof(struct crypto_aes_ctx),
+			.cra_priority = 300,
+			.cra_name = "cts(cbc(aes))",
+			.cra_driver_name = "cts-cbc-aes-riscv64-zvkned",
+			.cra_module = THIS_MODULE,
+		},
+	}
+};
+
+static struct skcipher_alg riscv64_zvkned_zvkb_aes_skcipher_alg = {
+	.setkey = riscv64_aes_setkey_skcipher,
+	.encrypt = riscv64_aes_ctr_crypt,
+	.decrypt = riscv64_aes_ctr_crypt,
+	.min_keysize = AES_MIN_KEY_SIZE,
+	.max_keysize = AES_MAX_KEY_SIZE,
+	.ivsize = AES_BLOCK_SIZE,
+	.chunksize = AES_BLOCK_SIZE,
+	.walksize = 4 * AES_BLOCK_SIZE, /* matches LMUL=4 */
+	.base = {
+		.cra_blocksize = 1,
+		.cra_ctxsize = sizeof(struct crypto_aes_ctx),
+		.cra_priority = 300,
+		.cra_name = "ctr(aes)",
+		.cra_driver_name = "ctr-aes-riscv64-zvkned-zvkb",
+		.cra_module = THIS_MODULE,
+	},
+};
+
+static struct skcipher_alg riscv64_zvkned_zvbb_zvkg_aes_skcipher_alg = {
+	.setkey = riscv64_aes_xts_setkey,
+	.encrypt = riscv64_aes_xts_encrypt,
+	.decrypt = riscv64_aes_xts_decrypt,
+	.min_keysize = 2 * AES_MIN_KEY_SIZE,
+	.max_keysize = 2 * AES_MAX_KEY_SIZE,
+	.ivsize = AES_BLOCK_SIZE,
+	.chunksize = AES_BLOCK_SIZE,
+	.walksize = 4 * AES_BLOCK_SIZE, /* matches LMUL=4 */
+	.base = {
+		.cra_blocksize = AES_BLOCK_SIZE,
+		.cra_ctxsize = sizeof(struct riscv64_aes_xts_ctx),
+		.cra_priority = 300,
+		.cra_name = "xts(aes)",
+		.cra_driver_name = "xts-aes-riscv64-zvkned-zvbb-zvkg",
+		.cra_module = THIS_MODULE,
+	},
+};
+
+static inline bool riscv64_aes_xts_supported(void)
+{
+	return riscv_isa_extension_available(NULL, ZVBB) &&
+	       riscv_isa_extension_available(NULL, ZVKG) &&
+	       riscv_vector_vlen() < 2048 /* Implementation limitation */;
+}
+
+static int __init riscv64_aes_mod_init(void)
+{
+	int err = -ENODEV;
+
+	if (riscv_isa_extension_available(NULL, ZVKNED) &&
+	    riscv_vector_vlen() >= 128) {
+		err = crypto_register_alg(&riscv64_zvkned_aes_cipher_alg);
+		if (err)
+			return err;
+
+		err = crypto_register_skciphers(
+			riscv64_zvkned_aes_skcipher_algs,
+			ARRAY_SIZE(riscv64_zvkned_aes_skcipher_algs));
+		if (err)
+			goto unregister_zvkned_cipher_alg;
+
+		if (riscv_isa_extension_available(NULL, ZVKB)) {
+			err = crypto_register_skcipher(
+				&riscv64_zvkned_zvkb_aes_skcipher_alg);
+			if (err)
+				goto unregister_zvkned_skcipher_algs;
+		}
+
+		if (riscv64_aes_xts_supported()) {
+			err = crypto_register_skcipher(
+				&riscv64_zvkned_zvbb_zvkg_aes_skcipher_alg);
+			if (err)
+				goto unregister_zvkned_zvkb_skcipher_alg;
+		}
+	}
+
+	return err;
+
+unregister_zvkned_zvkb_skcipher_alg:
+	if (riscv_isa_extension_available(NULL, ZVKB))
+		crypto_unregister_skcipher(&riscv64_zvkned_zvkb_aes_skcipher_alg);
+unregister_zvkned_skcipher_algs:
+	crypto_unregister_skciphers(riscv64_zvkned_aes_skcipher_algs,
+				    ARRAY_SIZE(riscv64_zvkned_aes_skcipher_algs));
+unregister_zvkned_cipher_alg:
+	crypto_unregister_alg(&riscv64_zvkned_aes_cipher_alg);
+	return err;
+}
+
+static void __exit riscv64_aes_mod_exit(void)
+{
+	if (riscv64_aes_xts_supported())
+		crypto_unregister_skcipher(&riscv64_zvkned_zvbb_zvkg_aes_skcipher_alg);
+	if (riscv_isa_extension_available(NULL, ZVKB))
+		crypto_unregister_skcipher(&riscv64_zvkned_zvkb_aes_skcipher_alg);
+	crypto_unregister_skciphers(riscv64_zvkned_aes_skcipher_algs,
+				    ARRAY_SIZE(riscv64_zvkned_aes_skcipher_algs));
+	crypto_unregister_alg(&riscv64_zvkned_aes_cipher_alg);
+}
+
+module_init(riscv64_aes_mod_init);
+module_exit(riscv64_aes_mod_exit);
+
+MODULE_DESCRIPTION("AES-ECB/CBC/CTS/CTR/XTS (RISC-V accelerated)");
+MODULE_AUTHOR("Jerry Shih <jerry.shih@sifive.com>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("aes");
+MODULE_ALIAS_CRYPTO("ecb(aes)");
+MODULE_ALIAS_CRYPTO("cbc(aes)");
+MODULE_ALIAS_CRYPTO("cts(cbc(aes))");
+MODULE_ALIAS_CRYPTO("ctr(aes)");
+MODULE_ALIAS_CRYPTO("xts(aes)");

arch/riscv/crypto/aes-riscv64-zvkned-zvbb-zvkg.S

@@ -0,0 +1,312 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128 && VLEN < 2048
+// - RISC-V Vector AES block cipher extension ('Zvkned')
+// - RISC-V Vector Bit-manipulation extension ('Zvbb')
+// - RISC-V Vector GCM/GMAC extension ('Zvkg')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvkned, +zvbb, +zvkg
+
+#include "aes-macros.S"
+
+#define KEYP		a0
+#define INP		a1
+#define OUTP		a2
+#define LEN		a3
+#define TWEAKP		a4
+
+#define LEN32		a5
+#define TAIL_LEN	a6
+#define VL		a7
+#define VLMAX		t4
+
+// v1-v15 contain the AES round keys, but they are used for temporaries before
+// the AES round keys have been loaded.
+#define TWEAKS		v16	// LMUL=4 (most of the time)
+#define TWEAKS_BREV	v20	// LMUL=4 (most of the time)
+#define MULTS_BREV	v24	// LMUL=4 (most of the time)
+#define TMP0		v28
+#define TMP1		v29
+#define TMP2		v30
+#define TMP3		v31
+
+// xts_init initializes the following values:
+//
+//	TWEAKS: N 128-bit tweaks T*(x^i) for i in 0..(N - 1)
+//	TWEAKS_BREV: same as TWEAKS, but bit-reversed
+//	MULTS_BREV: N 128-bit values x^N, bit-reversed.  Only if N > 1.
+//
+// N is the maximum number of blocks that will be processed per loop iteration,
+// computed using vsetvli.
+//
+// The field convention used by XTS is the same as that of GHASH, but with the
+// bits reversed within each byte.  The zvkg extension provides the vgmul
+// instruction which does multiplication in this field.  Therefore, for tweak
+// computation we use vgmul to do multiplications in parallel, instead of
+// serially multiplying by x using shifting+xoring.  Note that for this to work,
+// the inputs and outputs to vgmul must be bit-reversed (we do it with vbrev8).
+.macro	xts_init
+
+	// Load the first tweak T.
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vle32.v		TWEAKS, (TWEAKP)
+
+	// If there's only one block (or no blocks at all), then skip the tweak
+	// sequence computation because (at most) T itself is needed.
+	li		t0, 16
+	ble		LEN, t0, .Linit_single_block\@
+
+	// Save a copy of T bit-reversed in v12.
+	vbrev8.v	v12, TWEAKS
+
+	//
+	// Generate x^i for i in 0..(N - 1), i.e. 128-bit values 1 << i assuming
+	// that N <= 128.  Though, this code actually requires N < 64 (or
+	// equivalently VLEN < 2048) due to the use of 64-bit intermediate
+	// values here and in the x^N computation later.
+	//
+	vsetvli		VL, LEN32, e32, m4, ta, ma
+	srli		t0, VL, 2	// t0 = N (num blocks)
+	// Generate two sequences, each with N 32-bit values:
+	// v0=[1, 1, 1, ...] and v1=[0, 1, 2, ...].
+	vsetvli		zero, t0, e32, m1, ta, ma
+	vmv.v.i		v0, 1
+	vid.v		v1
+	// Use vzext to zero-extend the sequences to 64 bits.  Reinterpret them
+	// as two sequences, each with 2*N 32-bit values:
+	// v2=[1, 0, 1, 0, 1, 0, ...] and v4=[0, 0, 1, 0, 2, 0, ...].
+	vsetvli		zero, t0, e64, m2, ta, ma
+	vzext.vf2	v2, v0
+	vzext.vf2	v4, v1
+	slli		t1, t0, 1	// t1 = 2*N
+	vsetvli		zero, t1, e32, m2, ta, ma
+	// Use vwsll to compute [1<<0, 0<<0, 1<<1, 0<<0, 1<<2, 0<<0, ...],
+	// widening to 64 bits per element.  When reinterpreted as N 128-bit
+	// values, this is the needed sequence of 128-bit values 1 << i (x^i).
+	vwsll.vv	v8, v2, v4
+
+	// Copy the bit-reversed T to all N elements of TWEAKS_BREV, then
+	// multiply by x^i.  This gives the sequence T*(x^i), bit-reversed.
+	vsetvli		zero, LEN32, e32, m4, ta, ma
+	vmv.v.i		TWEAKS_BREV, 0
+	vaesz.vs	TWEAKS_BREV, v12
+	vbrev8.v	v8, v8
+	vgmul.vv	TWEAKS_BREV, v8
+
+	// Save a copy of the sequence T*(x^i) with the bit reversal undone.
+	vbrev8.v	TWEAKS, TWEAKS_BREV
+
+	// Generate N copies of x^N, i.e. 128-bit values 1 << N, bit-reversed.
+	li		t1, 1
+	sll		t1, t1, t0	// t1 = 1 << N
+	vsetivli	zero, 2, e64, m1, ta, ma
+	vmv.v.i		v0, 0
+	vsetivli	zero, 1, e64, m1, tu, ma
+	vmv.v.x		v0, t1
+	vbrev8.v	v0, v0
+	vsetvli		zero, LEN32, e32, m4, ta, ma
+	vmv.v.i		MULTS_BREV, 0
+	vaesz.vs	MULTS_BREV, v0
+
+	j		.Linit_done\@
+
+.Linit_single_block\@:
+	vbrev8.v	TWEAKS_BREV, TWEAKS
+.Linit_done\@:
+.endm
+
+// Set the first 128 bits of MULTS_BREV to 0x40, i.e. 'x' bit-reversed.  This is
+// the multiplier required to advance the tweak by one.
+.macro	load_x
+	li		t0, 0x40
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vmv.v.i		MULTS_BREV, 0
+	vsetivli	zero, 1, e8, m1, tu, ma
+	vmv.v.x		MULTS_BREV, t0
+.endm
+
+.macro	__aes_xts_crypt	enc, keylen
+	// With 16 < len <= 31, there's no main loop, just ciphertext stealing.
+	beqz		LEN32, .Lcts_without_main_loop\@
+
+	vsetvli		VLMAX, zero, e32, m4, ta, ma
+1:
+	vsetvli		VL, LEN32, e32, m4, ta, ma
+2:
+	// Encrypt or decrypt VL/4 blocks.
+	vle32.v		TMP0, (INP)
+	vxor.vv		TMP0, TMP0, TWEAKS
+	aes_crypt	TMP0, \enc, \keylen
+	vxor.vv		TMP0, TMP0, TWEAKS
+	vse32.v		TMP0, (OUTP)
+
+	// Update the pointers and the remaining length.
+	slli		t0, VL, 2
+	add		INP, INP, t0
+	add		OUTP, OUTP, t0
+	sub		LEN32, LEN32, VL
+
+	// Check whether more blocks remain.
+	beqz		LEN32, .Lmain_loop_done\@
+
+	// Compute the next sequence of tweaks by multiplying the previous
+	// sequence by x^N.  Store the result in both bit-reversed order and
+	// regular order (i.e. with the bit reversal undone).
+	vgmul.vv	TWEAKS_BREV, MULTS_BREV
+	vbrev8.v	TWEAKS, TWEAKS_BREV
+
+	// Since we compute the tweak multipliers x^N in advance, we require
+	// that each iteration process the same length except possibly the last.
+	// This conflicts slightly with the behavior allowed by RISC-V Vector
+	// Extension, where CPUs can select a lower length for both of the last
+	// two iterations.  E.g., vl might take the sequence of values
+	// [16, 16, 16, 12, 12], whereas we need [16, 16, 16, 16, 8] so that we
+	// can use x^4 again instead of computing x^3.  Therefore, we explicitly
+	// keep the vl at VLMAX if there is at least VLMAX remaining.
+	bge		LEN32, VLMAX, 2b
+	j		1b
+
+.Lmain_loop_done\@:
+	load_x
+
+	// Compute the next tweak.
+	addi		t0, VL, -4
+	vsetivli	zero, 4, e32, m4, ta, ma
+	vslidedown.vx	TWEAKS_BREV, TWEAKS_BREV, t0	// Extract last tweak
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vgmul.vv	TWEAKS_BREV, MULTS_BREV		// Advance to next tweak
+
+	bnez		TAIL_LEN, .Lcts\@
+
+	// Update *TWEAKP to contain the next tweak.
+	vbrev8.v	TWEAKS, TWEAKS_BREV
+	vse32.v		TWEAKS, (TWEAKP)
+	ret
+
+.Lcts_without_main_loop\@:
+	load_x
+.Lcts\@:
+	// TWEAKS_BREV now contains the next tweak.  Compute the one after that.
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vmv.v.v		TMP0, TWEAKS_BREV
+	vgmul.vv	TMP0, MULTS_BREV
+	// Undo the bit reversal of the next two tweaks and store them in TMP1
+	// and TMP2, such that TMP1 is the first needed and TMP2 the second.
+.if \enc
+	vbrev8.v	TMP1, TWEAKS_BREV
+	vbrev8.v	TMP2, TMP0
+.else
+	vbrev8.v	TMP1, TMP0
+	vbrev8.v	TMP2, TWEAKS_BREV
+.endif
+
+	// Encrypt/decrypt the last full block.
+	vle32.v		TMP0, (INP)
+	vxor.vv		TMP0, TMP0, TMP1
+	aes_crypt	TMP0, \enc, \keylen
+	vxor.vv		TMP0, TMP0, TMP1
+
+	// Swap the first TAIL_LEN bytes of the above result with the tail.
+	// Note that to support in-place encryption/decryption, the load from
+	// the input tail must happen before the store to the output tail.
+	addi		t0, INP, 16
+	addi		t1, OUTP, 16
+	vmv.v.v		TMP3, TMP0
+	vsetvli		zero, TAIL_LEN, e8, m1, tu, ma
+	vle8.v		TMP0, (t0)
+	vse8.v		TMP3, (t1)
+
+	// Encrypt/decrypt again and store the last full block.
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vxor.vv		TMP0, TMP0, TMP2
+	aes_crypt	TMP0, \enc, \keylen
+	vxor.vv		TMP0, TMP0, TMP2
+	vse32.v		TMP0, (OUTP)
+
+	ret
+.endm
+
+.macro	aes_xts_crypt	enc
+
+	// Check whether the length is a multiple of the AES block size.
+	andi		TAIL_LEN, LEN, 15
+	beqz		TAIL_LEN, 1f
+
+	// The length isn't a multiple of the AES block size, so ciphertext
+	// stealing will be required.  Ciphertext stealing involves special
+	// handling of the partial block and the last full block, so subtract
+	// the length of both from the length to be processed in the main loop.
+	sub		LEN, LEN, TAIL_LEN
+	addi		LEN, LEN, -16
+1:
+	srli		LEN32, LEN, 2
+	// LEN and LEN32 now contain the total length of the blocks that will be
+	// processed in the main loop, in bytes and 32-bit words respectively.
+
+	xts_init
+	aes_begin	KEYP, 128f, 192f
+	__aes_xts_crypt	\enc, 256
+128:
+	__aes_xts_crypt	\enc, 128
+192:
+	__aes_xts_crypt	\enc, 192
+.endm
+
+// void aes_xts_encrypt_zvkned_zvbb_zvkg(const struct crypto_aes_ctx *key,
+//					 const u8 *in, u8 *out, size_t len,
+//					 u8 tweak[16]);
+//
+// |key| is the data key.  |tweak| contains the next tweak; the encryption of
+// the original IV with the tweak key was already done.  This function supports
+// incremental computation, but |len| must always be >= 16 (AES_BLOCK_SIZE), and
+// |len| must be a multiple of 16 except on the last call.  If |len| is a
+// multiple of 16, then this function updates |tweak| to contain the next tweak.
+SYM_FUNC_START(aes_xts_encrypt_zvkned_zvbb_zvkg)
+	aes_xts_crypt	1
+SYM_FUNC_END(aes_xts_encrypt_zvkned_zvbb_zvkg)
+
+// Same prototype and calling convention as the encryption function
+SYM_FUNC_START(aes_xts_decrypt_zvkned_zvbb_zvkg)
+	aes_xts_crypt	0
+SYM_FUNC_END(aes_xts_decrypt_zvkned_zvbb_zvkg)

arch/riscv/crypto/aes-riscv64-zvkned-zvkb.S

@@ -0,0 +1,146 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector AES block cipher extension ('Zvkned')
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvkned, +zvkb
+
+#include "aes-macros.S"
+
+#define KEYP		a0
+#define INP		a1
+#define OUTP		a2
+#define LEN		a3
+#define IVP		a4
+
+#define LEN32		a5
+#define VL_E32		a6
+#define VL_BLOCKS	a7
+
+.macro	aes_ctr32_crypt	keylen
+	// LEN32 = number of blocks, rounded up, in 32-bit words.
+	addi		t0, LEN, 15
+	srli		t0, t0, 4
+	slli		LEN32, t0, 2
+
+	// Create a mask that selects the last 32-bit word of each 128-bit
+	// block.  This is the word that contains the (big-endian) counter.
+	li		t0, 0x88
+	vsetvli		t1, zero, e8, m1, ta, ma
+	vmv.v.x		v0, t0
+
+	// Load the IV into v31.  The last 32-bit word contains the counter.
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vle32.v		v31, (IVP)
+
+	// Convert the big-endian counter into little-endian.
+	vsetivli	zero, 4, e32, m1, ta, mu
+	vrev8.v		v31, v31, v0.t
+
+	// Splat the IV to v16 (with LMUL=4).  The number of copies is the
+	// maximum number of blocks that will be processed per iteration.
+	vsetvli		zero, LEN32, e32, m4, ta, ma
+	vmv.v.i		v16, 0
+	vaesz.vs	v16, v31
+
+	// v20 = [x, x, x, 0, x, x, x, 1, ...]
+	viota.m		v20, v0, v0.t
+	// v16 = [IV0, IV1, IV2, counter+0, IV0, IV1, IV2, counter+1, ...]
+	vsetvli		VL_E32, LEN32, e32, m4, ta, mu
+	vadd.vv		v16, v16, v20, v0.t
+
+	j 2f
+1:
+	// Set the number of blocks to process in this iteration.  vl=VL_E32 is
+	// the length in 32-bit words, i.e. 4 times the number of blocks.
+	vsetvli		VL_E32, LEN32, e32, m4, ta, mu
+
+	// Increment the counters by the number of blocks processed in the
+	// previous iteration.
+	vadd.vx		v16, v16, VL_BLOCKS, v0.t
+2:
+	// Prepare the AES inputs into v24.
+	vmv.v.v		v24, v16
+	vrev8.v		v24, v24, v0.t	// Convert counters back to big-endian.
+
+	// Encrypt the AES inputs to create the next portion of the keystream.
+	aes_encrypt	v24, \keylen
+
+	// XOR the data with the keystream.
+	vsetvli		t0, LEN, e8, m4, ta, ma
+	vle8.v		v20, (INP)
+	vxor.vv		v20, v20, v24
+	vse8.v		v20, (OUTP)
+
+	// Advance the pointers and update the remaining length.
+	add		INP, INP, t0
+	add		OUTP, OUTP, t0
+	sub		LEN, LEN, t0
+	sub		LEN32, LEN32, VL_E32
+	srli		VL_BLOCKS, VL_E32, 2
+
+	// Repeat if more data remains.
+	bnez		LEN, 1b
+
+	// Update *IVP to contain the next counter.
+	vsetivli	zero, 4, e32, m1, ta, mu
+	vadd.vx		v16, v16, VL_BLOCKS, v0.t
+	vrev8.v		v16, v16, v0.t	// Convert counters back to big-endian.
+	vse32.v		v16, (IVP)
+
+	ret
+.endm
+
+// void aes_ctr32_crypt_zvkned_zvkb(const struct crypto_aes_ctx *key,
+//				    const u8 *in, u8 *out, size_t len,
+//				    u8 iv[16]);
+SYM_FUNC_START(aes_ctr32_crypt_zvkned_zvkb)
+	aes_begin	KEYP, 128f, 192f
+	aes_ctr32_crypt	256
+128:
+	aes_ctr32_crypt	128
+192:
+	aes_ctr32_crypt	192
+SYM_FUNC_END(aes_ctr32_crypt_zvkned_zvkb)

arch/riscv/crypto/aes-riscv64-zvkned.S

@@ -0,0 +1,339 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com>
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector AES block cipher extension ('Zvkned')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvkned
+
+#include "aes-macros.S"
+
+#define KEYP		a0
+#define INP		a1
+#define OUTP		a2
+#define LEN		a3
+#define IVP		a4
+
+.macro	__aes_crypt_zvkned	enc, keylen
+	vle32.v		v16, (INP)
+	aes_crypt	v16, \enc, \keylen
+	vse32.v		v16, (OUTP)
+	ret
+.endm
+
+.macro	aes_crypt_zvkned	enc
+	aes_begin	KEYP, 128f, 192f
+	__aes_crypt_zvkned	\enc, 256
+128:
+	__aes_crypt_zvkned	\enc, 128
+192:
+	__aes_crypt_zvkned	\enc, 192
+.endm
+
+// void aes_encrypt_zvkned(const struct crypto_aes_ctx *key,
+//			   const u8 in[16], u8 out[16]);
+SYM_FUNC_START(aes_encrypt_zvkned)
+	aes_crypt_zvkned	1
+SYM_FUNC_END(aes_encrypt_zvkned)
+
+// Same prototype and calling convention as the encryption function
+SYM_FUNC_START(aes_decrypt_zvkned)
+	aes_crypt_zvkned	0
+SYM_FUNC_END(aes_decrypt_zvkned)
+
+.macro	__aes_ecb_crypt	enc, keylen
+	srli		t0, LEN, 2
+	// t0 is the remaining length in 32-bit words.  It's a multiple of 4.
+1:
+	vsetvli		t1, t0, e32, m8, ta, ma
+	sub		t0, t0, t1	// Subtract number of words processed
+	slli		t1, t1, 2	// Words to bytes
+	vle32.v		v16, (INP)
+	aes_crypt	v16, \enc, \keylen
+	vse32.v		v16, (OUTP)
+	add		INP, INP, t1
+	add		OUTP, OUTP, t1
+	bnez		t0, 1b
+
+	ret
+.endm
+
+.macro	aes_ecb_crypt	enc
+	aes_begin	KEYP, 128f, 192f
+	__aes_ecb_crypt	\enc, 256
+128:
+	__aes_ecb_crypt	\enc, 128
+192:
+	__aes_ecb_crypt	\enc, 192
+.endm
+
+// void aes_ecb_encrypt_zvkned(const struct crypto_aes_ctx *key,
+//			       const u8 *in, u8 *out, size_t len);
+//
+// |len| must be nonzero and a multiple of 16 (AES_BLOCK_SIZE).
+SYM_FUNC_START(aes_ecb_encrypt_zvkned)
+	aes_ecb_crypt	1
+SYM_FUNC_END(aes_ecb_encrypt_zvkned)
+
+// Same prototype and calling convention as the encryption function
+SYM_FUNC_START(aes_ecb_decrypt_zvkned)
+	aes_ecb_crypt	0
+SYM_FUNC_END(aes_ecb_decrypt_zvkned)
+
+.macro	aes_cbc_encrypt	keylen
+	vle32.v		v16, (IVP)	// Load IV
+1:
+	vle32.v		v17, (INP)	// Load plaintext block
+	vxor.vv		v16, v16, v17	// XOR with IV or prev ciphertext block
+	aes_encrypt	v16, \keylen	// Encrypt
+	vse32.v		v16, (OUTP)	// Store ciphertext block
+	addi		INP, INP, 16
+	addi		OUTP, OUTP, 16
+	addi		LEN, LEN, -16
+	bnez		LEN, 1b
+
+	vse32.v		v16, (IVP)	// Store next IV
+	ret
+.endm
+
+.macro	aes_cbc_decrypt	keylen
+	srli		LEN, LEN, 2	// Convert LEN from bytes to words
+	vle32.v		v16, (IVP)	// Load IV
+1:
+	vsetvli		t0, LEN, e32, m4, ta, ma
+	vle32.v		v20, (INP)	// Load ciphertext blocks
+	vslideup.vi	v16, v20, 4	// Setup prev ciphertext blocks
+	addi		t1, t0, -4
+	vslidedown.vx	v24, v20, t1	// Save last ciphertext block
+	aes_decrypt	v20, \keylen	// Decrypt the blocks
+	vxor.vv		v20, v20, v16	// XOR with prev ciphertext blocks
+	vse32.v		v20, (OUTP)	// Store plaintext blocks
+	vmv.v.v		v16, v24	// Next "IV" is last ciphertext block
+	slli		t1, t0, 2	// Words to bytes
+	add		INP, INP, t1
+	add		OUTP, OUTP, t1
+	sub		LEN, LEN, t0
+	bnez		LEN, 1b
+
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vse32.v		v16, (IVP)	// Store next IV
+	ret
+.endm
+
+// void aes_cbc_encrypt_zvkned(const struct crypto_aes_ctx *key,
+//			       const u8 *in, u8 *out, size_t len, u8 iv[16]);
+//
+// |len| must be nonzero and a multiple of 16 (AES_BLOCK_SIZE).
+SYM_FUNC_START(aes_cbc_encrypt_zvkned)
+	aes_begin	KEYP, 128f, 192f
+	aes_cbc_encrypt	256
+128:
+	aes_cbc_encrypt	128
+192:
+	aes_cbc_encrypt	192
+SYM_FUNC_END(aes_cbc_encrypt_zvkned)
+
+// Same prototype and calling convention as the encryption function
+SYM_FUNC_START(aes_cbc_decrypt_zvkned)
+	aes_begin	KEYP, 128f, 192f
+	aes_cbc_decrypt	256
+128:
+	aes_cbc_decrypt	128
+192:
+	aes_cbc_decrypt	192
+SYM_FUNC_END(aes_cbc_decrypt_zvkned)
+
+.macro	aes_cbc_cts_encrypt	keylen
+
+	// CBC-encrypt all blocks except the last.  But don't store the
+	// second-to-last block to the output buffer yet, since it will be
+	// handled specially in the ciphertext stealing step.  Exception: if the
+	// message is single-block, still encrypt the last (and only) block.
+	li		t0, 16
+	j		2f
+1:
+	vse32.v		v16, (OUTP)	// Store ciphertext block
+	addi		OUTP, OUTP, 16
+2:
+	vle32.v		v17, (INP)	// Load plaintext block
+	vxor.vv		v16, v16, v17	// XOR with IV or prev ciphertext block
+	aes_encrypt	v16, \keylen	// Encrypt
+	addi		INP, INP, 16
+	addi		LEN, LEN, -16
+	bgt		LEN, t0, 1b	// Repeat if more than one block remains
+
+	// Special case: if the message is a single block, just do CBC.
+	beqz		LEN, .Lcts_encrypt_done\@
+
+	// Encrypt the last two blocks using ciphertext stealing as follows:
+	//	C[n-1] = Encrypt(Encrypt(P[n-1] ^ C[n-2]) ^ P[n])
+	//	C[n] = Encrypt(P[n-1] ^ C[n-2])[0..LEN]
+	//
+	// C[i] denotes the i'th ciphertext block, and likewise P[i] the i'th
+	// plaintext block.  Block n, the last block, may be partial; its length
+	// is 1 <= LEN <= 16.  If there are only 2 blocks, C[n-2] means the IV.
+	//
+	// v16 already contains Encrypt(P[n-1] ^ C[n-2]).
+	// INP points to P[n].  OUTP points to where C[n-1] should go.
+	// To support in-place encryption, load P[n] before storing C[n].
+	addi		t0, OUTP, 16	// Get pointer to where C[n] should go
+	vsetvli		zero, LEN, e8, m1, tu, ma
+	vle8.v		v17, (INP)	// Load P[n]
+	vse8.v		v16, (t0)	// Store C[n]
+	vxor.vv		v16, v16, v17	// v16 = Encrypt(P[n-1] ^ C[n-2]) ^ P[n]
+	vsetivli	zero, 4, e32, m1, ta, ma
+	aes_encrypt	v16, \keylen
+.Lcts_encrypt_done\@:
+	vse32.v		v16, (OUTP)	// Store C[n-1] (or C[n] in single-block case)
+	ret
+.endm
+
+#define LEN32		t4 // Length of remaining full blocks in 32-bit words
+#define LEN_MOD16	t5 // Length of message in bytes mod 16
+
+.macro	aes_cbc_cts_decrypt	keylen
+	andi		LEN32, LEN, ~15
+	srli		LEN32, LEN32, 2
+	andi		LEN_MOD16, LEN, 15
+
+	// Save C[n-2] in v28 so that it's available later during the ciphertext
+	// stealing step.  If there are fewer than three blocks, C[n-2] means
+	// the IV, otherwise it means the third-to-last ciphertext block.
+	vmv.v.v		v28, v16	// IV
+	add		t0, LEN, -33
+	bltz		t0, .Lcts_decrypt_loop\@
+	andi		t0, t0, ~15
+	add		t0, t0, INP
+	vle32.v		v28, (t0)
+
+	// CBC-decrypt all full blocks.  For the last full block, or the last 2
+	// full blocks if the message is block-aligned, this doesn't write the
+	// correct output blocks (unless the message is only a single block),
+	// because it XORs the wrong values with the raw AES plaintexts.  But we
+	// fix this after this loop without redoing the AES decryptions.  This
+	// approach allows more of the AES decryptions to be parallelized.
+.Lcts_decrypt_loop\@:
+	vsetvli		t0, LEN32, e32, m4, ta, ma
+	addi		t1, t0, -4
+	vle32.v		v20, (INP)	// Load next set of ciphertext blocks
+	vmv.v.v		v24, v16	// Get IV or last ciphertext block of prev set
+	vslideup.vi	v24, v20, 4	// Setup prev ciphertext blocks
+	vslidedown.vx	v16, v20, t1	// Save last ciphertext block of this set
+	aes_decrypt	v20, \keylen	// Decrypt this set of blocks
+	vxor.vv		v24, v24, v20	// XOR prev ciphertext blocks with decrypted blocks
+	vse32.v		v24, (OUTP)	// Store this set of plaintext blocks
+	sub		LEN32, LEN32, t0
+	slli		t0, t0, 2	// Words to bytes
+	add		INP, INP, t0
+	add		OUTP, OUTP, t0
+	bnez		LEN32, .Lcts_decrypt_loop\@
+
+	vsetivli	zero, 4, e32, m4, ta, ma
+	vslidedown.vx	v20, v20, t1	// Extract raw plaintext of last full block
+	addi		t0, OUTP, -16	// Get pointer to last full plaintext block
+	bnez		LEN_MOD16, .Lcts_decrypt_non_block_aligned\@
+
+	// Special case: if the message is a single block, just do CBC.
+	li		t1, 16
+	beq		LEN, t1, .Lcts_decrypt_done\@
+
+	// Block-aligned message.  Just fix up the last 2 blocks.  We need:
+	//
+	//	P[n-1] = Decrypt(C[n]) ^ C[n-2]
+	//	P[n] = Decrypt(C[n-1]) ^ C[n]
+	//
+	// We have C[n] in v16, Decrypt(C[n]) in v20, and C[n-2] in v28.
+	// Together with Decrypt(C[n-1]) ^ C[n-2] from the output buffer, this
+	// is everything needed to fix the output without re-decrypting blocks.
+	addi		t1, OUTP, -32	// Get pointer to where P[n-1] should go
+	vxor.vv		v20, v20, v28	// Decrypt(C[n]) ^ C[n-2] == P[n-1]
+	vle32.v		v24, (t1)	// Decrypt(C[n-1]) ^ C[n-2]
+	vse32.v		v20, (t1)	// Store P[n-1]
+	vxor.vv		v20, v24, v16	// Decrypt(C[n-1]) ^ C[n-2] ^ C[n] == P[n] ^ C[n-2]
+	j		.Lcts_decrypt_finish\@
+
+.Lcts_decrypt_non_block_aligned\@:
+	// Decrypt the last two blocks using ciphertext stealing as follows:
+	//
+	//	P[n-1] = Decrypt(C[n] || Decrypt(C[n-1])[LEN_MOD16..16]) ^ C[n-2]
+	//	P[n] = (Decrypt(C[n-1]) ^ C[n])[0..LEN_MOD16]
+	//
+	// We already have Decrypt(C[n-1]) in v20 and C[n-2] in v28.
+	vmv.v.v		v16, v20	// v16 = Decrypt(C[n-1])
+	vsetvli		zero, LEN_MOD16, e8, m1, tu, ma
+	vle8.v		v20, (INP)	// v20 = C[n] || Decrypt(C[n-1])[LEN_MOD16..16]
+	vxor.vv		v16, v16, v20	// v16 = Decrypt(C[n-1]) ^ C[n]
+	vse8.v		v16, (OUTP)	// Store P[n]
+	vsetivli	zero, 4, e32, m1, ta, ma
+	aes_decrypt	v20, \keylen	// v20 = Decrypt(C[n] || Decrypt(C[n-1])[LEN_MOD16..16])
+.Lcts_decrypt_finish\@:
+	vxor.vv		v20, v20, v28	// XOR with C[n-2]
+	vse32.v		v20, (t0)	// Store last full plaintext block
+.Lcts_decrypt_done\@:
+	ret
+.endm
+
+.macro	aes_cbc_cts_crypt	keylen
+	vle32.v		v16, (IVP)	// Load IV
+	beqz		a5, .Lcts_decrypt\@
+	aes_cbc_cts_encrypt \keylen
+.Lcts_decrypt\@:
+	aes_cbc_cts_decrypt \keylen
+.endm
+
+// void aes_cbc_cts_crypt_zvkned(const struct crypto_aes_ctx *key,
+//			         const u8 *in, u8 *out, size_t len,
+//				 const u8 iv[16], bool enc);
+//
+// Encrypts or decrypts a message with the CS3 variant of AES-CBC-CTS.
+// This is the variant that unconditionally swaps the last two blocks.
+SYM_FUNC_START(aes_cbc_cts_crypt_zvkned)
+	aes_begin	KEYP, 128f, 192f
+	aes_cbc_cts_crypt 256
+128:
+	aes_cbc_cts_crypt 128
+192:
+	aes_cbc_cts_crypt 192
+SYM_FUNC_END(aes_cbc_cts_crypt_zvkned)

arch/riscv/crypto/chacha-riscv64-glue.c

@@ -0,0 +1,101 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * ChaCha20 using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/internal/chacha.h>
+#include <crypto/internal/skcipher.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+asmlinkage void chacha20_zvkb(const u32 key[8], const u8 *in, u8 *out,
+			      size_t len, const u32 iv[4]);
+
+static int riscv64_chacha20_crypt(struct skcipher_request *req)
+{
+	u32 iv[CHACHA_IV_SIZE / sizeof(u32)];
+	u8 block_buffer[CHACHA_BLOCK_SIZE];
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct skcipher_walk walk;
+	unsigned int nbytes;
+	unsigned int tail_bytes;
+	int err;
+
+	iv[0] = get_unaligned_le32(req->iv);
+	iv[1] = get_unaligned_le32(req->iv + 4);
+	iv[2] = get_unaligned_le32(req->iv + 8);
+	iv[3] = get_unaligned_le32(req->iv + 12);
+
+	err = skcipher_walk_virt(&walk, req, false);
+	while (walk.nbytes) {
+		nbytes = walk.nbytes & ~(CHACHA_BLOCK_SIZE - 1);
+		tail_bytes = walk.nbytes & (CHACHA_BLOCK_SIZE - 1);
+		kernel_vector_begin();
+		if (nbytes) {
+			chacha20_zvkb(ctx->key, walk.src.virt.addr,
+				      walk.dst.virt.addr, nbytes, iv);
+			iv[0] += nbytes / CHACHA_BLOCK_SIZE;
+		}
+		if (walk.nbytes == walk.total && tail_bytes > 0) {
+			memcpy(block_buffer, walk.src.virt.addr + nbytes,
+			       tail_bytes);
+			chacha20_zvkb(ctx->key, block_buffer, block_buffer,
+				      CHACHA_BLOCK_SIZE, iv);
+			memcpy(walk.dst.virt.addr + nbytes, block_buffer,
+			       tail_bytes);
+			tail_bytes = 0;
+		}
+		kernel_vector_end();
+
+		err = skcipher_walk_done(&walk, tail_bytes);
+	}
+
+	return err;
+}
+
+static struct skcipher_alg riscv64_chacha_alg = {
+	.setkey = chacha20_setkey,
+	.encrypt = riscv64_chacha20_crypt,
+	.decrypt = riscv64_chacha20_crypt,
+	.min_keysize = CHACHA_KEY_SIZE,
+	.max_keysize = CHACHA_KEY_SIZE,
+	.ivsize = CHACHA_IV_SIZE,
+	.chunksize = CHACHA_BLOCK_SIZE,
+	.walksize = 4 * CHACHA_BLOCK_SIZE,
+	.base = {
+		.cra_blocksize = 1,
+		.cra_ctxsize = sizeof(struct chacha_ctx),
+		.cra_priority = 300,
+		.cra_name = "chacha20",
+		.cra_driver_name = "chacha20-riscv64-zvkb",
+		.cra_module = THIS_MODULE,
+	},
+};
+
+static int __init riscv64_chacha_mod_init(void)
+{
+	if (riscv_isa_extension_available(NULL, ZVKB) &&
+	    riscv_vector_vlen() >= 128)
+		return crypto_register_skcipher(&riscv64_chacha_alg);
+
+	return -ENODEV;
+}
+
+static void __exit riscv64_chacha_mod_exit(void)
+{
+	crypto_unregister_skcipher(&riscv64_chacha_alg);
+}
+
+module_init(riscv64_chacha_mod_init);
+module_exit(riscv64_chacha_mod_exit);
+
+MODULE_DESCRIPTION("ChaCha20 (RISC-V accelerated)");
+MODULE_AUTHOR("Jerry Shih <jerry.shih@sifive.com>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("chacha20");

arch/riscv/crypto/chacha-riscv64-zvkb.S

@@ -0,0 +1,294 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvkb
+
+#define KEYP		a0
+#define INP		a1
+#define OUTP		a2
+#define LEN		a3
+#define IVP		a4
+
+#define CONSTS0		a5
+#define CONSTS1		a6
+#define CONSTS2		a7
+#define CONSTS3		t0
+#define TMP		t1
+#define VL		t2
+#define STRIDE		t3
+#define NROUNDS		t4
+#define KEY0		s0
+#define KEY1		s1
+#define KEY2		s2
+#define KEY3		s3
+#define KEY4		s4
+#define KEY5		s5
+#define KEY6		s6
+#define KEY7		s7
+#define COUNTER		s8
+#define NONCE0		s9
+#define NONCE1		s10
+#define NONCE2		s11
+
+.macro	chacha_round	a0, b0, c0, d0,  a1, b1, c1, d1, \
+			a2, b2, c2, d2,  a3, b3, c3, d3
+	// a += b; d ^= a; d = rol(d, 16);
+	vadd.vv		\a0, \a0, \b0
+	vadd.vv		\a1, \a1, \b1
+	vadd.vv		\a2, \a2, \b2
+	vadd.vv		\a3, \a3, \b3
+	vxor.vv		\d0, \d0, \a0
+	vxor.vv		\d1, \d1, \a1
+	vxor.vv		\d2, \d2, \a2
+	vxor.vv		\d3, \d3, \a3
+	vror.vi		\d0, \d0, 32 - 16
+	vror.vi		\d1, \d1, 32 - 16
+	vror.vi		\d2, \d2, 32 - 16
+	vror.vi		\d3, \d3, 32 - 16
+
+	// c += d; b ^= c; b = rol(b, 12);
+	vadd.vv		\c0, \c0, \d0
+	vadd.vv		\c1, \c1, \d1
+	vadd.vv		\c2, \c2, \d2
+	vadd.vv		\c3, \c3, \d3
+	vxor.vv		\b0, \b0, \c0
+	vxor.vv		\b1, \b1, \c1
+	vxor.vv		\b2, \b2, \c2
+	vxor.vv		\b3, \b3, \c3
+	vror.vi		\b0, \b0, 32 - 12
+	vror.vi		\b1, \b1, 32 - 12
+	vror.vi		\b2, \b2, 32 - 12
+	vror.vi		\b3, \b3, 32 - 12
+
+	// a += b; d ^= a; d = rol(d, 8);
+	vadd.vv		\a0, \a0, \b0
+	vadd.vv		\a1, \a1, \b1
+	vadd.vv		\a2, \a2, \b2
+	vadd.vv		\a3, \a3, \b3
+	vxor.vv		\d0, \d0, \a0
+	vxor.vv		\d1, \d1, \a1
+	vxor.vv		\d2, \d2, \a2
+	vxor.vv		\d3, \d3, \a3
+	vror.vi		\d0, \d0, 32 - 8
+	vror.vi		\d1, \d1, 32 - 8
+	vror.vi		\d2, \d2, 32 - 8
+	vror.vi		\d3, \d3, 32 - 8
+
+	// c += d; b ^= c; b = rol(b, 7);
+	vadd.vv		\c0, \c0, \d0
+	vadd.vv		\c1, \c1, \d1
+	vadd.vv		\c2, \c2, \d2
+	vadd.vv		\c3, \c3, \d3
+	vxor.vv		\b0, \b0, \c0
+	vxor.vv		\b1, \b1, \c1
+	vxor.vv		\b2, \b2, \c2
+	vxor.vv		\b3, \b3, \c3
+	vror.vi		\b0, \b0, 32 - 7
+	vror.vi		\b1, \b1, 32 - 7
+	vror.vi		\b2, \b2, 32 - 7
+	vror.vi		\b3, \b3, 32 - 7
+.endm
+
+// void chacha20_zvkb(const u32 key[8], const u8 *in, u8 *out, size_t len,
+//		      const u32 iv[4]);
+//
+// |len| must be nonzero and a multiple of 64 (CHACHA_BLOCK_SIZE).
+// The counter is treated as 32-bit, following the RFC7539 convention.
+SYM_FUNC_START(chacha20_zvkb)
+	srli		LEN, LEN, 6	// Bytes to blocks
+
+	addi		sp, sp, -96
+	sd		s0, 0(sp)
+	sd		s1, 8(sp)
+	sd		s2, 16(sp)
+	sd		s3, 24(sp)
+	sd		s4, 32(sp)
+	sd		s5, 40(sp)
+	sd		s6, 48(sp)
+	sd		s7, 56(sp)
+	sd		s8, 64(sp)
+	sd		s9, 72(sp)
+	sd		s10, 80(sp)
+	sd		s11, 88(sp)
+
+	li		STRIDE, 64
+
+	// Set up the initial state matrix in scalar registers.
+	li		CONSTS0, 0x61707865	// "expa" little endian
+	li		CONSTS1, 0x3320646e	// "nd 3" little endian
+	li		CONSTS2, 0x79622d32	// "2-by" little endian
+	li		CONSTS3, 0x6b206574	// "te k" little endian
+	lw		KEY0, 0(KEYP)
+	lw		KEY1, 4(KEYP)
+	lw		KEY2, 8(KEYP)
+	lw		KEY3, 12(KEYP)
+	lw		KEY4, 16(KEYP)
+	lw		KEY5, 20(KEYP)
+	lw		KEY6, 24(KEYP)
+	lw		KEY7, 28(KEYP)
+	lw		COUNTER, 0(IVP)
+	lw		NONCE0, 4(IVP)
+	lw		NONCE1, 8(IVP)
+	lw		NONCE2, 12(IVP)
+
+.Lblock_loop:
+	// Set vl to the number of blocks to process in this iteration.
+	vsetvli		VL, LEN, e32, m1, ta, ma
+
+	// Set up the initial state matrix for the next VL blocks in v0-v15.
+	// v{i} holds the i'th 32-bit word of the state matrix for all blocks.
+	// Note that only the counter word, at index 12, differs across blocks.
+	vmv.v.x		v0, CONSTS0
+	vmv.v.x		v1, CONSTS1
+	vmv.v.x		v2, CONSTS2
+	vmv.v.x		v3, CONSTS3
+	vmv.v.x		v4, KEY0
+	vmv.v.x		v5, KEY1
+	vmv.v.x		v6, KEY2
+	vmv.v.x		v7, KEY3
+	vmv.v.x		v8, KEY4
+	vmv.v.x		v9, KEY5
+	vmv.v.x		v10, KEY6
+	vmv.v.x		v11, KEY7
+	vid.v		v12
+	vadd.vx		v12, v12, COUNTER
+	vmv.v.x		v13, NONCE0
+	vmv.v.x		v14, NONCE1
+	vmv.v.x		v15, NONCE2
+
+	// Load the first half of the input data for each block into v16-v23.
+	// v{16+i} holds the i'th 32-bit word for all blocks.
+	vlsseg8e32.v	v16, (INP), STRIDE
+
+	li		NROUNDS, 20
+.Lnext_doubleround:
+	addi		NROUNDS, NROUNDS, -2
+	// column round
+	chacha_round	v0, v4, v8, v12, v1, v5, v9, v13, \
+			v2, v6, v10, v14, v3, v7, v11, v15
+	// diagonal round
+	chacha_round	v0, v5, v10, v15, v1, v6, v11, v12, \
+			v2, v7, v8, v13, v3, v4, v9, v14
+	bnez		NROUNDS, .Lnext_doubleround
+
+	// Load the second half of the input data for each block into v24-v31.
+	// v{24+i} holds the {8+i}'th 32-bit word for all blocks.
+	addi		TMP, INP, 32
+	vlsseg8e32.v	v24, (TMP), STRIDE
+
+	// Finalize the first half of the keystream for each block.
+	vadd.vx		v0, v0, CONSTS0
+	vadd.vx		v1, v1, CONSTS1
+	vadd.vx		v2, v2, CONSTS2
+	vadd.vx		v3, v3, CONSTS3
+	vadd.vx		v4, v4, KEY0
+	vadd.vx		v5, v5, KEY1
+	vadd.vx		v6, v6, KEY2
+	vadd.vx		v7, v7, KEY3
+
+	// Encrypt/decrypt the first half of the data for each block.
+	vxor.vv		v16, v16, v0
+	vxor.vv		v17, v17, v1
+	vxor.vv		v18, v18, v2
+	vxor.vv		v19, v19, v3
+	vxor.vv		v20, v20, v4
+	vxor.vv		v21, v21, v5
+	vxor.vv		v22, v22, v6
+	vxor.vv		v23, v23, v7
+
+	// Store the first half of the output data for each block.
+	vssseg8e32.v	v16, (OUTP), STRIDE
+
+	// Finalize the second half of the keystream for each block.
+	vadd.vx		v8, v8, KEY4
+	vadd.vx		v9, v9, KEY5
+	vadd.vx		v10, v10, KEY6
+	vadd.vx		v11, v11, KEY7
+	vid.v		v0
+	vadd.vx		v12, v12, COUNTER
+	vadd.vx		v13, v13, NONCE0
+	vadd.vx		v14, v14, NONCE1
+	vadd.vx		v15, v15, NONCE2
+	vadd.vv		v12, v12, v0
+
+	// Encrypt/decrypt the second half of the data for each block.
+	vxor.vv		v24, v24, v8
+	vxor.vv		v25, v25, v9
+	vxor.vv		v26, v26, v10
+	vxor.vv		v27, v27, v11
+	vxor.vv		v29, v29, v13
+	vxor.vv		v28, v28, v12
+	vxor.vv		v30, v30, v14
+	vxor.vv		v31, v31, v15
+
+	// Store the second half of the output data for each block.
+	addi		TMP, OUTP, 32
+	vssseg8e32.v	v24, (TMP), STRIDE
+
+	// Update the counter, the remaining number of blocks, and the input and
+	// output pointers according to the number of blocks processed (VL).
+	add		COUNTER, COUNTER, VL
+	sub		LEN, LEN, VL
+	slli		TMP, VL, 6
+	add		OUTP, OUTP, TMP
+	add		INP, INP, TMP
+	bnez		LEN, .Lblock_loop
+
+	ld		s0, 0(sp)
+	ld		s1, 8(sp)
+	ld		s2, 16(sp)
+	ld		s3, 24(sp)
+	ld		s4, 32(sp)
+	ld		s5, 40(sp)
+	ld		s6, 48(sp)
+	ld		s7, 56(sp)
+	ld		s8, 64(sp)
+	ld		s9, 72(sp)
+	ld		s10, 80(sp)
+	ld		s11, 88(sp)
+	addi		sp, sp, 96
+	ret
+SYM_FUNC_END(chacha20_zvkb)

arch/riscv/crypto/ghash-riscv64-glue.c

@@ -0,0 +1,168 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * GHASH using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2023 VRULL GmbH
+ * Author: Heiko Stuebner <heiko.stuebner@vrull.eu>
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/ghash.h>
+#include <crypto/internal/hash.h>
+#include <crypto/internal/simd.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+asmlinkage void ghash_zvkg(be128 *accumulator, const be128 *key, const u8 *data,
+			   size_t len);
+
+struct riscv64_ghash_tfm_ctx {
+	be128 key;
+};
+
+struct riscv64_ghash_desc_ctx {
+	be128 accumulator;
+	u8 buffer[GHASH_BLOCK_SIZE];
+	u32 bytes;
+};
+
+static int riscv64_ghash_setkey(struct crypto_shash *tfm, const u8 *key,
+				unsigned int keylen)
+{
+	struct riscv64_ghash_tfm_ctx *tctx = crypto_shash_ctx(tfm);
+
+	if (keylen != GHASH_BLOCK_SIZE)
+		return -EINVAL;
+
+	memcpy(&tctx->key, key, GHASH_BLOCK_SIZE);
+
+	return 0;
+}
+
+static int riscv64_ghash_init(struct shash_desc *desc)
+{
+	struct riscv64_ghash_desc_ctx *dctx = shash_desc_ctx(desc);
+
+	*dctx = (struct riscv64_ghash_desc_ctx){};
+
+	return 0;
+}
+
+static inline void
+riscv64_ghash_blocks(const struct riscv64_ghash_tfm_ctx *tctx,
+		     struct riscv64_ghash_desc_ctx *dctx,
+		     const u8 *src, size_t srclen)
+{
+	/* The srclen is nonzero and a multiple of 16. */
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		ghash_zvkg(&dctx->accumulator, &tctx->key, src, srclen);
+		kernel_vector_end();
+	} else {
+		do {
+			crypto_xor((u8 *)&dctx->accumulator, src,
+				   GHASH_BLOCK_SIZE);
+			gf128mul_lle(&dctx->accumulator, &tctx->key);
+			src += GHASH_BLOCK_SIZE;
+			srclen -= GHASH_BLOCK_SIZE;
+		} while (srclen);
+	}
+}
+
+static int riscv64_ghash_update(struct shash_desc *desc, const u8 *src,
+				unsigned int srclen)
+{
+	const struct riscv64_ghash_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
+	struct riscv64_ghash_desc_ctx *dctx = shash_desc_ctx(desc);
+	unsigned int len;
+
+	if (dctx->bytes) {
+		if (dctx->bytes + srclen < GHASH_BLOCK_SIZE) {
+			memcpy(dctx->buffer + dctx->bytes, src, srclen);
+			dctx->bytes += srclen;
+			return 0;
+		}
+		memcpy(dctx->buffer + dctx->bytes, src,
+		       GHASH_BLOCK_SIZE - dctx->bytes);
+		riscv64_ghash_blocks(tctx, dctx, dctx->buffer,
+				     GHASH_BLOCK_SIZE);
+		src += GHASH_BLOCK_SIZE - dctx->bytes;
+		srclen -= GHASH_BLOCK_SIZE - dctx->bytes;
+		dctx->bytes = 0;
+	}
+
+	len = round_down(srclen, GHASH_BLOCK_SIZE);
+	if (len) {
+		riscv64_ghash_blocks(tctx, dctx, src, len);
+		src += len;
+		srclen -= len;
+	}
+
+	if (srclen) {
+		memcpy(dctx->buffer, src, srclen);
+		dctx->bytes = srclen;
+	}
+
+	return 0;
+}
+
+static int riscv64_ghash_final(struct shash_desc *desc, u8 *out)
+{
+	const struct riscv64_ghash_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
+	struct riscv64_ghash_desc_ctx *dctx = shash_desc_ctx(desc);
+	int i;
+
+	if (dctx->bytes) {
+		for (i = dctx->bytes; i < GHASH_BLOCK_SIZE; i++)
+			dctx->buffer[i] = 0;
+
+		riscv64_ghash_blocks(tctx, dctx, dctx->buffer,
+				     GHASH_BLOCK_SIZE);
+	}
+
+	memcpy(out, &dctx->accumulator, GHASH_DIGEST_SIZE);
+	return 0;
+}
+
+static struct shash_alg riscv64_ghash_alg = {
+	.init = riscv64_ghash_init,
+	.update = riscv64_ghash_update,
+	.final = riscv64_ghash_final,
+	.setkey = riscv64_ghash_setkey,
+	.descsize = sizeof(struct riscv64_ghash_desc_ctx),
+	.digestsize = GHASH_DIGEST_SIZE,
+	.base = {
+		.cra_blocksize = GHASH_BLOCK_SIZE,
+		.cra_ctxsize = sizeof(struct riscv64_ghash_tfm_ctx),
+		.cra_priority = 300,
+		.cra_name = "ghash",
+		.cra_driver_name = "ghash-riscv64-zvkg",
+		.cra_module = THIS_MODULE,
+	},
+};
+
+static int __init riscv64_ghash_mod_init(void)
+{
+	if (riscv_isa_extension_available(NULL, ZVKG) &&
+	    riscv_vector_vlen() >= 128)
+		return crypto_register_shash(&riscv64_ghash_alg);
+
+	return -ENODEV;
+}
+
+static void __exit riscv64_ghash_mod_exit(void)
+{
+	crypto_unregister_shash(&riscv64_ghash_alg);
+}
+
+module_init(riscv64_ghash_mod_init);
+module_exit(riscv64_ghash_mod_exit);
+
+MODULE_DESCRIPTION("GHASH (RISC-V accelerated)");
+MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("ghash");

arch/riscv/crypto/ghash-riscv64-zvkg.S

@@ -0,0 +1,72 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector GCM/GMAC extension ('Zvkg')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvkg
+
+#define ACCUMULATOR	a0
+#define KEY		a1
+#define DATA		a2
+#define LEN		a3
+
+// void ghash_zvkg(be128 *accumulator, const be128 *key, const u8 *data,
+//		   size_t len);
+//
+// |len| must be nonzero and a multiple of 16 (GHASH_BLOCK_SIZE).
+SYM_FUNC_START(ghash_zvkg)
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vle32.v		v1, (ACCUMULATOR)
+	vle32.v		v2, (KEY)
+.Lnext_block:
+	vle32.v		v3, (DATA)
+	vghsh.vv	v1, v2, v3
+	addi		DATA, DATA, 16
+	addi		LEN, LEN, -16
+	bnez		LEN, .Lnext_block
+
+	vse32.v		v1, (ACCUMULATOR)
+	ret
+SYM_FUNC_END(ghash_zvkg)

arch/riscv/crypto/sha256-riscv64-glue.c

@@ -0,0 +1,137 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * SHA-256 and SHA-224 using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2022 VRULL GmbH
+ * Author: Heiko Stuebner <heiko.stuebner@vrull.eu>
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/internal/hash.h>
+#include <crypto/internal/simd.h>
+#include <crypto/sha256_base.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+/*
+ * Note: the asm function only uses the 'state' field of struct sha256_state.
+ * It is assumed to be the first field.
+ */
+asmlinkage void sha256_transform_zvknha_or_zvknhb_zvkb(
+	struct sha256_state *state, const u8 *data, int num_blocks);
+
+static int riscv64_sha256_update(struct shash_desc *desc, const u8 *data,
+				 unsigned int len)
+{
+	/*
+	 * Ensure struct sha256_state begins directly with the SHA-256
+	 * 256-bit internal state, as this is what the asm function expects.
+	 */
+	BUILD_BUG_ON(offsetof(struct sha256_state, state) != 0);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		sha256_base_do_update(desc, data, len,
+				      sha256_transform_zvknha_or_zvknhb_zvkb);
+		kernel_vector_end();
+	} else {
+		crypto_sha256_update(desc, data, len);
+	}
+	return 0;
+}
+
+static int riscv64_sha256_finup(struct shash_desc *desc, const u8 *data,
+				unsigned int len, u8 *out)
+{
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		if (len)
+			sha256_base_do_update(
+				desc, data, len,
+				sha256_transform_zvknha_or_zvknhb_zvkb);
+		sha256_base_do_finalize(
+			desc, sha256_transform_zvknha_or_zvknhb_zvkb);
+		kernel_vector_end();
+
+		return sha256_base_finish(desc, out);
+	}
+
+	return crypto_sha256_finup(desc, data, len, out);
+}
+
+static int riscv64_sha256_final(struct shash_desc *desc, u8 *out)
+{
+	return riscv64_sha256_finup(desc, NULL, 0, out);
+}
+
+static int riscv64_sha256_digest(struct shash_desc *desc, const u8 *data,
+				 unsigned int len, u8 *out)
+{
+	return sha256_base_init(desc) ?:
+	       riscv64_sha256_finup(desc, data, len, out);
+}
+
+static struct shash_alg riscv64_sha256_algs[] = {
+	{
+		.init = sha256_base_init,
+		.update = riscv64_sha256_update,
+		.final = riscv64_sha256_final,
+		.finup = riscv64_sha256_finup,
+		.digest = riscv64_sha256_digest,
+		.descsize = sizeof(struct sha256_state),
+		.digestsize = SHA256_DIGEST_SIZE,
+		.base = {
+			.cra_blocksize = SHA256_BLOCK_SIZE,
+			.cra_priority = 300,
+			.cra_name = "sha256",
+			.cra_driver_name = "sha256-riscv64-zvknha_or_zvknhb-zvkb",
+			.cra_module = THIS_MODULE,
+		},
+	}, {
+		.init = sha224_base_init,
+		.update = riscv64_sha256_update,
+		.final = riscv64_sha256_final,
+		.finup = riscv64_sha256_finup,
+		.descsize = sizeof(struct sha256_state),
+		.digestsize = SHA224_DIGEST_SIZE,
+		.base = {
+			.cra_blocksize = SHA224_BLOCK_SIZE,
+			.cra_priority = 300,
+			.cra_name = "sha224",
+			.cra_driver_name = "sha224-riscv64-zvknha_or_zvknhb-zvkb",
+			.cra_module = THIS_MODULE,
+		},
+	},
+};
+
+static int __init riscv64_sha256_mod_init(void)
+{
+	/* Both zvknha and zvknhb provide the SHA-256 instructions. */
+	if ((riscv_isa_extension_available(NULL, ZVKNHA) ||
+	     riscv_isa_extension_available(NULL, ZVKNHB)) &&
+	    riscv_isa_extension_available(NULL, ZVKB) &&
+	    riscv_vector_vlen() >= 128)
+		return crypto_register_shashes(riscv64_sha256_algs,
+					       ARRAY_SIZE(riscv64_sha256_algs));
+
+	return -ENODEV;
+}
+
+static void __exit riscv64_sha256_mod_exit(void)
+{
+	crypto_unregister_shashes(riscv64_sha256_algs,
+				  ARRAY_SIZE(riscv64_sha256_algs));
+}
+
+module_init(riscv64_sha256_mod_init);
+module_exit(riscv64_sha256_mod_exit);
+
+MODULE_DESCRIPTION("SHA-256 (RISC-V accelerated)");
+MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("sha256");
+MODULE_ALIAS_CRYPTO("sha224");

arch/riscv/crypto/sha256-riscv64-zvknha_or_zvknhb-zvkb.S

@@ -0,0 +1,225 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector SHA-2 Secure Hash extension ('Zvknha' or 'Zvknhb')
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/cfi_types.h>
+
+.text
+.option arch, +zvknha, +zvkb
+
+#define STATEP		a0
+#define DATA		a1
+#define NUM_BLOCKS	a2
+
+#define STATEP_C	a3
+
+#define MASK		v0
+#define INDICES		v1
+#define W0		v2
+#define W1		v3
+#define W2		v4
+#define W3		v5
+#define VTMP		v6
+#define FEBA		v7
+#define HGDC		v8
+#define K0		v10
+#define K1		v11
+#define K2		v12
+#define K3		v13
+#define K4		v14
+#define K5		v15
+#define K6		v16
+#define K7		v17
+#define K8		v18
+#define K9		v19
+#define K10		v20
+#define K11		v21
+#define K12		v22
+#define K13		v23
+#define K14		v24
+#define K15		v25
+#define PREV_FEBA	v26
+#define PREV_HGDC	v27
+
+// Do 4 rounds of SHA-256.  w0 contains the current 4 message schedule words.
+//
+// If not all the message schedule words have been computed yet, then this also
+// computes 4 more message schedule words.  w1-w3 contain the next 3 groups of 4
+// message schedule words; this macro computes the group after w3 and writes it
+// to w0.  This means that the next (w0, w1, w2, w3) is the current (w1, w2, w3,
+// w0), so the caller must cycle through the registers accordingly.
+.macro	sha256_4rounds	last, k, w0, w1, w2, w3
+	vadd.vv		VTMP, \k, \w0
+	vsha2cl.vv	HGDC, FEBA, VTMP
+	vsha2ch.vv	FEBA, HGDC, VTMP
+.if !\last
+	vmerge.vvm	VTMP, \w2, \w1, MASK
+	vsha2ms.vv	\w0, VTMP, \w3
+.endif
+.endm
+
+.macro	sha256_16rounds	last, k0, k1, k2, k3
+	sha256_4rounds	\last, \k0, W0, W1, W2, W3
+	sha256_4rounds	\last, \k1, W1, W2, W3, W0
+	sha256_4rounds	\last, \k2, W2, W3, W0, W1
+	sha256_4rounds	\last, \k3, W3, W0, W1, W2
+.endm
+
+// void sha256_transform_zvknha_or_zvknhb_zvkb(u32 state[8], const u8 *data,
+//					       int num_blocks);
+SYM_TYPED_FUNC_START(sha256_transform_zvknha_or_zvknhb_zvkb)
+
+	// Load the round constants into K0-K15.
+	vsetivli	zero, 4, e32, m1, ta, ma
+	la		t0, K256
+	vle32.v		K0, (t0)
+	addi		t0, t0, 16
+	vle32.v		K1, (t0)
+	addi		t0, t0, 16
+	vle32.v		K2, (t0)
+	addi		t0, t0, 16
+	vle32.v		K3, (t0)
+	addi		t0, t0, 16
+	vle32.v		K4, (t0)
+	addi		t0, t0, 16
+	vle32.v		K5, (t0)
+	addi		t0, t0, 16
+	vle32.v		K6, (t0)
+	addi		t0, t0, 16
+	vle32.v		K7, (t0)
+	addi		t0, t0, 16
+	vle32.v		K8, (t0)
+	addi		t0, t0, 16
+	vle32.v		K9, (t0)
+	addi		t0, t0, 16
+	vle32.v		K10, (t0)
+	addi		t0, t0, 16
+	vle32.v		K11, (t0)
+	addi		t0, t0, 16
+	vle32.v		K12, (t0)
+	addi		t0, t0, 16
+	vle32.v		K13, (t0)
+	addi		t0, t0, 16
+	vle32.v		K14, (t0)
+	addi		t0, t0, 16
+	vle32.v		K15, (t0)
+
+	// Setup mask for the vmerge to replace the first word (idx==0) in
+	// message scheduling.  There are 4 words, so an 8-bit mask suffices.
+	vsetivli	zero, 1, e8, m1, ta, ma
+	vmv.v.i		MASK, 0x01
+
+	// Load the state.  The state is stored as {a,b,c,d,e,f,g,h}, but we
+	// need {f,e,b,a},{h,g,d,c}.  The dst vtype is e32m1 and the index vtype
+	// is e8mf4.  We use index-load with the i8 indices {20, 16, 4, 0},
+	// loaded using the 32-bit little endian value 0x00041014.
+	li		t0, 0x00041014
+	vsetivli	zero, 1, e32, m1, ta, ma
+	vmv.v.x		INDICES, t0
+	addi		STATEP_C, STATEP, 8
+	vsetivli	zero, 4, e32, m1, ta, ma
+	vluxei8.v	FEBA, (STATEP), INDICES
+	vluxei8.v	HGDC, (STATEP_C), INDICES
+
+.Lnext_block:
+	addi		NUM_BLOCKS, NUM_BLOCKS, -1
+
+	// Save the previous state, as it's needed later.
+	vmv.v.v		PREV_FEBA, FEBA
+	vmv.v.v		PREV_HGDC, HGDC
+
+	// Load the next 512-bit message block and endian-swap each 32-bit word.
+	vle32.v		W0, (DATA)
+	vrev8.v		W0, W0
+	addi		DATA, DATA, 16
+	vle32.v		W1, (DATA)
+	vrev8.v		W1, W1
+	addi		DATA, DATA, 16
+	vle32.v		W2, (DATA)
+	vrev8.v		W2, W2
+	addi		DATA, DATA, 16
+	vle32.v		W3, (DATA)
+	vrev8.v		W3, W3
+	addi		DATA, DATA, 16
+
+	// Do the 64 rounds of SHA-256.
+	sha256_16rounds	0, K0, K1, K2, K3
+	sha256_16rounds	0, K4, K5, K6, K7
+	sha256_16rounds	0, K8, K9, K10, K11
+	sha256_16rounds	1, K12, K13, K14, K15
+
+	// Add the previous state.
+	vadd.vv		FEBA, FEBA, PREV_FEBA
+	vadd.vv		HGDC, HGDC, PREV_HGDC
+
+	// Repeat if more blocks remain.
+	bnez		NUM_BLOCKS, .Lnext_block
+
+	// Store the new state and return.
+	vsuxei8.v	FEBA, (STATEP), INDICES
+	vsuxei8.v	HGDC, (STATEP_C), INDICES
+	ret
+SYM_FUNC_END(sha256_transform_zvknha_or_zvknhb_zvkb)
+
+.section ".rodata"
+.p2align 2
+.type K256, @object
+K256:
+	.word		0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5
+	.word		0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5
+	.word		0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3
+	.word		0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174
+	.word		0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc
+	.word		0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da
+	.word		0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7
+	.word		0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967
+	.word		0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13
+	.word		0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85
+	.word		0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3
+	.word		0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070
+	.word		0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5
+	.word		0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3
+	.word		0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208
+	.word		0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
+.size K256, . - K256

arch/riscv/crypto/sha512-riscv64-glue.c

@@ -0,0 +1,133 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * SHA-512 and SHA-384 using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2023 VRULL GmbH
+ * Author: Heiko Stuebner <heiko.stuebner@vrull.eu>
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/internal/hash.h>
+#include <crypto/internal/simd.h>
+#include <crypto/sha512_base.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+/*
+ * Note: the asm function only uses the 'state' field of struct sha512_state.
+ * It is assumed to be the first field.
+ */
+asmlinkage void sha512_transform_zvknhb_zvkb(
+	struct sha512_state *state, const u8 *data, int num_blocks);
+
+static int riscv64_sha512_update(struct shash_desc *desc, const u8 *data,
+				 unsigned int len)
+{
+	/*
+	 * Ensure struct sha512_state begins directly with the SHA-512
+	 * 512-bit internal state, as this is what the asm function expects.
+	 */
+	BUILD_BUG_ON(offsetof(struct sha512_state, state) != 0);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		sha512_base_do_update(desc, data, len,
+				      sha512_transform_zvknhb_zvkb);
+		kernel_vector_end();
+	} else {
+		crypto_sha512_update(desc, data, len);
+	}
+	return 0;
+}
+
+static int riscv64_sha512_finup(struct shash_desc *desc, const u8 *data,
+				unsigned int len, u8 *out)
+{
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		if (len)
+			sha512_base_do_update(desc, data, len,
+					      sha512_transform_zvknhb_zvkb);
+		sha512_base_do_finalize(desc, sha512_transform_zvknhb_zvkb);
+		kernel_vector_end();
+
+		return sha512_base_finish(desc, out);
+	}
+
+	return crypto_sha512_finup(desc, data, len, out);
+}
+
+static int riscv64_sha512_final(struct shash_desc *desc, u8 *out)
+{
+	return riscv64_sha512_finup(desc, NULL, 0, out);
+}
+
+static int riscv64_sha512_digest(struct shash_desc *desc, const u8 *data,
+				 unsigned int len, u8 *out)
+{
+	return sha512_base_init(desc) ?:
+	       riscv64_sha512_finup(desc, data, len, out);
+}
+
+static struct shash_alg riscv64_sha512_algs[] = {
+	{
+		.init = sha512_base_init,
+		.update = riscv64_sha512_update,
+		.final = riscv64_sha512_final,
+		.finup = riscv64_sha512_finup,
+		.digest = riscv64_sha512_digest,
+		.descsize = sizeof(struct sha512_state),
+		.digestsize = SHA512_DIGEST_SIZE,
+		.base = {
+			.cra_blocksize = SHA512_BLOCK_SIZE,
+			.cra_priority = 300,
+			.cra_name = "sha512",
+			.cra_driver_name = "sha512-riscv64-zvknhb-zvkb",
+			.cra_module = THIS_MODULE,
+		},
+	}, {
+		.init = sha384_base_init,
+		.update = riscv64_sha512_update,
+		.final = riscv64_sha512_final,
+		.finup = riscv64_sha512_finup,
+		.descsize = sizeof(struct sha512_state),
+		.digestsize = SHA384_DIGEST_SIZE,
+		.base = {
+			.cra_blocksize = SHA384_BLOCK_SIZE,
+			.cra_priority = 300,
+			.cra_name = "sha384",
+			.cra_driver_name = "sha384-riscv64-zvknhb-zvkb",
+			.cra_module = THIS_MODULE,
+		},
+	},
+};
+
+static int __init riscv64_sha512_mod_init(void)
+{
+	if (riscv_isa_extension_available(NULL, ZVKNHB) &&
+	    riscv_isa_extension_available(NULL, ZVKB) &&
+	    riscv_vector_vlen() >= 128)
+		return crypto_register_shashes(riscv64_sha512_algs,
+					       ARRAY_SIZE(riscv64_sha512_algs));
+
+	return -ENODEV;
+}
+
+static void __exit riscv64_sha512_mod_exit(void)
+{
+	crypto_unregister_shashes(riscv64_sha512_algs,
+				  ARRAY_SIZE(riscv64_sha512_algs));
+}
+
+module_init(riscv64_sha512_mod_init);
+module_exit(riscv64_sha512_mod_exit);
+
+MODULE_DESCRIPTION("SHA-512 (RISC-V accelerated)");
+MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("sha512");
+MODULE_ALIAS_CRYPTO("sha384");

arch/riscv/crypto/sha512-riscv64-zvknhb-zvkb.S

@@ -0,0 +1,203 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector SHA-2 Secure Hash extension ('Zvknhb')
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/cfi_types.h>
+
+.text
+.option arch, +zvknhb, +zvkb
+
+#define STATEP		a0
+#define DATA		a1
+#define NUM_BLOCKS	a2
+
+#define STATEP_C	a3
+#define K		a4
+
+#define MASK		v0
+#define INDICES		v1
+#define W0		v10	// LMUL=2
+#define W1		v12	// LMUL=2
+#define W2		v14	// LMUL=2
+#define W3		v16	// LMUL=2
+#define VTMP		v20	// LMUL=2
+#define FEBA		v22	// LMUL=2
+#define HGDC		v24	// LMUL=2
+#define PREV_FEBA	v26	// LMUL=2
+#define PREV_HGDC	v28	// LMUL=2
+
+// Do 4 rounds of SHA-512.  w0 contains the current 4 message schedule words.
+//
+// If not all the message schedule words have been computed yet, then this also
+// computes 4 more message schedule words.  w1-w3 contain the next 3 groups of 4
+// message schedule words; this macro computes the group after w3 and writes it
+// to w0.  This means that the next (w0, w1, w2, w3) is the current (w1, w2, w3,
+// w0), so the caller must cycle through the registers accordingly.
+.macro	sha512_4rounds	last, w0, w1, w2, w3
+	vle64.v		VTMP, (K)
+	addi		K, K, 32
+	vadd.vv		VTMP, VTMP, \w0
+	vsha2cl.vv	HGDC, FEBA, VTMP
+	vsha2ch.vv	FEBA, HGDC, VTMP
+.if !\last
+	vmerge.vvm	VTMP, \w2, \w1, MASK
+	vsha2ms.vv	\w0, VTMP, \w3
+.endif
+.endm
+
+.macro	sha512_16rounds	last
+	sha512_4rounds	\last, W0, W1, W2, W3
+	sha512_4rounds	\last, W1, W2, W3, W0
+	sha512_4rounds	\last, W2, W3, W0, W1
+	sha512_4rounds	\last, W3, W0, W1, W2
+.endm
+
+// void sha512_transform_zvknhb_zvkb(u64 state[8], const u8 *data,
+//				     int num_blocks);
+SYM_TYPED_FUNC_START(sha512_transform_zvknhb_zvkb)
+
+	// Setup mask for the vmerge to replace the first word (idx==0) in
+	// message scheduling.  There are 4 words, so an 8-bit mask suffices.
+	vsetivli	zero, 1, e8, m1, ta, ma
+	vmv.v.i		MASK, 0x01
+
+	// Load the state.  The state is stored as {a,b,c,d,e,f,g,h}, but we
+	// need {f,e,b,a},{h,g,d,c}.  The dst vtype is e64m2 and the index vtype
+	// is e8mf4.  We use index-load with the i8 indices {40, 32, 8, 0},
+	// loaded using the 32-bit little endian value 0x00082028.
+	li		t0, 0x00082028
+	vsetivli	zero, 1, e32, m1, ta, ma
+	vmv.v.x		INDICES, t0
+	addi		STATEP_C, STATEP, 16
+	vsetivli	zero, 4, e64, m2, ta, ma
+	vluxei8.v	FEBA, (STATEP), INDICES
+	vluxei8.v	HGDC, (STATEP_C), INDICES
+
+.Lnext_block:
+	la		K, K512
+	addi		NUM_BLOCKS, NUM_BLOCKS, -1
+
+	// Save the previous state, as it's needed later.
+	vmv.v.v		PREV_FEBA, FEBA
+	vmv.v.v		PREV_HGDC, HGDC
+
+	// Load the next 1024-bit message block and endian-swap each 64-bit word
+	vle64.v		W0, (DATA)
+	vrev8.v		W0, W0
+	addi		DATA, DATA, 32
+	vle64.v		W1, (DATA)
+	vrev8.v		W1, W1
+	addi		DATA, DATA, 32
+	vle64.v		W2, (DATA)
+	vrev8.v		W2, W2
+	addi		DATA, DATA, 32
+	vle64.v		W3, (DATA)
+	vrev8.v		W3, W3
+	addi		DATA, DATA, 32
+
+	// Do the 80 rounds of SHA-512.
+	sha512_16rounds 0
+	sha512_16rounds 0
+	sha512_16rounds 0
+	sha512_16rounds 0
+	sha512_16rounds 1
+
+	// Add the previous state.
+	vadd.vv		FEBA, FEBA, PREV_FEBA
+	vadd.vv		HGDC, HGDC, PREV_HGDC
+
+	// Repeat if more blocks remain.
+	bnez		NUM_BLOCKS, .Lnext_block
+
+	// Store the new state and return.
+	vsuxei8.v	FEBA, (STATEP), INDICES
+	vsuxei8.v	HGDC, (STATEP_C), INDICES
+	ret
+SYM_FUNC_END(sha512_transform_zvknhb_zvkb)
+
+.section ".rodata"
+.p2align 3
+.type K512, @object
+K512:
+	.dword		0x428a2f98d728ae22, 0x7137449123ef65cd
+	.dword		0xb5c0fbcfec4d3b2f, 0xe9b5dba58189dbbc
+	.dword		0x3956c25bf348b538, 0x59f111f1b605d019
+	.dword		0x923f82a4af194f9b, 0xab1c5ed5da6d8118
+	.dword		0xd807aa98a3030242, 0x12835b0145706fbe
+	.dword		0x243185be4ee4b28c, 0x550c7dc3d5ffb4e2
+	.dword		0x72be5d74f27b896f, 0x80deb1fe3b1696b1
+	.dword		0x9bdc06a725c71235, 0xc19bf174cf692694
+	.dword		0xe49b69c19ef14ad2, 0xefbe4786384f25e3
+	.dword		0x0fc19dc68b8cd5b5, 0x240ca1cc77ac9c65
+	.dword		0x2de92c6f592b0275, 0x4a7484aa6ea6e483
+	.dword		0x5cb0a9dcbd41fbd4, 0x76f988da831153b5
+	.dword		0x983e5152ee66dfab, 0xa831c66d2db43210
+	.dword		0xb00327c898fb213f, 0xbf597fc7beef0ee4
+	.dword		0xc6e00bf33da88fc2, 0xd5a79147930aa725
+	.dword		0x06ca6351e003826f, 0x142929670a0e6e70
+	.dword		0x27b70a8546d22ffc, 0x2e1b21385c26c926
+	.dword		0x4d2c6dfc5ac42aed, 0x53380d139d95b3df
+	.dword		0x650a73548baf63de, 0x766a0abb3c77b2a8
+	.dword		0x81c2c92e47edaee6, 0x92722c851482353b
+	.dword		0xa2bfe8a14cf10364, 0xa81a664bbc423001
+	.dword		0xc24b8b70d0f89791, 0xc76c51a30654be30
+	.dword		0xd192e819d6ef5218, 0xd69906245565a910
+	.dword		0xf40e35855771202a, 0x106aa07032bbd1b8
+	.dword		0x19a4c116b8d2d0c8, 0x1e376c085141ab53
+	.dword		0x2748774cdf8eeb99, 0x34b0bcb5e19b48a8
+	.dword		0x391c0cb3c5c95a63, 0x4ed8aa4ae3418acb
+	.dword		0x5b9cca4f7763e373, 0x682e6ff3d6b2b8a3
+	.dword		0x748f82ee5defb2fc, 0x78a5636f43172f60
+	.dword		0x84c87814a1f0ab72, 0x8cc702081a6439ec
+	.dword		0x90befffa23631e28, 0xa4506cebde82bde9
+	.dword		0xbef9a3f7b2c67915, 0xc67178f2e372532b
+	.dword		0xca273eceea26619c, 0xd186b8c721c0c207
+	.dword		0xeada7dd6cde0eb1e, 0xf57d4f7fee6ed178
+	.dword		0x06f067aa72176fba, 0x0a637dc5a2c898a6
+	.dword		0x113f9804bef90dae, 0x1b710b35131c471b
+	.dword		0x28db77f523047d84, 0x32caab7b40c72493
+	.dword		0x3c9ebe0a15c9bebc, 0x431d67c49c100d4c
+	.dword		0x4cc5d4becb3e42b6, 0x597f299cfc657e2a
+	.dword		0x5fcb6fab3ad6faec, 0x6c44198c4a475817
+.size K512, . - K512

arch/riscv/crypto/sm3-riscv64-glue.c

@@ -0,0 +1,112 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * SM3 using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2023 VRULL GmbH
+ * Author: Heiko Stuebner <heiko.stuebner@vrull.eu>
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/internal/hash.h>
+#include <crypto/internal/simd.h>
+#include <crypto/sm3_base.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+/*
+ * Note: the asm function only uses the 'state' field of struct sm3_state.
+ * It is assumed to be the first field.
+ */
+asmlinkage void sm3_transform_zvksh_zvkb(
+	struct sm3_state *state, const u8 *data, int num_blocks);
+
+static int riscv64_sm3_update(struct shash_desc *desc, const u8 *data,
+			      unsigned int len)
+{
+	/*
+	 * Ensure struct sm3_state begins directly with the SM3
+	 * 256-bit internal state, as this is what the asm function expects.
+	 */
+	BUILD_BUG_ON(offsetof(struct sm3_state, state) != 0);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		sm3_base_do_update(desc, data, len, sm3_transform_zvksh_zvkb);
+		kernel_vector_end();
+	} else {
+		sm3_update(shash_desc_ctx(desc), data, len);
+	}
+	return 0;
+}
+
+static int riscv64_sm3_finup(struct shash_desc *desc, const u8 *data,
+			     unsigned int len, u8 *out)
+{
+	struct sm3_state *ctx;
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		if (len)
+			sm3_base_do_update(desc, data, len,
+					   sm3_transform_zvksh_zvkb);
+		sm3_base_do_finalize(desc, sm3_transform_zvksh_zvkb);
+		kernel_vector_end();
+
+		return sm3_base_finish(desc, out);
+	}
+
+	ctx = shash_desc_ctx(desc);
+	if (len)
+		sm3_update(ctx, data, len);
+	sm3_final(ctx, out);
+
+	return 0;
+}
+
+static int riscv64_sm3_final(struct shash_desc *desc, u8 *out)
+{
+	return riscv64_sm3_finup(desc, NULL, 0, out);
+}
+
+static struct shash_alg riscv64_sm3_alg = {
+	.init = sm3_base_init,
+	.update = riscv64_sm3_update,
+	.final = riscv64_sm3_final,
+	.finup = riscv64_sm3_finup,
+	.descsize = sizeof(struct sm3_state),
+	.digestsize = SM3_DIGEST_SIZE,
+	.base = {
+		.cra_blocksize = SM3_BLOCK_SIZE,
+		.cra_priority = 300,
+		.cra_name = "sm3",
+		.cra_driver_name = "sm3-riscv64-zvksh-zvkb",
+		.cra_module = THIS_MODULE,
+	},
+};
+
+static int __init riscv64_sm3_mod_init(void)
+{
+	if (riscv_isa_extension_available(NULL, ZVKSH) &&
+	    riscv_isa_extension_available(NULL, ZVKB) &&
+	    riscv_vector_vlen() >= 128)
+		return crypto_register_shash(&riscv64_sm3_alg);
+
+	return -ENODEV;
+}
+
+static void __exit riscv64_sm3_mod_exit(void)
+{
+	crypto_unregister_shash(&riscv64_sm3_alg);
+}
+
+module_init(riscv64_sm3_mod_init);
+module_exit(riscv64_sm3_mod_exit);
+
+MODULE_DESCRIPTION("SM3 (RISC-V accelerated)");
+MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("sm3");

arch/riscv/crypto/sm3-riscv64-zvksh-zvkb.S

@@ -0,0 +1,123 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector SM3 Secure Hash extension ('Zvksh')
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/cfi_types.h>
+
+.text
+.option arch, +zvksh, +zvkb
+
+#define STATEP		a0
+#define DATA		a1
+#define NUM_BLOCKS	a2
+
+#define STATE		v0	// LMUL=2
+#define PREV_STATE	v2	// LMUL=2
+#define W0		v4	// LMUL=2
+#define W1		v6	// LMUL=2
+#define VTMP		v8	// LMUL=2
+
+.macro	sm3_8rounds	i, w0, w1
+	// Do 4 rounds using W_{0+i}..W_{7+i}.
+	vsm3c.vi	STATE, \w0, \i + 0
+	vslidedown.vi	VTMP, \w0, 2
+	vsm3c.vi	STATE, VTMP, \i + 1
+
+	// Compute W_{4+i}..W_{11+i}.
+	vslidedown.vi	VTMP, \w0, 4
+	vslideup.vi	VTMP, \w1, 4
+
+	// Do 4 rounds using W_{4+i}..W_{11+i}.
+	vsm3c.vi	STATE, VTMP, \i + 2
+	vslidedown.vi	VTMP, VTMP, 2
+	vsm3c.vi	STATE, VTMP, \i + 3
+
+.if \i < 28
+	// Compute W_{16+i}..W_{23+i}.
+	vsm3me.vv	\w0, \w1, \w0
+.endif
+	// For the next 8 rounds, w0 and w1 are swapped.
+.endm
+
+// void sm3_transform_zvksh_zvkb(u32 state[8], const u8 *data, int num_blocks);
+SYM_TYPED_FUNC_START(sm3_transform_zvksh_zvkb)
+
+	// Load the state and endian-swap each 32-bit word.
+	vsetivli	zero, 8, e32, m2, ta, ma
+	vle32.v		STATE, (STATEP)
+	vrev8.v		STATE, STATE
+
+.Lnext_block:
+	addi		NUM_BLOCKS, NUM_BLOCKS, -1
+
+	// Save the previous state, as it's needed later.
+	vmv.v.v		PREV_STATE, STATE
+
+	// Load the next 512-bit message block into W0-W1.
+	vle32.v		W0, (DATA)
+	addi		DATA, DATA, 32
+	vle32.v		W1, (DATA)
+	addi		DATA, DATA, 32
+
+	// Do the 64 rounds of SM3.
+	sm3_8rounds	0, W0, W1
+	sm3_8rounds	4, W1, W0
+	sm3_8rounds	8, W0, W1
+	sm3_8rounds	12, W1, W0
+	sm3_8rounds	16, W0, W1
+	sm3_8rounds	20, W1, W0
+	sm3_8rounds	24, W0, W1
+	sm3_8rounds	28, W1, W0
+
+	// XOR in the previous state.
+	vxor.vv		STATE, STATE, PREV_STATE
+
+	// Repeat if more blocks remain.
+	bnez		NUM_BLOCKS, .Lnext_block
+
+	// Store the new state and return.
+	vrev8.v		STATE, STATE
+	vse32.v		STATE, (STATEP)
+	ret
+SYM_FUNC_END(sm3_transform_zvksh_zvkb)

arch/riscv/crypto/sm4-riscv64-glue.c

@@ -0,0 +1,107 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * SM4 using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2023 VRULL GmbH
+ * Author: Heiko Stuebner <heiko.stuebner@vrull.eu>
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih@sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/internal/cipher.h>
+#include <crypto/internal/simd.h>
+#include <crypto/sm4.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+asmlinkage void sm4_expandkey_zvksed_zvkb(const u8 user_key[SM4_KEY_SIZE],
+					  u32 rkey_enc[SM4_RKEY_WORDS],
+					  u32 rkey_dec[SM4_RKEY_WORDS]);
+asmlinkage void sm4_crypt_zvksed_zvkb(const u32 rkey[SM4_RKEY_WORDS],
+				      const u8 in[SM4_BLOCK_SIZE],
+				      u8 out[SM4_BLOCK_SIZE]);
+
+static int riscv64_sm4_setkey(struct crypto_tfm *tfm, const u8 *key,
+			      unsigned int keylen)
+{
+	struct sm4_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	if (crypto_simd_usable()) {
+		if (keylen != SM4_KEY_SIZE)
+			return -EINVAL;
+		kernel_vector_begin();
+		sm4_expandkey_zvksed_zvkb(key, ctx->rkey_enc, ctx->rkey_dec);
+		kernel_vector_end();
+		return 0;
+	}
+	return sm4_expandkey(ctx, key, keylen);
+}
+
+static void riscv64_sm4_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	const struct sm4_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		sm4_crypt_zvksed_zvkb(ctx->rkey_enc, src, dst);
+		kernel_vector_end();
+	} else {
+		sm4_crypt_block(ctx->rkey_enc, dst, src);
+	}
+}
+
+static void riscv64_sm4_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	const struct sm4_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	if (crypto_simd_usable()) {
+		kernel_vector_begin();
+		sm4_crypt_zvksed_zvkb(ctx->rkey_dec, src, dst);
+		kernel_vector_end();
+	} else {
+		sm4_crypt_block(ctx->rkey_dec, dst, src);
+	}
+}
+
+static struct crypto_alg riscv64_sm4_alg = {
+	.cra_flags = CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize = SM4_BLOCK_SIZE,
+	.cra_ctxsize = sizeof(struct sm4_ctx),
+	.cra_priority = 300,
+	.cra_name = "sm4",
+	.cra_driver_name = "sm4-riscv64-zvksed-zvkb",
+	.cra_cipher = {
+		.cia_min_keysize = SM4_KEY_SIZE,
+		.cia_max_keysize = SM4_KEY_SIZE,
+		.cia_setkey = riscv64_sm4_setkey,
+		.cia_encrypt = riscv64_sm4_encrypt,
+		.cia_decrypt = riscv64_sm4_decrypt,
+	},
+	.cra_module = THIS_MODULE,
+};
+
+static int __init riscv64_sm4_mod_init(void)
+{
+	if (riscv_isa_extension_available(NULL, ZVKSED) &&
+	    riscv_isa_extension_available(NULL, ZVKB) &&
+	    riscv_vector_vlen() >= 128)
+		return crypto_register_alg(&riscv64_sm4_alg);
+
+	return -ENODEV;
+}
+
+static void __exit riscv64_sm4_mod_exit(void)
+{
+	crypto_unregister_alg(&riscv64_sm4_alg);
+}
+
+module_init(riscv64_sm4_mod_init);
+module_exit(riscv64_sm4_mod_exit);
+
+MODULE_DESCRIPTION("SM4 (RISC-V accelerated)");
+MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("sm4");

arch/riscv/crypto/sm4-riscv64-zvksed-zvkb.S

@@ -0,0 +1,117 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Christoph Müllner <christoph.muellner@vrull.eu>
+// Copyright (c) 2023, Jerry Shih <jerry.shih@sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector SM4 Block Cipher extension ('Zvksed')
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvksed, +zvkb
+
+// void sm4_expandkey_zksed_zvkb(const u8 user_key[16], u32 rkey_enc[32],
+//				 u32 rkey_dec[32]);
+SYM_FUNC_START(sm4_expandkey_zvksed_zvkb)
+	vsetivli	zero, 4, e32, m1, ta, ma
+
+	// Load the user key.
+	vle32.v		v1, (a0)
+	vrev8.v		v1, v1
+
+	// XOR the user key with the family key.
+	la		t0, FAMILY_KEY
+	vle32.v		v2, (t0)
+	vxor.vv		v1, v1, v2
+
+	// Compute the round keys.  Store them in forwards order in rkey_enc
+	// and in reverse order in rkey_dec.
+	addi		a2, a2, 31*4
+	li		t0, -4
+	.set		i, 0
+.rept 8
+	vsm4k.vi	v1, v1, i
+	vse32.v		v1, (a1)	// Store to rkey_enc.
+	vsse32.v	v1, (a2), t0	// Store to rkey_dec.
+.if i < 7
+	addi		a1, a1, 16
+	addi		a2, a2, -16
+.endif
+	.set		i, i + 1
+.endr
+
+	ret
+SYM_FUNC_END(sm4_expandkey_zvksed_zvkb)
+
+// void sm4_crypt_zvksed_zvkb(const u32 rkey[32], const u8 in[16], u8 out[16]);
+SYM_FUNC_START(sm4_crypt_zvksed_zvkb)
+	vsetivli	zero, 4, e32, m1, ta, ma
+
+	// Load the input data.
+	vle32.v		v1, (a1)
+	vrev8.v		v1, v1
+
+	// Do the 32 rounds of SM4, 4 at a time.
+	.set		i, 0
+.rept 8
+	vle32.v		v2, (a0)
+	vsm4r.vs	v1, v2
+.if i < 7
+	addi		a0, a0, 16
+.endif
+	.set		i, i + 1
+.endr
+
+	// Store the output data (in reverse element order).
+	vrev8.v		v1, v1
+	li		t0, -4
+	addi		a2, a2, 12
+	vsse32.v	v1, (a2), t0
+
+	ret
+SYM_FUNC_END(sm4_crypt_zvksed_zvkb)
+
+.section ".rodata"
+.p2align 2
+.type FAMILY_KEY, @object
+FAMILY_KEY:
+	.word 0xA3B1BAC6, 0x56AA3350, 0x677D9197, 0xB27022DC
+.size FAMILY_KEY, . - FAMILY_KEY

arch/riscv/errata/andes/errata.c

@@ -18,9 +18,9 @@
 #include <asm/sbi.h>
 #include <asm/vendorid_list.h>
 
-#define ANDESTECH_AX45MP_MARCHID	0x8000000000008a45UL
-#define ANDESTECH_AX45MP_MIMPID		0x500UL
-#define ANDESTECH_SBI_EXT_ANDES		0x0900031E
+#define ANDES_AX45MP_MARCHID		0x8000000000008a45UL
+#define ANDES_AX45MP_MIMPID		0x500UL
+#define ANDES_SBI_EXT_ANDES		0x0900031E
 
 #define ANDES_SBI_EXT_IOCP_SW_WORKAROUND	1
 
@@ -32,7 +32,7 @@ static long ax45mp_iocp_sw_workaround(void)
 	 * ANDES_SBI_EXT_IOCP_SW_WORKAROUND SBI EXT checks if the IOCP is missing and
 	 * cache is controllable only then CMO will be applied to the platform.
 	 */
-	ret = sbi_ecall(ANDESTECH_SBI_EXT_ANDES, ANDES_SBI_EXT_IOCP_SW_WORKAROUND,
+	ret = sbi_ecall(ANDES_SBI_EXT_ANDES, ANDES_SBI_EXT_IOCP_SW_WORKAROUND,
 			0, 0, 0, 0, 0, 0);
 
 	return ret.error ? 0 : ret.value;
@@ -50,7 +50,7 @@ static void errata_probe_iocp(unsigned int stage, unsigned long arch_id, unsigne
 
 	done = true;
 
-	if (arch_id != ANDESTECH_AX45MP_MARCHID || impid != ANDESTECH_AX45MP_MIMPID)
+	if (arch_id != ANDES_AX45MP_MARCHID || impid != ANDES_AX45MP_MIMPID)
 		return;
 
 	if (!ax45mp_iocp_sw_workaround())

arch/riscv/include/asm/asm.h

@@ -183,6 +183,16 @@
 	REG_L x31, PT_T6(sp)
 	.endm
 
+/* Annotate a function as being unsuitable for kprobes. */
+#ifdef CONFIG_KPROBES
+#define ASM_NOKPROBE(name)				\
+	.pushsection "_kprobe_blacklist", "aw";		\
+	RISCV_PTR name;					\
+	.popsection
+#else
+#define ASM_NOKPROBE(name)
+#endif
+
 #endif /* __ASSEMBLY__ */
 
 #endif /* _ASM_RISCV_ASM_H */

arch/riscv/include/asm/atomic.h

@@ -17,7 +17,6 @@
 #endif
 
 #include <asm/cmpxchg.h>
-#include <asm/barrier.h>
 
 #define __atomic_acquire_fence()					\
 	__asm__ __volatile__(RISCV_ACQUIRE_BARRIER "" ::: "memory")
@@ -207,7 +206,7 @@ static __always_inline int arch_atomic_fetch_add_unless(atomic_t *v, int a, int
 		"	add      %[rc], %[p], %[a]\n"
 		"	sc.w.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		: [a]"r" (a), [u]"r" (u)
@@ -228,7 +227,7 @@ static __always_inline s64 arch_atomic64_fetch_add_unless(atomic64_t *v, s64 a,
 		"	add      %[rc], %[p], %[a]\n"
 		"	sc.d.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		: [a]"r" (a), [u]"r" (u)
@@ -248,7 +247,7 @@ static __always_inline bool arch_atomic_inc_unless_negative(atomic_t *v)
 		"	addi      %[rc], %[p], 1\n"
 		"	sc.w.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -268,7 +267,7 @@ static __always_inline bool arch_atomic_dec_unless_positive(atomic_t *v)
 		"	addi      %[rc], %[p], -1\n"
 		"	sc.w.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -288,7 +287,7 @@ static __always_inline int arch_atomic_dec_if_positive(atomic_t *v)
 		"	bltz     %[rc], 1f\n"
 		"	sc.w.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -310,7 +309,7 @@ static __always_inline bool arch_atomic64_inc_unless_negative(atomic64_t *v)
 		"	addi      %[rc], %[p], 1\n"
 		"	sc.d.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -331,7 +330,7 @@ static __always_inline bool arch_atomic64_dec_unless_positive(atomic64_t *v)
 		"	addi      %[rc], %[p], -1\n"
 		"	sc.d.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -352,7 +351,7 @@ static __always_inline s64 arch_atomic64_dec_if_positive(atomic64_t *v)
 		"	bltz     %[rc], 1f\n"
 		"	sc.d.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:

arch/riscv/include/asm/barrier.h

@@ -11,28 +11,27 @@
 #define _ASM_RISCV_BARRIER_H
 
 #ifndef __ASSEMBLY__
+#include <asm/fence.h>
 
 #define nop()		__asm__ __volatile__ ("nop")
 #define __nops(n)	".rept	" #n "\nnop\n.endr\n"
 #define nops(n)		__asm__ __volatile__ (__nops(n))
 
-#define RISCV_FENCE(p, s) \
-	__asm__ __volatile__ ("fence " #p "," #s : : : "memory")
 
 /* These barriers need to enforce ordering on both devices or memory. */
-#define mb()		RISCV_FENCE(iorw,iorw)
-#define rmb()		RISCV_FENCE(ir,ir)
-#define wmb()		RISCV_FENCE(ow,ow)
+#define __mb()		RISCV_FENCE(iorw, iorw)
+#define __rmb()		RISCV_FENCE(ir, ir)
+#define __wmb()		RISCV_FENCE(ow, ow)
 
 /* These barriers do not need to enforce ordering on devices, just memory. */
-#define __smp_mb()	RISCV_FENCE(rw,rw)
-#define __smp_rmb()	RISCV_FENCE(r,r)
-#define __smp_wmb()	RISCV_FENCE(w,w)
+#define __smp_mb()	RISCV_FENCE(rw, rw)
+#define __smp_rmb()	RISCV_FENCE(r, r)
+#define __smp_wmb()	RISCV_FENCE(w, w)
 
 #define __smp_store_release(p, v)					\
 do {									\
 	compiletime_assert_atomic_type(*p);				\
-	RISCV_FENCE(rw,w);						\
+	RISCV_FENCE(rw, w);						\
 	WRITE_ONCE(*p, v);						\
 } while (0)
 
@@ -40,7 +39,7 @@ do {									\
 ({									\
 	typeof(*p) ___p1 = READ_ONCE(*p);				\
 	compiletime_assert_atomic_type(*p);				\
-	RISCV_FENCE(r,rw);						\
+	RISCV_FENCE(r, rw);						\
 	___p1;								\
 })
 
@@ -69,7 +68,7 @@ do {									\
  * instances the scheduler pairs this with an mb(), so nothing is necessary on
  * the new hart.
  */
-#define smp_mb__after_spinlock()	RISCV_FENCE(iorw,iorw)
+#define smp_mb__after_spinlock()	RISCV_FENCE(iorw, iorw)
 
 #include <asm-generic/barrier.h>
 

arch/riscv/include/asm/bitops.h

@@ -22,6 +22,16 @@
 #include <asm-generic/bitops/fls.h>
 
 #else
+#define __HAVE_ARCH___FFS
+#define __HAVE_ARCH___FLS
+#define __HAVE_ARCH_FFS
+#define __HAVE_ARCH_FLS
+
+#include <asm-generic/bitops/__ffs.h>
+#include <asm-generic/bitops/__fls.h>
+#include <asm-generic/bitops/ffs.h>
+#include <asm-generic/bitops/fls.h>
+
 #include <asm/alternative-macros.h>
 #include <asm/hwcap.h>
 
@@ -37,8 +47,6 @@
 
 static __always_inline unsigned long variable__ffs(unsigned long word)
 {
-	int num;
-
 	asm goto(ALTERNATIVE("j %l[legacy]", "nop", 0,
 				      RISCV_ISA_EXT_ZBB, 1)
 			  : : : : legacy);
@@ -52,32 +60,7 @@ static __always_inline unsigned long variable__ffs(unsigned long word)
 	return word;
 
 legacy:
-	num = 0;
-#if BITS_PER_LONG == 64
-	if ((word & 0xffffffff) == 0) {
-		num += 32;
-		word >>= 32;
-	}
-#endif
-	if ((word & 0xffff) == 0) {
-		num += 16;
-		word >>= 16;
-	}
-	if ((word & 0xff) == 0) {
-		num += 8;
-		word >>= 8;
-	}
-	if ((word & 0xf) == 0) {
-		num += 4;
-		word >>= 4;
-	}
-	if ((word & 0x3) == 0) {
-		num += 2;
-		word >>= 2;
-	}
-	if ((word & 0x1) == 0)
-		num += 1;
-	return num;
+	return generic___ffs(word);
 }
 
 /**
@@ -93,8 +76,6 @@ static __always_inline unsigned long variable__ffs(unsigned long word)
 
 static __always_inline unsigned long variable__fls(unsigned long word)
 {
-	int num;
-
 	asm goto(ALTERNATIVE("j %l[legacy]", "nop", 0,
 				      RISCV_ISA_EXT_ZBB, 1)
 			  : : : : legacy);
@@ -108,32 +89,7 @@ static __always_inline unsigned long variable__fls(unsigned long word)
 	return BITS_PER_LONG - 1 - word;
 
 legacy:
-	num = BITS_PER_LONG - 1;
-#if BITS_PER_LONG == 64
-	if (!(word & (~0ul << 32))) {
-		num -= 32;
-		word <<= 32;
-	}
-#endif
-	if (!(word & (~0ul << (BITS_PER_LONG - 16)))) {
-		num -= 16;
-		word <<= 16;
-	}
-	if (!(word & (~0ul << (BITS_PER_LONG - 8)))) {
-		num -= 8;
-		word <<= 8;
-	}
-	if (!(word & (~0ul << (BITS_PER_LONG - 4)))) {
-		num -= 4;
-		word <<= 4;
-	}
-	if (!(word & (~0ul << (BITS_PER_LONG - 2)))) {
-		num -= 2;
-		word <<= 2;
-	}
-	if (!(word & (~0ul << (BITS_PER_LONG - 1))))
-		num -= 1;
-	return num;
+	return generic___fls(word);
 }
 
 /**
@@ -149,46 +105,23 @@ static __always_inline unsigned long variable__fls(unsigned long word)
 
 static __always_inline int variable_ffs(int x)
 {
-	int r;
-
-	if (!x)
-		return 0;
-
 	asm goto(ALTERNATIVE("j %l[legacy]", "nop", 0,
 				      RISCV_ISA_EXT_ZBB, 1)
 			  : : : : legacy);
 
+	if (!x)
+		return 0;
+
 	asm volatile (".option push\n"
 		      ".option arch,+zbb\n"
 		      CTZW "%0, %1\n"
 		      ".option pop\n"
-		      : "=r" (r) : "r" (x) :);
+		      : "=r" (x) : "r" (x) :);
 
-	return r + 1;
+	return x + 1;
 
 legacy:
-	r = 1;
-	if (!(x & 0xffff)) {
-		x >>= 16;
-		r += 16;
-	}
-	if (!(x & 0xff)) {
-		x >>= 8;
-		r += 8;
-	}
-	if (!(x & 0xf)) {
-		x >>= 4;
-		r += 4;
-	}
-	if (!(x & 3)) {
-		x >>= 2;
-		r += 2;
-	}
-	if (!(x & 1)) {
-		x >>= 1;
-		r += 1;
-	}
-	return r;
+	return generic_ffs(x);
 }
 
 /**
@@ -204,46 +137,23 @@ static __always_inline int variable_ffs(int x)
 
 static __always_inline int variable_fls(unsigned int x)
 {
-	int r;
-
-	if (!x)
-		return 0;
-
 	asm goto(ALTERNATIVE("j %l[legacy]", "nop", 0,
 				      RISCV_ISA_EXT_ZBB, 1)
 			  : : : : legacy);
 
+	if (!x)
+		return 0;
+
 	asm volatile (".option push\n"
 		      ".option arch,+zbb\n"
 		      CLZW "%0, %1\n"
 		      ".option pop\n"
-		      : "=r" (r) : "r" (x) :);
+		      : "=r" (x) : "r" (x) :);
 
-	return 32 - r;
+	return 32 - x;
 
 legacy:
-	r = 32;
-	if (!(x & 0xffff0000u)) {
-		x <<= 16;
-		r -= 16;
-	}
-	if (!(x & 0xff000000u)) {
-		x <<= 8;
-		r -= 8;
-	}
-	if (!(x & 0xf0000000u)) {
-		x <<= 4;
-		r -= 4;
-	}
-	if (!(x & 0xc0000000u)) {
-		x <<= 2;
-		r -= 2;
-	}
-	if (!(x & 0x80000000u)) {
-		x <<= 1;
-		r -= 1;
-	}
-	return r;
+	return generic_fls(x);
 }
 
 /**

arch/riscv/include/asm/cmpxchg.h

@@ -8,7 +8,6 @@
 
 #include <linux/bug.h>
 
-#include <asm/barrier.h>
 #include <asm/fence.h>
 
 #define __xchg_relaxed(ptr, new, size)					\
@@ -313,7 +312,7 @@
 			"	bne  %0, %z3, 1f\n"			\
 			"	sc.w.rl %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
-			"	fence rw, rw\n"				\
+			RISCV_FULL_BARRIER				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" ((long)__old), "rJ" (__new)		\
@@ -325,7 +324,7 @@
 			"	bne %0, %z3, 1f\n"			\
 			"	sc.d.rl %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
-			"	fence rw, rw\n"				\
+			RISCV_FULL_BARRIER				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" (__old), "rJ" (__new)			\

arch/riscv/include/asm/compat.h

@@ -14,9 +14,28 @@
 
 static inline int is_compat_task(void)
 {
+	if (!IS_ENABLED(CONFIG_COMPAT))
+		return 0;
+
 	return test_thread_flag(TIF_32BIT);
 }
 
+static inline int is_compat_thread(struct thread_info *thread)
+{
+	if (!IS_ENABLED(CONFIG_COMPAT))
+		return 0;
+
+	return test_ti_thread_flag(thread, TIF_32BIT);
+}
+
+static inline void set_compat_task(bool is_compat)
+{
+	if (is_compat)
+		set_thread_flag(TIF_32BIT);
+	else
+		clear_thread_flag(TIF_32BIT);
+}
+
 struct compat_user_regs_struct {
 	compat_ulong_t pc;
 	compat_ulong_t ra;

arch/riscv/include/asm/cpufeature.h

@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 /*
- * Copyright 2022-2023 Rivos, Inc
+ * Copyright 2022-2024 Rivos, Inc
  */
 
 #ifndef _ASM_CPUFEATURE_H
@@ -28,29 +28,38 @@ struct riscv_isainfo {
 
 DECLARE_PER_CPU(struct riscv_cpuinfo, riscv_cpuinfo);
 
-DECLARE_PER_CPU(long, misaligned_access_speed);
-
 /* Per-cpu ISA extensions. */
 extern struct riscv_isainfo hart_isa[NR_CPUS];
 
 void riscv_user_isa_enable(void);
 
-#ifdef CONFIG_RISCV_MISALIGNED
-bool unaligned_ctl_available(void);
-bool check_unaligned_access_emulated(int cpu);
+#if defined(CONFIG_RISCV_MISALIGNED)
+bool check_unaligned_access_emulated_all_cpus(void);
 void unaligned_emulation_finish(void);
+bool unaligned_ctl_available(void);
+DECLARE_PER_CPU(long, misaligned_access_speed);
 #else
 static inline bool unaligned_ctl_available(void)
 {
 	return false;
 }
+#endif
 
-static inline bool check_unaligned_access_emulated(int cpu)
+#if defined(CONFIG_RISCV_PROBE_UNALIGNED_ACCESS)
+DECLARE_STATIC_KEY_FALSE(fast_unaligned_access_speed_key);
+
+static __always_inline bool has_fast_unaligned_accesses(void)
 {
-	return false;
+	return static_branch_likely(&fast_unaligned_access_speed_key);
+}
+#else
+static __always_inline bool has_fast_unaligned_accesses(void)
+{
+	if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS))
+		return true;
+	else
+		return false;
 }
-
-static inline void unaligned_emulation_finish(void) {}
 #endif
 
 unsigned long riscv_get_elf_hwcap(void);
@@ -135,6 +144,4 @@ static __always_inline bool riscv_cpu_has_extension_unlikely(int cpu, const unsi
 	return __riscv_isa_extension_available(hart_isa[cpu].isa, ext);
 }
 
-DECLARE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
-
 #endif

arch/riscv/include/asm/elf.h

@@ -53,13 +53,9 @@ extern bool compat_elf_check_arch(Elf32_Ehdr *hdr);
 #define ELF_ET_DYN_BASE		((DEFAULT_MAP_WINDOW / 3) * 2)
 
 #ifdef CONFIG_64BIT
-#ifdef CONFIG_COMPAT
-#define STACK_RND_MASK		(test_thread_flag(TIF_32BIT) ? \
+#define STACK_RND_MASK		(is_compat_task() ? \
 				 0x7ff >> (PAGE_SHIFT - 12) : \
 				 0x3ffff >> (PAGE_SHIFT - 12))
-#else
-#define STACK_RND_MASK		(0x3ffff >> (PAGE_SHIFT - 12))
-#endif
 #endif
 
 /*
@@ -139,10 +135,7 @@ do {							\
 #ifdef CONFIG_COMPAT
 
 #define SET_PERSONALITY(ex)					\
-do {    if ((ex).e_ident[EI_CLASS] == ELFCLASS32)		\
-		set_thread_flag(TIF_32BIT);			\
-	else							\
-		clear_thread_flag(TIF_32BIT);			\
+do {	set_compat_task((ex).e_ident[EI_CLASS] == ELFCLASS32);	\
 	if (personality(current->personality) != PER_LINUX32)	\
 		set_personality(PER_LINUX |			\
 			(current->personality & (~PER_MASK)));	\

arch/riscv/include/asm/errata_list.h

@@ -12,8 +12,8 @@
 #include <asm/vendorid_list.h>
 
 #ifdef CONFIG_ERRATA_ANDES
-#define ERRATA_ANDESTECH_NO_IOCP	0
-#define ERRATA_ANDESTECH_NUMBER		1
+#define ERRATA_ANDES_NO_IOCP 0
+#define ERRATA_ANDES_NUMBER 1
 #endif
 
 #ifdef CONFIG_ERRATA_SIFIVE
@@ -112,15 +112,6 @@ asm volatile(ALTERNATIVE(						\
 #define THEAD_C9XX_RV_IRQ_PMU			17
 #define THEAD_C9XX_CSR_SCOUNTEROF		0x5c5
 
-#define ALT_SBI_PMU_OVERFLOW(__ovl)					\
-asm volatile(ALTERNATIVE(						\
-	"csrr %0, " __stringify(CSR_SSCOUNTOVF),			\
-	"csrr %0, " __stringify(THEAD_C9XX_CSR_SCOUNTEROF),		\
-		THEAD_VENDOR_ID, ERRATA_THEAD_PMU,			\
-		CONFIG_ERRATA_THEAD_PMU)				\
-	: "=r" (__ovl) :						\
-	: "memory")
-
 #endif /* __ASSEMBLY__ */
 
 #endif

arch/riscv/include/asm/fence.h

@@ -1,12 +1,18 @@
 #ifndef _ASM_RISCV_FENCE_H
 #define _ASM_RISCV_FENCE_H
 
+#define RISCV_FENCE_ASM(p, s)		"\tfence " #p "," #s "\n"
+#define RISCV_FENCE(p, s) \
+	({ __asm__ __volatile__ (RISCV_FENCE_ASM(p, s) : : : "memory"); })
+
 #ifdef CONFIG_SMP
-#define RISCV_ACQUIRE_BARRIER		"\tfence r , rw\n"
-#define RISCV_RELEASE_BARRIER		"\tfence rw,  w\n"
+#define RISCV_ACQUIRE_BARRIER		RISCV_FENCE_ASM(r, rw)
+#define RISCV_RELEASE_BARRIER		RISCV_FENCE_ASM(rw, w)
+#define RISCV_FULL_BARRIER		RISCV_FENCE_ASM(rw, rw)
 #else
 #define RISCV_ACQUIRE_BARRIER
 #define RISCV_RELEASE_BARRIER
+#define RISCV_FULL_BARRIER
 #endif
 
 #endif	/* _ASM_RISCV_FENCE_H */

arch/riscv/include/asm/hwcap.h

@@ -80,6 +80,7 @@
 #define RISCV_ISA_EXT_ZFA		71
 #define RISCV_ISA_EXT_ZTSO		72
 #define RISCV_ISA_EXT_ZACAS		73
+#define RISCV_ISA_EXT_XANDESPMU		74
 
 #define RISCV_ISA_EXT_XLINUXENVCFG	127
 

arch/riscv/include/asm/io.h

@@ -47,10 +47,10 @@
  * sufficient to ensure this works sanely on controllers that support I/O
  * writes.
  */
-#define __io_pbr()	__asm__ __volatile__ ("fence io,i"  : : : "memory");
-#define __io_par(v)	__asm__ __volatile__ ("fence i,ior" : : : "memory");
-#define __io_pbw()	__asm__ __volatile__ ("fence iow,o" : : : "memory");
-#define __io_paw()	__asm__ __volatile__ ("fence o,io"  : : : "memory");
+#define __io_pbr()	RISCV_FENCE(io, i)
+#define __io_par(v)	RISCV_FENCE(i, ior)
+#define __io_pbw()	RISCV_FENCE(iow, o)
+#define __io_paw()	RISCV_FENCE(o, io)
 
 /*
  * Accesses from a single hart to a single I/O address must be ordered.  This

arch/riscv/include/asm/membarrier.h

@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef _ASM_RISCV_MEMBARRIER_H
+#define _ASM_RISCV_MEMBARRIER_H
+
+static inline void membarrier_arch_switch_mm(struct mm_struct *prev,
+					     struct mm_struct *next,
+					     struct task_struct *tsk)
+{
+	/*
+	 * Only need the full barrier when switching between processes.
+	 * Barrier when switching from kernel to userspace is not
+	 * required here, given that it is implied by mmdrop(). Barrier
+	 * when switching from userspace to kernel is not needed after
+	 * store to rq->curr.
+	 */
+	if (IS_ENABLED(CONFIG_SMP) &&
+	    likely(!(atomic_read(&next->membarrier_state) &
+		     (MEMBARRIER_STATE_PRIVATE_EXPEDITED |
+		      MEMBARRIER_STATE_GLOBAL_EXPEDITED)) || !prev))
+		return;
+
+	/*
+	 * The membarrier system call requires a full memory barrier
+	 * after storing to rq->curr, before going back to user-space.
+	 *
+	 * This barrier is also needed for the SYNC_CORE command when
+	 * switching between processes; in particular, on a transition
+	 * from a thread belonging to another mm to a thread belonging
+	 * to the mm for which a membarrier SYNC_CORE is done on CPU0:
+	 *
+	 *   - [CPU0] sets all bits in the mm icache_stale_mask (in
+	 *     prepare_sync_core_cmd());
+	 *
+	 *   - [CPU1] stores to rq->curr (by the scheduler);
+	 *
+	 *   - [CPU0] loads rq->curr within membarrier and observes
+	 *     cpu_rq(1)->curr->mm != mm, so the IPI is skipped on
+	 *     CPU1; this means membarrier relies on switch_mm() to
+	 *     issue the sync-core;
+	 *
+	 *   - [CPU1] switch_mm() loads icache_stale_mask; if the bit
+	 *     is zero, switch_mm() may incorrectly skip the sync-core.
+	 *
+	 * Matches a full barrier in the proximity of the membarrier
+	 * system call entry.
+	 */
+	smp_mb();
+}
+
+#endif /* _ASM_RISCV_MEMBARRIER_H */

arch/riscv/include/asm/mmio.h

@@ -12,6 +12,7 @@
 #define _ASM_RISCV_MMIO_H
 
 #include <linux/types.h>
+#include <asm/fence.h>
 #include <asm/mmiowb.h>
 
 /* Generic IO read/write.  These perform native-endian accesses. */
@@ -131,8 +132,8 @@ static inline u64 __raw_readq(const volatile void __iomem *addr)
  * doesn't define any ordering between the memory space and the I/O space.
  */
 #define __io_br()	do {} while (0)
-#define __io_ar(v)	({ __asm__ __volatile__ ("fence i,ir" : : : "memory"); })
-#define __io_bw()	({ __asm__ __volatile__ ("fence w,o" : : : "memory"); })
+#define __io_ar(v)	RISCV_FENCE(i, ir)
+#define __io_bw()	RISCV_FENCE(w, o)
 #define __io_aw()	mmiowb_set_pending()
 
 #define readb(c)	({ u8  __v; __io_br(); __v = readb_cpu(c); __io_ar(__v); __v; })

arch/riscv/include/asm/mmiowb.h

@@ -7,7 +7,7 @@
  * "o,w" is sufficient to ensure that all writes to the device have completed
  * before the write to the spinlock is allowed to commit.
  */
-#define mmiowb()	__asm__ __volatile__ ("fence o,w" : : : "memory");
+#define mmiowb()	RISCV_FENCE(o, w)
 
 #include <linux/smp.h>
 #include <asm-generic/mmiowb.h>

arch/riscv/include/asm/pgalloc.h

@@ -95,13 +95,19 @@ static inline void pud_free(struct mm_struct *mm, pud_t *pud)
 		__pud_free(mm, pud);
 }
 
-#define __pud_free_tlb(tlb, pud, addr)					\
-do {									\
-	if (pgtable_l4_enabled) {					\
-		pagetable_pud_dtor(virt_to_ptdesc(pud));		\
-		tlb_remove_page_ptdesc((tlb), virt_to_ptdesc(pud));	\
-	}								\
-} while (0)
+static inline void __pud_free_tlb(struct mmu_gather *tlb, pud_t *pud,
+				  unsigned long addr)
+{
+	if (pgtable_l4_enabled) {
+		struct ptdesc *ptdesc = virt_to_ptdesc(pud);
+
+		pagetable_pud_dtor(ptdesc);
+		if (riscv_use_ipi_for_rfence())
+			tlb_remove_page_ptdesc(tlb, ptdesc);
+		else
+			tlb_remove_ptdesc(tlb, ptdesc);
+	}
+}
 
 #define p4d_alloc_one p4d_alloc_one
 static inline p4d_t *p4d_alloc_one(struct mm_struct *mm, unsigned long addr)
@@ -130,11 +136,16 @@ static inline void p4d_free(struct mm_struct *mm, p4d_t *p4d)
 		__p4d_free(mm, p4d);
 }
 
-#define __p4d_free_tlb(tlb, p4d, addr)					\
-do {									\
-	if (pgtable_l5_enabled)						\
-		tlb_remove_page_ptdesc((tlb), virt_to_ptdesc(p4d));	\
-} while (0)
+static inline void __p4d_free_tlb(struct mmu_gather *tlb, p4d_t *p4d,
+				  unsigned long addr)
+{
+	if (pgtable_l5_enabled) {
+		if (riscv_use_ipi_for_rfence())
+			tlb_remove_page_ptdesc(tlb, virt_to_ptdesc(p4d));
+		else
+			tlb_remove_ptdesc(tlb, virt_to_ptdesc(p4d));
+	}
+}
 #endif /* __PAGETABLE_PMD_FOLDED */
 
 static inline void sync_kernel_mappings(pgd_t *pgd)
@@ -159,19 +170,31 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm)
 
 #ifndef __PAGETABLE_PMD_FOLDED
 
-#define __pmd_free_tlb(tlb, pmd, addr)				\
-do {								\
-	pagetable_pmd_dtor(virt_to_ptdesc(pmd));		\
-	tlb_remove_page_ptdesc((tlb), virt_to_ptdesc(pmd));	\
-} while (0)
+static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd,
+				  unsigned long addr)
+{
+	struct ptdesc *ptdesc = virt_to_ptdesc(pmd);
+
+	pagetable_pmd_dtor(ptdesc);
+	if (riscv_use_ipi_for_rfence())
+		tlb_remove_page_ptdesc(tlb, ptdesc);
+	else
+		tlb_remove_ptdesc(tlb, ptdesc);
+}
 
 #endif /* __PAGETABLE_PMD_FOLDED */
 
-#define __pte_free_tlb(tlb, pte, buf)			\
-do {							\
-	pagetable_pte_dtor(page_ptdesc(pte));		\
-	tlb_remove_page_ptdesc((tlb), page_ptdesc(pte));\
-} while (0)
+static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
+				  unsigned long addr)
+{
+	struct ptdesc *ptdesc = page_ptdesc(pte);
+
+	pagetable_pte_dtor(ptdesc);
+	if (riscv_use_ipi_for_rfence())
+		tlb_remove_page_ptdesc(tlb, ptdesc);
+	else
+		tlb_remove_ptdesc(tlb, ptdesc);
+}
 #endif /* CONFIG_MMU */
 
 #endif /* _ASM_RISCV_PGALLOC_H */

arch/riscv/include/asm/pgtable.h

@@ -127,16 +127,10 @@
 #define VA_USER_SV48 (UL(1) << (VA_BITS_SV48 - 1))
 #define VA_USER_SV57 (UL(1) << (VA_BITS_SV57 - 1))
 
-#ifdef CONFIG_COMPAT
 #define MMAP_VA_BITS_64 ((VA_BITS >= VA_BITS_SV48) ? VA_BITS_SV48 : VA_BITS)
 #define MMAP_MIN_VA_BITS_64 (VA_BITS_SV39)
 #define MMAP_VA_BITS (is_compat_task() ? VA_BITS_SV32 : MMAP_VA_BITS_64)
 #define MMAP_MIN_VA_BITS (is_compat_task() ? VA_BITS_SV32 : MMAP_MIN_VA_BITS_64)
-#else
-#define MMAP_VA_BITS ((VA_BITS >= VA_BITS_SV48) ? VA_BITS_SV48 : VA_BITS)
-#define MMAP_MIN_VA_BITS (VA_BITS_SV39)
-#endif /* CONFIG_COMPAT */
-
 #else
 #include <asm/pgtable-32.h>
 #endif /* CONFIG_64BIT */
@@ -439,9 +433,11 @@ static inline pte_t pte_mkhuge(pte_t pte)
 	return pte;
 }
 
+#ifdef CONFIG_RISCV_ISA_SVNAPOT
 #define pte_leaf_size(pte)	(pte_napot(pte) ?				\
 					napot_cont_size(napot_cont_order(pte)) :\
 					PAGE_SIZE)
+#endif
 
 #ifdef CONFIG_NUMA_BALANCING
 /*
@@ -517,12 +513,12 @@ static inline void set_pte(pte_t *ptep, pte_t pteval)
 	WRITE_ONCE(*ptep, pteval);
 }
 
-void flush_icache_pte(pte_t pte);
+void flush_icache_pte(struct mm_struct *mm, pte_t pte);
 
-static inline void __set_pte_at(pte_t *ptep, pte_t pteval)
+static inline void __set_pte_at(struct mm_struct *mm, pte_t *ptep, pte_t pteval)
 {
 	if (pte_present(pteval) && pte_exec(pteval))
-		flush_icache_pte(pteval);
+		flush_icache_pte(mm, pteval);
 
 	set_pte(ptep, pteval);
 }
@@ -535,7 +531,7 @@ static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
 	page_table_check_ptes_set(mm, ptep, pteval, nr);
 
 	for (;;) {
-		__set_pte_at(ptep, pteval);
+		__set_pte_at(mm, ptep, pteval);
 		if (--nr == 0)
 			break;
 		ptep++;
@@ -547,7 +543,7 @@ static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
 static inline void pte_clear(struct mm_struct *mm,
 	unsigned long addr, pte_t *ptep)
 {
-	__set_pte_at(ptep, __pte(0));
+	__set_pte_at(mm, ptep, __pte(0));
 }
 
 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS	/* defined in mm/pgtable.c */
@@ -662,6 +658,12 @@ static inline int pmd_write(pmd_t pmd)
 	return pte_write(pmd_pte(pmd));
 }
 
+#define pud_write pud_write
+static inline int pud_write(pud_t pud)
+{
+	return pte_write(pud_pte(pud));
+}
+
 #define pmd_dirty pmd_dirty
 static inline int pmd_dirty(pmd_t pmd)
 {
@@ -713,14 +715,14 @@ static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
 				pmd_t *pmdp, pmd_t pmd)
 {
 	page_table_check_pmd_set(mm, pmdp, pmd);
-	return __set_pte_at((pte_t *)pmdp, pmd_pte(pmd));
+	return __set_pte_at(mm, (pte_t *)pmdp, pmd_pte(pmd));
 }
 
 static inline void set_pud_at(struct mm_struct *mm, unsigned long addr,
 				pud_t *pudp, pud_t pud)
 {
 	page_table_check_pud_set(mm, pudp, pud);
-	return __set_pte_at((pte_t *)pudp, pud_pte(pud));
+	return __set_pte_at(mm, (pte_t *)pudp, pud_pte(pud));
 }
 
 #ifdef CONFIG_PAGE_TABLE_CHECK
@@ -871,8 +873,8 @@ static inline pte_t pte_swp_clear_exclusive(pte_t pte)
 #define TASK_SIZE_MIN	(PGDIR_SIZE_L3 * PTRS_PER_PGD / 2)
 
 #ifdef CONFIG_COMPAT
-#define TASK_SIZE_32	(_AC(0x80000000, UL))
-#define TASK_SIZE	(test_thread_flag(TIF_32BIT) ? \
+#define TASK_SIZE_32	(_AC(0x80000000, UL) - PAGE_SIZE)
+#define TASK_SIZE	(is_compat_task() ? \
 			 TASK_SIZE_32 : TASK_SIZE_64)
 #else
 #define TASK_SIZE	TASK_SIZE_64

arch/riscv/include/asm/processor.h

@@ -14,22 +14,21 @@
 
 #include <asm/ptrace.h>
 
-#ifdef CONFIG_64BIT
-#define DEFAULT_MAP_WINDOW	(UL(1) << (MMAP_VA_BITS - 1))
-#define STACK_TOP_MAX		TASK_SIZE
-
+/*
+ * addr is a hint to the maximum userspace address that mmap should provide, so
+ * this macro needs to return the largest address space available so that
+ * mmap_end < addr, being mmap_end the top of that address space.
+ * See Documentation/arch/riscv/vm-layout.rst for more details.
+ */
 #define arch_get_mmap_end(addr, len, flags)			\
 ({								\
 	unsigned long mmap_end;					\
 	typeof(addr) _addr = (addr);				\
-	if ((_addr) == 0 || (IS_ENABLED(CONFIG_COMPAT) && is_compat_task())) \
+	if ((_addr) == 0 || is_compat_task() ||			\
+	    ((_addr + len) > BIT(VA_BITS - 1)))			\
 		mmap_end = STACK_TOP_MAX;			\
-	else if ((_addr) >= VA_USER_SV57)			\
-		mmap_end = STACK_TOP_MAX;			\
-	else if ((((_addr) >= VA_USER_SV48)) && (VA_BITS >= VA_BITS_SV48)) \
-		mmap_end = VA_USER_SV48;			\
 	else							\
-		mmap_end = VA_USER_SV39;			\
+		mmap_end = (_addr + len);			\
 	mmap_end;						\
 })
 
@@ -39,17 +38,17 @@
 	typeof(addr) _addr = (addr);				\
 	typeof(base) _base = (base);				\
 	unsigned long rnd_gap = DEFAULT_MAP_WINDOW - (_base);	\
-	if ((_addr) == 0 || (IS_ENABLED(CONFIG_COMPAT) && is_compat_task())) \
+	if ((_addr) == 0 || is_compat_task() || 		\
+	    ((_addr + len) > BIT(VA_BITS - 1)))			\
 		mmap_base = (_base);				\
-	else if (((_addr) >= VA_USER_SV57) && (VA_BITS >= VA_BITS_SV57)) \
-		mmap_base = VA_USER_SV57 - rnd_gap;		\
-	else if ((((_addr) >= VA_USER_SV48)) && (VA_BITS >= VA_BITS_SV48)) \
-		mmap_base = VA_USER_SV48 - rnd_gap;		\
 	else							\
-		mmap_base = VA_USER_SV39 - rnd_gap;		\
+		mmap_base = (_addr + len) - rnd_gap;		\
 	mmap_base;						\
 })
 
+#ifdef CONFIG_64BIT
+#define DEFAULT_MAP_WINDOW	(UL(1) << (MMAP_VA_BITS - 1))
+#define STACK_TOP_MAX		TASK_SIZE_64
 #else
 #define DEFAULT_MAP_WINDOW	TASK_SIZE
 #define STACK_TOP_MAX		TASK_SIZE

arch/riscv/include/asm/simd.h

@@ -34,9 +34,9 @@ static __must_check inline bool may_use_simd(void)
 		return false;
 
 	/*
-	 * Nesting is acheived in preempt_v by spreading the control for
+	 * Nesting is achieved in preempt_v by spreading the control for
 	 * preemptible and non-preemptible kernel-mode Vector into two fields.
-	 * Always try to match with prempt_v if kernel V-context exists. Then,
+	 * Always try to match with preempt_v if kernel V-context exists. Then,
 	 * fallback to check non preempt_v if nesting happens, or if the config
 	 * is not set.
 	 */

arch/riscv/include/asm/suspend.h

@@ -56,4 +56,7 @@ int hibernate_resume_nonboot_cpu_disable(void);
 asmlinkage void hibernate_restore_image(unsigned long resume_satp, unsigned long satp_temp,
 					unsigned long cpu_resume);
 asmlinkage int hibernate_core_restore_code(void);
+bool riscv_sbi_hsm_is_supported(void);
+bool riscv_sbi_suspend_state_is_valid(u32 state);
+int riscv_sbi_hart_suspend(u32 state);
 #endif

arch/riscv/include/asm/sync_core.h

@@ -0,0 +1,29 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_RISCV_SYNC_CORE_H
+#define _ASM_RISCV_SYNC_CORE_H
+
+/*
+ * RISC-V implements return to user-space through an xRET instruction,
+ * which is not core serializing.
+ */
+static inline void sync_core_before_usermode(void)
+{
+	asm volatile ("fence.i" ::: "memory");
+}
+
+#ifdef CONFIG_SMP
+/*
+ * Ensure the next switch_mm() on every CPU issues a core serializing
+ * instruction for the given @mm.
+ */
+static inline void prepare_sync_core_cmd(struct mm_struct *mm)
+{
+	cpumask_setall(&mm->context.icache_stale_mask);
+}
+#else
+static inline void prepare_sync_core_cmd(struct mm_struct *mm)
+{
+}
+#endif /* CONFIG_SMP */
+
+#endif /* _ASM_RISCV_SYNC_CORE_H */

arch/riscv/include/asm/syscall_wrapper.h

@@ -12,26 +12,52 @@
 
 asmlinkage long __riscv_sys_ni_syscall(const struct pt_regs *);
 
-#define SC_RISCV_REGS_TO_ARGS(x, ...)				\
-	__MAP(x,__SC_ARGS					\
-	      ,,regs->orig_a0,,regs->a1,,regs->a2		\
+#ifdef CONFIG_64BIT
+
+#define __SYSCALL_SE_DEFINEx(x, prefix, name, ...)					\
+	static long __se_##prefix##name(__MAP(x,__SC_LONG,__VA_ARGS__));		\
+	static long __se_##prefix##name(__MAP(x,__SC_LONG,__VA_ARGS__))
+
+#define SC_RISCV_REGS_TO_ARGS(x, ...)							\
+	__MAP(x,__SC_ARGS								\
+	      ,,regs->orig_a0,,regs->a1,,regs->a2					\
 	      ,,regs->a3,,regs->a4,,regs->a5,,regs->a6)
 
+#else
+/*
+ * Use type aliasing to ensure registers a0-a6 are correctly passed to the syscall
+ * implementation when >word-size arguments are used.
+ */
+#define __SYSCALL_SE_DEFINEx(x, prefix, name, ...)					\
+	__diag_push();									\
+	__diag_ignore(GCC, 8, "-Wattribute-alias",					\
+			"Type aliasing is used to sanitize syscall arguments");		\
+	static long __se_##prefix##name(ulong, ulong, ulong, ulong, ulong, ulong, 	\
+					ulong)						\
+			__attribute__((alias(__stringify(___se_##prefix##name))));	\
+	__diag_pop();									\
+	static long noinline ___se_##prefix##name(__MAP(x,__SC_LONG,__VA_ARGS__));	\
+	static long ___se_##prefix##name(__MAP(x,__SC_LONG,__VA_ARGS__))
+
+#define SC_RISCV_REGS_TO_ARGS(x, ...) \
+	regs->orig_a0,regs->a1,regs->a2,regs->a3,regs->a4,regs->a5,regs->a6
+
+#endif /* CONFIG_64BIT */
+
 #ifdef CONFIG_COMPAT
 
 #define COMPAT_SYSCALL_DEFINEx(x, name, ...)						\
 	asmlinkage long __riscv_compat_sys##name(const struct pt_regs *regs);		\
 	ALLOW_ERROR_INJECTION(__riscv_compat_sys##name, ERRNO);				\
-	static long __se_compat_sys##name(__MAP(x,__SC_LONG,__VA_ARGS__));		\
 	static inline long __do_compat_sys##name(__MAP(x,__SC_DECL,__VA_ARGS__));	\
+	__SYSCALL_SE_DEFINEx(x, compat_sys, name, __VA_ARGS__)				\
+	{										\
+		return __do_compat_sys##name(__MAP(x,__SC_DELOUSE,__VA_ARGS__));	\
+	}										\
 	asmlinkage long __riscv_compat_sys##name(const struct pt_regs *regs)		\
 	{										\
 		return __se_compat_sys##name(SC_RISCV_REGS_TO_ARGS(x,__VA_ARGS__));	\
 	}										\
-	static long __se_compat_sys##name(__MAP(x,__SC_LONG,__VA_ARGS__))		\
-	{										\
-		return __do_compat_sys##name(__MAP(x,__SC_DELOUSE,__VA_ARGS__));	\
-	}										\
 	static inline long __do_compat_sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))
 
 #define COMPAT_SYSCALL_DEFINE0(sname)							\
@@ -51,19 +77,18 @@ asmlinkage long __riscv_sys_ni_syscall(const struct pt_regs *);
 #define __SYSCALL_DEFINEx(x, name, ...)						\
 	asmlinkage long __riscv_sys##name(const struct pt_regs *regs);		\
 	ALLOW_ERROR_INJECTION(__riscv_sys##name, ERRNO);			\
-	static long __se_sys##name(__MAP(x,__SC_LONG,__VA_ARGS__));		\
 	static inline long __do_sys##name(__MAP(x,__SC_DECL,__VA_ARGS__));	\
-	asmlinkage long __riscv_sys##name(const struct pt_regs *regs)		\
-	{									\
-		return __se_sys##name(SC_RISCV_REGS_TO_ARGS(x,__VA_ARGS__));	\
-	}									\
-	static long __se_sys##name(__MAP(x,__SC_LONG,__VA_ARGS__))		\
+	__SYSCALL_SE_DEFINEx(x, sys, name, __VA_ARGS__)				\
 	{									\
 		long ret = __do_sys##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
 		__MAP(x,__SC_TEST,__VA_ARGS__);					\
 		__PROTECT(x, ret,__MAP(x,__SC_ARGS,__VA_ARGS__));		\
 		return ret;							\
 	}									\
+	asmlinkage long __riscv_sys##name(const struct pt_regs *regs)		\
+	{									\
+		return __se_sys##name(SC_RISCV_REGS_TO_ARGS(x,__VA_ARGS__));	\
+	}									\
 	static inline long __do_sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))
 
 #define SYSCALL_DEFINE0(sname)							\

arch/riscv/include/asm/tlb.h

@@ -10,6 +10,24 @@ struct mmu_gather;
 
 static void tlb_flush(struct mmu_gather *tlb);
 
+#ifdef CONFIG_MMU
+#include <linux/swap.h>
+
+/*
+ * While riscv platforms with riscv_ipi_for_rfence as true require an IPI to
+ * perform TLB shootdown, some platforms with riscv_ipi_for_rfence as false use
+ * SBI to perform TLB shootdown. To keep software pagetable walkers safe in this
+ * case we switch to RCU based table free (MMU_GATHER_RCU_TABLE_FREE). See the
+ * comment below 'ifdef CONFIG_MMU_GATHER_RCU_TABLE_FREE' in include/asm-generic/tlb.h
+ * for more details.
+ */
+static inline void __tlb_remove_table(void *table)
+{
+	free_page_and_swap_cache(table);
+}
+
+#endif /* CONFIG_MMU */
+
 #define tlb_flush tlb_flush
 #include <asm-generic/tlb.h>
 

arch/riscv/include/asm/vector.h

@@ -284,4 +284,15 @@ static inline bool riscv_v_vstate_ctrl_user_allowed(void) { return false; }
 
 #endif /* CONFIG_RISCV_ISA_V */
 
+/*
+ * Return the implementation's vlen value.
+ *
+ * riscv_v_vsize contains the value of "32 vector registers with vlenb length"
+ * so rebuild the vlen value in bits from it.
+ */
+static inline int riscv_vector_vlen(void)
+{
+	return riscv_v_vsize / 32 * 8;
+}
+
 #endif /* ! __ASM_RISCV_VECTOR_H */

arch/riscv/include/asm/vendorid_list.h

@@ -5,7 +5,7 @@
 #ifndef ASM_VENDOR_LIST_H
 #define ASM_VENDOR_LIST_H
 
-#define ANDESTECH_VENDOR_ID	0x31e
+#define ANDES_VENDOR_ID		0x31e
 #define SIFIVE_VENDOR_ID	0x489
 #define THEAD_VENDOR_ID		0x5b7
 

arch/riscv/kernel/Makefile

@@ -39,7 +39,6 @@ extra-y += vmlinux.lds
 obj-y	+= head.o
 obj-y	+= soc.o
 obj-$(CONFIG_RISCV_ALTERNATIVE) += alternative.o
-obj-y	+= copy-unaligned.o
 obj-y	+= cpu.o
 obj-y	+= cpufeature.o
 obj-y	+= entry.o
@@ -64,6 +63,9 @@ obj-y	+= tests/
 obj-$(CONFIG_MMU) += vdso.o vdso/
 
 obj-$(CONFIG_RISCV_MISALIGNED)	+= traps_misaligned.o
+obj-$(CONFIG_RISCV_MISALIGNED)	+= unaligned_access_speed.o
+obj-$(CONFIG_RISCV_PROBE_UNALIGNED_ACCESS)	+= copy-unaligned.o
+
 obj-$(CONFIG_FPU)		+= fpu.o
 obj-$(CONFIG_RISCV_ISA_V)	+= vector.o
 obj-$(CONFIG_RISCV_ISA_V)	+= kernel_mode_vector.o

arch/riscv/kernel/alternative.c

@@ -43,7 +43,7 @@ static void riscv_fill_cpu_mfr_info(struct cpu_manufacturer_info_t *cpu_mfr_info
 
 	switch (cpu_mfr_info->vendor_id) {
 #ifdef CONFIG_ERRATA_ANDES
-	case ANDESTECH_VENDOR_ID:
+	case ANDES_VENDOR_ID:
 		cpu_mfr_info->patch_func = andes_errata_patch_func;
 		break;
 #endif

arch/riscv/kernel/cpufeature.c

@@ -11,7 +11,6 @@
 #include <linux/cpu.h>
 #include <linux/cpuhotplug.h>
 #include <linux/ctype.h>
-#include <linux/jump_label.h>
 #include <linux/log2.h>
 #include <linux/memory.h>
 #include <linux/module.h>
@@ -21,21 +20,13 @@
 #include <asm/cacheflush.h>
 #include <asm/cpufeature.h>
 #include <asm/hwcap.h>
-#include <asm/hwprobe.h>
 #include <asm/patch.h>
 #include <asm/processor.h>
 #include <asm/sbi.h>
 #include <asm/vector.h>
 
-#include "copy-unaligned.h"
-
 #define NUM_ALPHA_EXTS ('z' - 'a' + 1)
 
-#define MISALIGNED_ACCESS_JIFFIES_LG2 1
-#define MISALIGNED_BUFFER_SIZE 0x4000
-#define MISALIGNED_BUFFER_ORDER get_order(MISALIGNED_BUFFER_SIZE)
-#define MISALIGNED_COPY_SIZE ((MISALIGNED_BUFFER_SIZE / 2) - 0x80)
-
 unsigned long elf_hwcap __read_mostly;
 
 /* Host ISA bitmap */
@@ -44,11 +35,6 @@ static DECLARE_BITMAP(riscv_isa, RISCV_ISA_EXT_MAX) __read_mostly;
 /* Per-cpu ISA extensions. */
 struct riscv_isainfo hart_isa[NR_CPUS];
 
-/* Performance information */
-DEFINE_PER_CPU(long, misaligned_access_speed);
-
-static cpumask_t fast_misaligned_access;
-
 /**
  * riscv_isa_extension_base() - Get base extension word
  *
@@ -318,6 +304,7 @@ const struct riscv_isa_ext_data riscv_isa_ext[] = {
 	__RISCV_ISA_EXT_DATA(svinval, RISCV_ISA_EXT_SVINVAL),
 	__RISCV_ISA_EXT_DATA(svnapot, RISCV_ISA_EXT_SVNAPOT),
 	__RISCV_ISA_EXT_DATA(svpbmt, RISCV_ISA_EXT_SVPBMT),
+	__RISCV_ISA_EXT_DATA(xandespmu, RISCV_ISA_EXT_XANDESPMU),
 };
 
 const size_t riscv_isa_ext_count = ARRAY_SIZE(riscv_isa_ext);
@@ -731,247 +718,6 @@ unsigned long riscv_get_elf_hwcap(void)
 	return hwcap;
 }
 
-static int check_unaligned_access(void *param)
-{
-	int cpu = smp_processor_id();
-	u64 start_cycles, end_cycles;
-	u64 word_cycles;
-	u64 byte_cycles;
-	int ratio;
-	unsigned long start_jiffies, now;
-	struct page *page = param;
-	void *dst;
-	void *src;
-	long speed = RISCV_HWPROBE_MISALIGNED_SLOW;
-
-	if (check_unaligned_access_emulated(cpu))
-		return 0;
-
-	/* Make an unaligned destination buffer. */
-	dst = (void *)((unsigned long)page_address(page) | 0x1);
-	/* Unalign src as well, but differently (off by 1 + 2 = 3). */
-	src = dst + (MISALIGNED_BUFFER_SIZE / 2);
-	src += 2;
-	word_cycles = -1ULL;
-	/* Do a warmup. */
-	__riscv_copy_words_unaligned(dst, src, MISALIGNED_COPY_SIZE);
-	preempt_disable();
-	start_jiffies = jiffies;
-	while ((now = jiffies) == start_jiffies)
-		cpu_relax();
-
-	/*
-	 * For a fixed amount of time, repeatedly try the function, and take
-	 * the best time in cycles as the measurement.
-	 */
-	while (time_before(jiffies, now + (1 << MISALIGNED_ACCESS_JIFFIES_LG2))) {
-		start_cycles = get_cycles64();
-		/* Ensure the CSR read can't reorder WRT to the copy. */
-		mb();
-		__riscv_copy_words_unaligned(dst, src, MISALIGNED_COPY_SIZE);
-		/* Ensure the copy ends before the end time is snapped. */
-		mb();
-		end_cycles = get_cycles64();
-		if ((end_cycles - start_cycles) < word_cycles)
-			word_cycles = end_cycles - start_cycles;
-	}
-
-	byte_cycles = -1ULL;
-	__riscv_copy_bytes_unaligned(dst, src, MISALIGNED_COPY_SIZE);
-	start_jiffies = jiffies;
-	while ((now = jiffies) == start_jiffies)
-		cpu_relax();
-
-	while (time_before(jiffies, now + (1 << MISALIGNED_ACCESS_JIFFIES_LG2))) {
-		start_cycles = get_cycles64();
-		mb();
-		__riscv_copy_bytes_unaligned(dst, src, MISALIGNED_COPY_SIZE);
-		mb();
-		end_cycles = get_cycles64();
-		if ((end_cycles - start_cycles) < byte_cycles)
-			byte_cycles = end_cycles - start_cycles;
-	}
-
-	preempt_enable();
-
-	/* Don't divide by zero. */
-	if (!word_cycles || !byte_cycles) {
-		pr_warn("cpu%d: rdtime lacks granularity needed to measure unaligned access speed\n",
-			cpu);
-
-		return 0;
-	}
-
-	if (word_cycles < byte_cycles)
-		speed = RISCV_HWPROBE_MISALIGNED_FAST;
-
-	ratio = div_u64((byte_cycles * 100), word_cycles);
-	pr_info("cpu%d: Ratio of byte access time to unaligned word access is %d.%02d, unaligned accesses are %s\n",
-		cpu,
-		ratio / 100,
-		ratio % 100,
-		(speed == RISCV_HWPROBE_MISALIGNED_FAST) ? "fast" : "slow");
-
-	per_cpu(misaligned_access_speed, cpu) = speed;
-
-	/*
-	 * Set the value of fast_misaligned_access of a CPU. These operations
-	 * are atomic to avoid race conditions.
-	 */
-	if (speed == RISCV_HWPROBE_MISALIGNED_FAST)
-		cpumask_set_cpu(cpu, &fast_misaligned_access);
-	else
-		cpumask_clear_cpu(cpu, &fast_misaligned_access);
-
-	return 0;
-}
-
-static void check_unaligned_access_nonboot_cpu(void *param)
-{
-	unsigned int cpu = smp_processor_id();
-	struct page **pages = param;
-
-	if (smp_processor_id() != 0)
-		check_unaligned_access(pages[cpu]);
-}
-
-DEFINE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
-
-static void modify_unaligned_access_branches(cpumask_t *mask, int weight)
-{
-	if (cpumask_weight(mask) == weight)
-		static_branch_enable_cpuslocked(&fast_misaligned_access_speed_key);
-	else
-		static_branch_disable_cpuslocked(&fast_misaligned_access_speed_key);
-}
-
-static void set_unaligned_access_static_branches_except_cpu(int cpu)
-{
-	/*
-	 * Same as set_unaligned_access_static_branches, except excludes the
-	 * given CPU from the result. When a CPU is hotplugged into an offline
-	 * state, this function is called before the CPU is set to offline in
-	 * the cpumask, and thus the CPU needs to be explicitly excluded.
-	 */
-
-	cpumask_t fast_except_me;
-
-	cpumask_and(&fast_except_me, &fast_misaligned_access, cpu_online_mask);
-	cpumask_clear_cpu(cpu, &fast_except_me);
-
-	modify_unaligned_access_branches(&fast_except_me, num_online_cpus() - 1);
-}
-
-static void set_unaligned_access_static_branches(void)
-{
-	/*
-	 * This will be called after check_unaligned_access_all_cpus so the
-	 * result of unaligned access speed for all CPUs will be available.
-	 *
-	 * To avoid the number of online cpus changing between reading
-	 * cpu_online_mask and calling num_online_cpus, cpus_read_lock must be
-	 * held before calling this function.
-	 */
-
-	cpumask_t fast_and_online;
-
-	cpumask_and(&fast_and_online, &fast_misaligned_access, cpu_online_mask);
-
-	modify_unaligned_access_branches(&fast_and_online, num_online_cpus());
-}
-
-static int lock_and_set_unaligned_access_static_branch(void)
-{
-	cpus_read_lock();
-	set_unaligned_access_static_branches();
-	cpus_read_unlock();
-
-	return 0;
-}
-
-arch_initcall_sync(lock_and_set_unaligned_access_static_branch);
-
-static int riscv_online_cpu(unsigned int cpu)
-{
-	static struct page *buf;
-
-	/* We are already set since the last check */
-	if (per_cpu(misaligned_access_speed, cpu) != RISCV_HWPROBE_MISALIGNED_UNKNOWN)
-		goto exit;
-
-	buf = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
-	if (!buf) {
-		pr_warn("Allocation failure, not measuring misaligned performance\n");
-		return -ENOMEM;
-	}
-
-	check_unaligned_access(buf);
-	__free_pages(buf, MISALIGNED_BUFFER_ORDER);
-
-exit:
-	set_unaligned_access_static_branches();
-
-	return 0;
-}
-
-static int riscv_offline_cpu(unsigned int cpu)
-{
-	set_unaligned_access_static_branches_except_cpu(cpu);
-
-	return 0;
-}
-
-/* Measure unaligned access on all CPUs present at boot in parallel. */
-static int check_unaligned_access_all_cpus(void)
-{
-	unsigned int cpu;
-	unsigned int cpu_count = num_possible_cpus();
-	struct page **bufs = kzalloc(cpu_count * sizeof(struct page *),
-				     GFP_KERNEL);
-
-	if (!bufs) {
-		pr_warn("Allocation failure, not measuring misaligned performance\n");
-		return 0;
-	}
-
-	/*
-	 * Allocate separate buffers for each CPU so there's no fighting over
-	 * cache lines.
-	 */
-	for_each_cpu(cpu, cpu_online_mask) {
-		bufs[cpu] = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
-		if (!bufs[cpu]) {
-			pr_warn("Allocation failure, not measuring misaligned performance\n");
-			goto out;
-		}
-	}
-
-	/* Check everybody except 0, who stays behind to tend jiffies. */
-	on_each_cpu(check_unaligned_access_nonboot_cpu, bufs, 1);
-
-	/* Check core 0. */
-	smp_call_on_cpu(0, check_unaligned_access, bufs[0], true);
-
-	/*
-	 * Setup hotplug callbacks for any new CPUs that come online or go
-	 * offline.
-	 */
-	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
-				  riscv_online_cpu, riscv_offline_cpu);
-
-out:
-	unaligned_emulation_finish();
-	for_each_cpu(cpu, cpu_online_mask) {
-		if (bufs[cpu])
-			__free_pages(bufs[cpu], MISALIGNED_BUFFER_ORDER);
-	}
-
-	kfree(bufs);
-	return 0;
-}
-
-arch_initcall(check_unaligned_access_all_cpus);
-
 void riscv_user_isa_enable(void)
 {
 	if (riscv_cpu_has_extension_unlikely(smp_processor_id(), RISCV_ISA_EXT_ZICBOZ))

arch/riscv/kernel/entry.S

@@ -111,6 +111,7 @@ SYM_CODE_START(handle_exception)
 1:
 	tail do_trap_unknown
 SYM_CODE_END(handle_exception)
+ASM_NOKPROBE(handle_exception)
 
 /*
  * The ret_from_exception must be called with interrupt disabled. Here is the
@@ -184,6 +185,7 @@ SYM_CODE_START_NOALIGN(ret_from_exception)
 	sret
 #endif
 SYM_CODE_END(ret_from_exception)
+ASM_NOKPROBE(ret_from_exception)
 
 #ifdef CONFIG_VMAP_STACK
 SYM_CODE_START_LOCAL(handle_kernel_stack_overflow)
@@ -219,6 +221,7 @@ SYM_CODE_START_LOCAL(handle_kernel_stack_overflow)
 	move a0, sp
 	tail handle_bad_stack
 SYM_CODE_END(handle_kernel_stack_overflow)
+ASM_NOKPROBE(handle_kernel_stack_overflow)
 #endif
 
 SYM_CODE_START(ret_from_fork)

arch/riscv/kernel/pi/Makefile

@@ -9,6 +9,9 @@ KBUILD_CFLAGS	:= $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) -fpie \
 		   -fno-asynchronous-unwind-tables -fno-unwind-tables \
 		   $(call cc-option,-fno-addrsig)
 
+# Disable LTO
+KBUILD_CFLAGS	:= $(filter-out $(CC_FLAGS_LTO), $(KBUILD_CFLAGS))
+
 KBUILD_CFLAGS	+= -mcmodel=medany
 
 CFLAGS_cmdline_early.o += -D__NO_FORTIFY

arch/riscv/kernel/ptrace.c

@@ -377,14 +377,14 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
 
 	return ret;
 }
+#else
+static const struct user_regset_view compat_riscv_user_native_view = {};
 #endif /* CONFIG_COMPAT */
 
 const struct user_regset_view *task_user_regset_view(struct task_struct *task)
 {
-#ifdef CONFIG_COMPAT
-	if (test_tsk_thread_flag(task, TIF_32BIT))
+	if (is_compat_thread(&task->thread_info))
 		return &compat_riscv_user_native_view;
 	else
-#endif
 		return &riscv_user_native_view;
 }

arch/riscv/kernel/smpboot.c

@@ -28,7 +28,6 @@
 
 #include <asm/cpufeature.h>
 #include <asm/cpu_ops.h>
-#include <asm/cpufeature.h>
 #include <asm/irq.h>
 #include <asm/mmu_context.h>
 #include <asm/numa.h>

arch/riscv/kernel/suspend.c

@@ -132,4 +132,53 @@ static int __init sbi_system_suspend_init(void)
 }
 
 arch_initcall(sbi_system_suspend_init);
+
+static int sbi_suspend_finisher(unsigned long suspend_type,
+				unsigned long resume_addr,
+				unsigned long opaque)
+{
+	struct sbiret ret;
+
+	ret = sbi_ecall(SBI_EXT_HSM, SBI_EXT_HSM_HART_SUSPEND,
+			suspend_type, resume_addr, opaque, 0, 0, 0);
+
+	return (ret.error) ? sbi_err_map_linux_errno(ret.error) : 0;
+}
+
+int riscv_sbi_hart_suspend(u32 state)
+{
+	if (state & SBI_HSM_SUSP_NON_RET_BIT)
+		return cpu_suspend(state, sbi_suspend_finisher);
+	else
+		return sbi_suspend_finisher(state, 0, 0);
+}
+
+bool riscv_sbi_suspend_state_is_valid(u32 state)
+{
+	if (state > SBI_HSM_SUSPEND_RET_DEFAULT &&
+	    state < SBI_HSM_SUSPEND_RET_PLATFORM)
+		return false;
+
+	if (state > SBI_HSM_SUSPEND_NON_RET_DEFAULT &&
+	    state < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
+		return false;
+
+	return true;
+}
+
+bool riscv_sbi_hsm_is_supported(void)
+{
+	/*
+	 * The SBI HSM suspend function is only available when:
+	 * 1) SBI version is 0.3 or higher
+	 * 2) SBI HSM extension is available
+	 */
+	if (sbi_spec_version < sbi_mk_version(0, 3) ||
+	    !sbi_probe_extension(SBI_EXT_HSM)) {
+		pr_info("HSM suspend not available\n");
+		return false;
+	}
+
+	return true;
+}
 #endif /* CONFIG_RISCV_SBI */

arch/riscv/kernel/sys_hwprobe.c

@@ -147,6 +147,7 @@ static bool hwprobe_ext0_has(const struct cpumask *cpus, unsigned long ext)
 	return (pair.value & ext);
 }
 
+#if defined(CONFIG_RISCV_PROBE_UNALIGNED_ACCESS)
 static u64 hwprobe_misaligned(const struct cpumask *cpus)
 {
 	int cpu;
@@ -169,6 +170,18 @@ static u64 hwprobe_misaligned(const struct cpumask *cpus)
 
 	return perf;
 }
+#else
+static u64 hwprobe_misaligned(const struct cpumask *cpus)
+{
+	if (IS_ENABLED(CONFIG_RISCV_EFFICIENT_UNALIGNED_ACCESS))
+		return RISCV_HWPROBE_MISALIGNED_FAST;
+
+	if (IS_ENABLED(CONFIG_RISCV_EMULATED_UNALIGNED_ACCESS) && unaligned_ctl_available())
+		return RISCV_HWPROBE_MISALIGNED_EMULATED;
+
+	return RISCV_HWPROBE_MISALIGNED_SLOW;
+}
+#endif
 
 static void hwprobe_one_pair(struct riscv_hwprobe *pair,
 			     const struct cpumask *cpus)

arch/riscv/kernel/traps.c

@@ -6,6 +6,7 @@
 #include <linux/cpu.h>
 #include <linux/kernel.h>
 #include <linux/init.h>
+#include <linux/randomize_kstack.h>
 #include <linux/sched.h>
 #include <linux/sched/debug.h>
 #include <linux/sched/signal.h>
@@ -310,7 +311,8 @@ asmlinkage __visible __trap_section void do_trap_break(struct pt_regs *regs)
 	}
 }
 
-asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
+asmlinkage __visible __trap_section  __no_stack_protector
+void do_trap_ecall_u(struct pt_regs *regs)
 {
 	if (user_mode(regs)) {
 		long syscall = regs->a7;
@@ -322,10 +324,23 @@ asmlinkage __visible __trap_section void do_trap_ecall_u(struct pt_regs *regs)
 
 		syscall = syscall_enter_from_user_mode(regs, syscall);
 
+		add_random_kstack_offset();
+
 		if (syscall >= 0 && syscall < NR_syscalls)
 			syscall_handler(regs, syscall);
 		else if (syscall != -1)
 			regs->a0 = -ENOSYS;
+		/*
+		 * Ultimately, this value will get limited by KSTACK_OFFSET_MAX(),
+		 * so the maximum stack offset is 1k bytes (10 bits).
+		 *
+		 * The actual entropy will be further reduced by the compiler when
+		 * applying stack alignment constraints: 16-byte (i.e. 4-bit) aligned
+		 * for RV32I or RV64I.
+		 *
+		 * The resulting 6 bits of entropy is seen in SP[9:4].
+		 */
+		choose_random_kstack_offset(get_random_u16());
 
 		syscall_exit_to_user_mode(regs);
 	} else {

arch/riscv/kernel/traps_misaligned.c

@@ -413,7 +413,9 @@ int handle_misaligned_load(struct pt_regs *regs)
 
 	perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, addr);
 
+#ifdef CONFIG_RISCV_PROBE_UNALIGNED_ACCESS
 	*this_cpu_ptr(&misaligned_access_speed) = RISCV_HWPROBE_MISALIGNED_EMULATED;
+#endif
 
 	if (!unaligned_enabled)
 		return -1;
@@ -596,7 +598,7 @@ int handle_misaligned_store(struct pt_regs *regs)
 	return 0;
 }
 
-bool check_unaligned_access_emulated(int cpu)
+static bool check_unaligned_access_emulated(int cpu)
 {
 	long *mas_ptr = per_cpu_ptr(&misaligned_access_speed, cpu);
 	unsigned long tmp_var, tmp_val;
@@ -623,7 +625,7 @@ bool check_unaligned_access_emulated(int cpu)
 	return misaligned_emu_detected;
 }
 
-void unaligned_emulation_finish(void)
+bool check_unaligned_access_emulated_all_cpus(void)
 {
 	int cpu;
 
@@ -632,13 +634,12 @@ void unaligned_emulation_finish(void)
 	 * accesses emulated since tasks requesting such control can run on any
 	 * CPU.
 	 */
-	for_each_present_cpu(cpu) {
-		if (per_cpu(misaligned_access_speed, cpu) !=
-					RISCV_HWPROBE_MISALIGNED_EMULATED) {
-			return;
-		}
-	}
+	for_each_online_cpu(cpu)
+		if (!check_unaligned_access_emulated(cpu))
+			return false;
+
 	unaligned_ctl = true;
+	return true;
 }
 
 bool unaligned_ctl_available(void)

arch/riscv/kernel/unaligned_access_speed.c

@@ -0,0 +1,281 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2024 Rivos Inc.
+ */
+
+#include <linux/cpu.h>
+#include <linux/cpumask.h>
+#include <linux/jump_label.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/types.h>
+#include <asm/cpufeature.h>
+#include <asm/hwprobe.h>
+
+#include "copy-unaligned.h"
+
+#define MISALIGNED_ACCESS_JIFFIES_LG2 1
+#define MISALIGNED_BUFFER_SIZE 0x4000
+#define MISALIGNED_BUFFER_ORDER get_order(MISALIGNED_BUFFER_SIZE)
+#define MISALIGNED_COPY_SIZE ((MISALIGNED_BUFFER_SIZE / 2) - 0x80)
+
+DEFINE_PER_CPU(long, misaligned_access_speed);
+
+#ifdef CONFIG_RISCV_PROBE_UNALIGNED_ACCESS
+static cpumask_t fast_misaligned_access;
+static int check_unaligned_access(void *param)
+{
+	int cpu = smp_processor_id();
+	u64 start_cycles, end_cycles;
+	u64 word_cycles;
+	u64 byte_cycles;
+	int ratio;
+	unsigned long start_jiffies, now;
+	struct page *page = param;
+	void *dst;
+	void *src;
+	long speed = RISCV_HWPROBE_MISALIGNED_SLOW;
+
+	if (per_cpu(misaligned_access_speed, cpu) != RISCV_HWPROBE_MISALIGNED_UNKNOWN)
+		return 0;
+
+	/* Make an unaligned destination buffer. */
+	dst = (void *)((unsigned long)page_address(page) | 0x1);
+	/* Unalign src as well, but differently (off by 1 + 2 = 3). */
+	src = dst + (MISALIGNED_BUFFER_SIZE / 2);
+	src += 2;
+	word_cycles = -1ULL;
+	/* Do a warmup. */
+	__riscv_copy_words_unaligned(dst, src, MISALIGNED_COPY_SIZE);
+	preempt_disable();
+	start_jiffies = jiffies;
+	while ((now = jiffies) == start_jiffies)
+		cpu_relax();
+
+	/*
+	 * For a fixed amount of time, repeatedly try the function, and take
+	 * the best time in cycles as the measurement.
+	 */
+	while (time_before(jiffies, now + (1 << MISALIGNED_ACCESS_JIFFIES_LG2))) {
+		start_cycles = get_cycles64();
+		/* Ensure the CSR read can't reorder WRT to the copy. */
+		mb();
+		__riscv_copy_words_unaligned(dst, src, MISALIGNED_COPY_SIZE);
+		/* Ensure the copy ends before the end time is snapped. */
+		mb();
+		end_cycles = get_cycles64();
+		if ((end_cycles - start_cycles) < word_cycles)
+			word_cycles = end_cycles - start_cycles;
+	}
+
+	byte_cycles = -1ULL;
+	__riscv_copy_bytes_unaligned(dst, src, MISALIGNED_COPY_SIZE);
+	start_jiffies = jiffies;
+	while ((now = jiffies) == start_jiffies)
+		cpu_relax();
+
+	while (time_before(jiffies, now + (1 << MISALIGNED_ACCESS_JIFFIES_LG2))) {
+		start_cycles = get_cycles64();
+		mb();
+		__riscv_copy_bytes_unaligned(dst, src, MISALIGNED_COPY_SIZE);
+		mb();
+		end_cycles = get_cycles64();
+		if ((end_cycles - start_cycles) < byte_cycles)
+			byte_cycles = end_cycles - start_cycles;
+	}
+
+	preempt_enable();
+
+	/* Don't divide by zero. */
+	if (!word_cycles || !byte_cycles) {
+		pr_warn("cpu%d: rdtime lacks granularity needed to measure unaligned access speed\n",
+			cpu);
+
+		return 0;
+	}
+
+	if (word_cycles < byte_cycles)
+		speed = RISCV_HWPROBE_MISALIGNED_FAST;
+
+	ratio = div_u64((byte_cycles * 100), word_cycles);
+	pr_info("cpu%d: Ratio of byte access time to unaligned word access is %d.%02d, unaligned accesses are %s\n",
+		cpu,
+		ratio / 100,
+		ratio % 100,
+		(speed == RISCV_HWPROBE_MISALIGNED_FAST) ? "fast" : "slow");
+
+	per_cpu(misaligned_access_speed, cpu) = speed;
+
+	/*
+	 * Set the value of fast_misaligned_access of a CPU. These operations
+	 * are atomic to avoid race conditions.
+	 */
+	if (speed == RISCV_HWPROBE_MISALIGNED_FAST)
+		cpumask_set_cpu(cpu, &fast_misaligned_access);
+	else
+		cpumask_clear_cpu(cpu, &fast_misaligned_access);
+
+	return 0;
+}
+
+static void check_unaligned_access_nonboot_cpu(void *param)
+{
+	unsigned int cpu = smp_processor_id();
+	struct page **pages = param;
+
+	if (smp_processor_id() != 0)
+		check_unaligned_access(pages[cpu]);
+}
+
+DEFINE_STATIC_KEY_FALSE(fast_unaligned_access_speed_key);
+
+static void modify_unaligned_access_branches(cpumask_t *mask, int weight)
+{
+	if (cpumask_weight(mask) == weight)
+		static_branch_enable_cpuslocked(&fast_unaligned_access_speed_key);
+	else
+		static_branch_disable_cpuslocked(&fast_unaligned_access_speed_key);
+}
+
+static void set_unaligned_access_static_branches_except_cpu(int cpu)
+{
+	/*
+	 * Same as set_unaligned_access_static_branches, except excludes the
+	 * given CPU from the result. When a CPU is hotplugged into an offline
+	 * state, this function is called before the CPU is set to offline in
+	 * the cpumask, and thus the CPU needs to be explicitly excluded.
+	 */
+
+	cpumask_t fast_except_me;
+
+	cpumask_and(&fast_except_me, &fast_misaligned_access, cpu_online_mask);
+	cpumask_clear_cpu(cpu, &fast_except_me);
+
+	modify_unaligned_access_branches(&fast_except_me, num_online_cpus() - 1);
+}
+
+static void set_unaligned_access_static_branches(void)
+{
+	/*
+	 * This will be called after check_unaligned_access_all_cpus so the
+	 * result of unaligned access speed for all CPUs will be available.
+	 *
+	 * To avoid the number of online cpus changing between reading
+	 * cpu_online_mask and calling num_online_cpus, cpus_read_lock must be
+	 * held before calling this function.
+	 */
+
+	cpumask_t fast_and_online;
+
+	cpumask_and(&fast_and_online, &fast_misaligned_access, cpu_online_mask);
+
+	modify_unaligned_access_branches(&fast_and_online, num_online_cpus());
+}
+
+static int lock_and_set_unaligned_access_static_branch(void)
+{
+	cpus_read_lock();
+	set_unaligned_access_static_branches();
+	cpus_read_unlock();
+
+	return 0;
+}
+
+arch_initcall_sync(lock_and_set_unaligned_access_static_branch);
+
+static int riscv_online_cpu(unsigned int cpu)
+{
+	static struct page *buf;
+
+	/* We are already set since the last check */
+	if (per_cpu(misaligned_access_speed, cpu) != RISCV_HWPROBE_MISALIGNED_UNKNOWN)
+		goto exit;
+
+	buf = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
+	if (!buf) {
+		pr_warn("Allocation failure, not measuring misaligned performance\n");
+		return -ENOMEM;
+	}
+
+	check_unaligned_access(buf);
+	__free_pages(buf, MISALIGNED_BUFFER_ORDER);
+
+exit:
+	set_unaligned_access_static_branches();
+
+	return 0;
+}
+
+static int riscv_offline_cpu(unsigned int cpu)
+{
+	set_unaligned_access_static_branches_except_cpu(cpu);
+
+	return 0;
+}
+
+/* Measure unaligned access speed on all CPUs present at boot in parallel. */
+static int check_unaligned_access_speed_all_cpus(void)
+{
+	unsigned int cpu;
+	unsigned int cpu_count = num_possible_cpus();
+	struct page **bufs = kcalloc(cpu_count, sizeof(*bufs), GFP_KERNEL);
+
+	if (!bufs) {
+		pr_warn("Allocation failure, not measuring misaligned performance\n");
+		return 0;
+	}
+
+	/*
+	 * Allocate separate buffers for each CPU so there's no fighting over
+	 * cache lines.
+	 */
+	for_each_cpu(cpu, cpu_online_mask) {
+		bufs[cpu] = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
+		if (!bufs[cpu]) {
+			pr_warn("Allocation failure, not measuring misaligned performance\n");
+			goto out;
+		}
+	}
+
+	/* Check everybody except 0, who stays behind to tend jiffies. */
+	on_each_cpu(check_unaligned_access_nonboot_cpu, bufs, 1);
+
+	/* Check core 0. */
+	smp_call_on_cpu(0, check_unaligned_access, bufs[0], true);
+
+	/*
+	 * Setup hotplug callbacks for any new CPUs that come online or go
+	 * offline.
+	 */
+	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
+				  riscv_online_cpu, riscv_offline_cpu);
+
+out:
+	for_each_cpu(cpu, cpu_online_mask) {
+		if (bufs[cpu])
+			__free_pages(bufs[cpu], MISALIGNED_BUFFER_ORDER);
+	}
+
+	kfree(bufs);
+	return 0;
+}
+
+static int check_unaligned_access_all_cpus(void)
+{
+	bool all_cpus_emulated = check_unaligned_access_emulated_all_cpus();
+
+	if (!all_cpus_emulated)
+		return check_unaligned_access_speed_all_cpus();
+
+	return 0;
+}
+#else /* CONFIG_RISCV_PROBE_UNALIGNED_ACCESS */
+static int check_unaligned_access_all_cpus(void)
+{
+	check_unaligned_access_emulated_all_cpus();
+
+	return 0;
+}
+#endif
+
+arch_initcall(check_unaligned_access_all_cpus);

arch/riscv/lib/csum.c

@@ -3,7 +3,7 @@
  * Checksum library
  *
  * Influenced by arch/arm64/lib/csum.c
- * Copyright (C) 2023 Rivos Inc.
+ * Copyright (C) 2023-2024 Rivos Inc.
  */
 #include <linux/bitops.h>
 #include <linux/compiler.h>
@@ -318,10 +318,7 @@ unsigned int do_csum(const unsigned char *buff, int len)
 	 * branches. The largest chunk of overlap was delegated into the
 	 * do_csum_common function.
 	 */
-	if (static_branch_likely(&fast_misaligned_access_speed_key))
-		return do_csum_no_alignment(buff, len);
-
-	if (((unsigned long)buff & OFFSET_MASK) == 0)
+	if (has_fast_unaligned_accesses() || (((unsigned long)buff & OFFSET_MASK) == 0))
 		return do_csum_no_alignment(buff, len);
 
 	return do_csum_with_alignment(buff, len);

arch/riscv/lib/uaccess_vector.S

@@ -1,7 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 
 #include <linux/linkage.h>
-#include <asm-generic/export.h>
 #include <asm/asm.h>
 #include <asm/asm-extable.h>
 #include <asm/csr.h>

arch/riscv/mm/cacheflush.c

@@ -82,12 +82,12 @@ void flush_icache_mm(struct mm_struct *mm, bool local)
 #endif /* CONFIG_SMP */
 
 #ifdef CONFIG_MMU
-void flush_icache_pte(pte_t pte)
+void flush_icache_pte(struct mm_struct *mm, pte_t pte)
 {
 	struct folio *folio = page_folio(pte_page(pte));
 
 	if (!test_bit(PG_dcache_clean, &folio->flags)) {
-		flush_icache_all();
+		flush_icache_mm(mm, false);
 		set_bit(PG_dcache_clean, &folio->flags);
 	}
 }

arch/riscv/mm/context.c

@@ -323,6 +323,8 @@ void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	if (unlikely(prev == next))
 		return;
 
+	membarrier_arch_switch_mm(prev, next, task);
+
 	/*
 	 * Mark the current MM context as inactive, and the next as
 	 * active.  This is at least used by the icache flushing

arch/riscv/mm/init.c

@@ -764,6 +764,11 @@ static int __init print_no5lvl(char *p)
 }
 early_param("no5lvl", print_no5lvl);
 
+static void __init set_mmap_rnd_bits_max(void)
+{
+	mmap_rnd_bits_max = MMAP_VA_BITS - PAGE_SHIFT - 3;
+}
+
 /*
  * There is a simple way to determine if 4-level is supported by the
  * underlying hardware: establish 1:1 mapping in 4-level page table mode
@@ -1078,6 +1083,7 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa)
 
 #if defined(CONFIG_64BIT) && !defined(CONFIG_XIP_KERNEL)
 	set_satp_mode(dtb_pa);
+	set_mmap_rnd_bits_max();
 #endif
 
 	/*

arch/riscv/mm/pgtable.c

@@ -10,7 +10,7 @@ int ptep_set_access_flags(struct vm_area_struct *vma,
 			  pte_t entry, int dirty)
 {
 	if (!pte_same(ptep_get(ptep), entry))
-		__set_pte_at(ptep, entry);
+		__set_pte_at(vma->vm_mm, ptep, entry);
 	/*
 	 * update_mmu_cache will unconditionally execute, handling both
 	 * the case that the PTE changed and the spurious fault case.

crypto/Kconfig

@@ -1497,6 +1497,9 @@ endif
 if PPC
 source "arch/powerpc/crypto/Kconfig"
 endif
+if RISCV
+source "arch/riscv/crypto/Kconfig"
+endif
 if S390
 source "arch/s390/crypto/Kconfig"
 endif

drivers/acpi/Kconfig

@@ -286,7 +286,7 @@ config ACPI_CPPC_LIB
 
 config ACPI_PROCESSOR
 	tristate "Processor"
-	depends on X86 || ARM64 || LOONGARCH
+	depends on X86 || ARM64 || LOONGARCH || RISCV
 	select ACPI_PROCESSOR_IDLE
 	select ACPI_CPU_FREQ_PSS if X86 || LOONGARCH
 	select THERMAL

drivers/acpi/riscv/Makefile

@@ -1,2 +1,4 @@
 # SPDX-License-Identifier: GPL-2.0-only
-obj-y 	+= rhct.o
+obj-y					+= rhct.o
+obj-$(CONFIG_ACPI_PROCESSOR_IDLE)	+= cpuidle.o
+obj-$(CONFIG_ACPI_CPPC_LIB)		+= cppc.o

drivers/acpi/riscv/cppc.c

@@ -0,0 +1,157 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Implement CPPC FFH helper routines for RISC-V.
+ *
+ * Copyright (C) 2024 Ventana Micro Systems Inc.
+ */
+
+#include <acpi/cppc_acpi.h>
+#include <asm/csr.h>
+#include <asm/sbi.h>
+
+#define SBI_EXT_CPPC 0x43505043
+
+/* CPPC interfaces defined in SBI spec */
+#define SBI_CPPC_PROBE			0x0
+#define SBI_CPPC_READ			0x1
+#define SBI_CPPC_READ_HI		0x2
+#define SBI_CPPC_WRITE			0x3
+
+/* RISC-V FFH definitions from RISC-V FFH spec */
+#define FFH_CPPC_TYPE(r)		(((r) & GENMASK_ULL(63, 60)) >> 60)
+#define FFH_CPPC_SBI_REG(r)		((r) & GENMASK(31, 0))
+#define FFH_CPPC_CSR_NUM(r)		((r) & GENMASK(11, 0))
+
+#define FFH_CPPC_SBI			0x1
+#define FFH_CPPC_CSR			0x2
+
+struct sbi_cppc_data {
+	u64 val;
+	u32 reg;
+	struct sbiret ret;
+};
+
+static bool cppc_ext_present;
+
+static int __init sbi_cppc_init(void)
+{
+	if (sbi_spec_version >= sbi_mk_version(2, 0) &&
+	    sbi_probe_extension(SBI_EXT_CPPC) > 0) {
+		pr_info("SBI CPPC extension detected\n");
+		cppc_ext_present = true;
+	} else {
+		pr_info("SBI CPPC extension NOT detected!!\n");
+		cppc_ext_present = false;
+	}
+
+	return 0;
+}
+device_initcall(sbi_cppc_init);
+
+static void sbi_cppc_read(void *read_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)read_data;
+
+	data->ret = sbi_ecall(SBI_EXT_CPPC, SBI_CPPC_READ,
+			      data->reg, 0, 0, 0, 0, 0);
+}
+
+static void sbi_cppc_write(void *write_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)write_data;
+
+	data->ret = sbi_ecall(SBI_EXT_CPPC, SBI_CPPC_WRITE,
+			      data->reg, data->val, 0, 0, 0, 0);
+}
+
+static void cppc_ffh_csr_read(void *read_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)read_data;
+
+	switch (data->reg) {
+	/* Support only TIME CSR for now */
+	case CSR_TIME:
+		data->ret.value = csr_read(CSR_TIME);
+		data->ret.error = 0;
+		break;
+	default:
+		data->ret.error = -EINVAL;
+		break;
+	}
+}
+
+static void cppc_ffh_csr_write(void *write_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)write_data;
+
+	data->ret.error = -EINVAL;
+}
+
+/*
+ * Refer to drivers/acpi/cppc_acpi.c for the description of the functions
+ * below.
+ */
+bool cpc_ffh_supported(void)
+{
+	return true;
+}
+
+int cpc_read_ffh(int cpu, struct cpc_reg *reg, u64 *val)
+{
+	struct sbi_cppc_data data;
+
+	if (WARN_ON_ONCE(irqs_disabled()))
+		return -EPERM;
+
+	if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_SBI) {
+		if (!cppc_ext_present)
+			return -EINVAL;
+
+		data.reg = FFH_CPPC_SBI_REG(reg->address);
+
+		smp_call_function_single(cpu, sbi_cppc_read, &data, 1);
+
+		*val = data.ret.value;
+
+		return (data.ret.error) ? sbi_err_map_linux_errno(data.ret.error) : 0;
+	} else if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_CSR) {
+		data.reg = FFH_CPPC_CSR_NUM(reg->address);
+
+		smp_call_function_single(cpu, cppc_ffh_csr_read, &data, 1);
+
+		*val = data.ret.value;
+
+		return (data.ret.error) ? sbi_err_map_linux_errno(data.ret.error) : 0;
+	}
+
+	return -EINVAL;
+}
+
+int cpc_write_ffh(int cpu, struct cpc_reg *reg, u64 val)
+{
+	struct sbi_cppc_data data;
+
+	if (WARN_ON_ONCE(irqs_disabled()))
+		return -EPERM;
+
+	if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_SBI) {
+		if (!cppc_ext_present)
+			return -EINVAL;
+
+		data.reg = FFH_CPPC_SBI_REG(reg->address);
+		data.val = val;
+
+		smp_call_function_single(cpu, sbi_cppc_write, &data, 1);
+
+		return (data.ret.error) ? sbi_err_map_linux_errno(data.ret.error) : 0;
+	} else if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_CSR) {
+		data.reg = FFH_CPPC_CSR_NUM(reg->address);
+		data.val = val;
+
+		smp_call_function_single(cpu, cppc_ffh_csr_write, &data, 1);
+
+		return (data.ret.error) ? sbi_err_map_linux_errno(data.ret.error) : 0;
+	}
+
+	return -EINVAL;
+}

drivers/acpi/riscv/cpuidle.c

@@ -0,0 +1,81 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2024, Ventana Micro Systems Inc
+ *	Author: Sunil V L <sunilvl@ventanamicro.com>
+ *
+ */
+
+#include <linux/acpi.h>
+#include <acpi/processor.h>
+#include <linux/cpu_pm.h>
+#include <linux/cpuidle.h>
+#include <linux/suspend.h>
+#include <asm/cpuidle.h>
+#include <asm/sbi.h>
+#include <asm/suspend.h>
+
+#define RISCV_FFH_LPI_TYPE_MASK	GENMASK_ULL(63, 60)
+#define RISCV_FFH_LPI_RSVD_MASK	GENMASK_ULL(59, 32)
+
+#define RISCV_FFH_LPI_TYPE_SBI	BIT_ULL(60)
+
+static int acpi_cpu_init_idle(unsigned int cpu)
+{
+	int i;
+	struct acpi_lpi_state *lpi;
+	struct acpi_processor *pr = per_cpu(processors, cpu);
+
+	if (unlikely(!pr || !pr->flags.has_lpi))
+		return -EINVAL;
+
+	if (!riscv_sbi_hsm_is_supported())
+		return -ENODEV;
+
+	if (pr->power.count <= 1)
+		return -ENODEV;
+
+	for (i = 1; i < pr->power.count; i++) {
+		u32 state;
+
+		lpi = &pr->power.lpi_states[i];
+
+		/*
+		 * Validate Entry Method as per FFH spec.
+		 * bits[63:60] should be 0x1
+		 * bits[59:32] should be 0x0
+		 * bits[31:0] represent a SBI power_state
+		 */
+		if (((lpi->address & RISCV_FFH_LPI_TYPE_MASK) != RISCV_FFH_LPI_TYPE_SBI) ||
+		    (lpi->address & RISCV_FFH_LPI_RSVD_MASK)) {
+			pr_warn("Invalid LPI entry method %#llx\n", lpi->address);
+			return -EINVAL;
+		}
+
+		state = lpi->address;
+		if (!riscv_sbi_suspend_state_is_valid(state)) {
+			pr_warn("Invalid SBI power state %#x\n", state);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+int acpi_processor_ffh_lpi_probe(unsigned int cpu)
+{
+	return acpi_cpu_init_idle(cpu);
+}
+
+int acpi_processor_ffh_lpi_enter(struct acpi_lpi_state *lpi)
+{
+	u32 state = lpi->address;
+
+	if (state & SBI_HSM_SUSP_NON_RET_BIT)
+		return CPU_PM_CPU_IDLE_ENTER_PARAM(riscv_sbi_hart_suspend,
+						   lpi->index,
+						   state);
+	else
+		return CPU_PM_CPU_IDLE_ENTER_RETENTION_PARAM(riscv_sbi_hart_suspend,
+							     lpi->index,
+							     state);
+}

drivers/clocksource/timer-clint.c

@@ -131,7 +131,7 @@ static int clint_timer_starting_cpu(unsigned int cpu)
 	struct clock_event_device *ce = per_cpu_ptr(&clint_clock_event, cpu);
 
 	ce->cpumask = cpumask_of(cpu);
-	clockevents_config_and_register(ce, clint_timer_freq, 100, 0x7fffffff);
+	clockevents_config_and_register(ce, clint_timer_freq, 100, ULONG_MAX);
 
 	enable_percpu_irq(clint_timer_irq,
 			  irq_get_trigger_type(clint_timer_irq));

drivers/clocksource/timer-riscv.c

@@ -114,7 +114,7 @@ static int riscv_timer_starting_cpu(unsigned int cpu)
 		ce->features |= CLOCK_EVT_FEAT_C3STOP;
 	if (static_branch_likely(&riscv_sstc_available))
 		ce->rating = 450;
-	clockevents_config_and_register(ce, riscv_timebase, 100, 0x7fffffff);
+	clockevents_config_and_register(ce, riscv_timebase, 100, ULONG_MAX);
 
 	enable_percpu_irq(riscv_clock_event_irq,
 			  irq_get_trigger_type(riscv_clock_event_irq));

drivers/cpufreq/Kconfig

@@ -302,4 +302,33 @@ config QORIQ_CPUFREQ
 	  which are capable of changing the CPU's frequency dynamically.
 
 endif
+
+config ACPI_CPPC_CPUFREQ
+	tristate "CPUFreq driver based on the ACPI CPPC spec"
+	depends on ACPI_PROCESSOR
+	depends on ARM || ARM64 || RISCV
+	select ACPI_CPPC_LIB
+	help
+	  This adds a CPUFreq driver which uses CPPC methods
+	  as described in the ACPIv5.1 spec. CPPC stands for
+	  Collaborative Processor Performance Controls. It
+	  is based on an abstract continuous scale of CPU
+	  performance values which allows the remote power
+	  processor to flexibly optimize for power and
+	  performance. CPPC relies on power management firmware
+	  support for its operation.
+
+	  If in doubt, say N.
+
+config ACPI_CPPC_CPUFREQ_FIE
+	bool "Frequency Invariance support for CPPC cpufreq driver"
+	depends on ACPI_CPPC_CPUFREQ && GENERIC_ARCH_TOPOLOGY
+	depends on ARM || ARM64 || RISCV
+	default y
+	help
+	  This extends frequency invariance support in the CPPC cpufreq driver,
+	  by using CPPC delivered and reference performance counters.
+
+	  If in doubt, say N.
+
 endmenu

drivers/cpufreq/Kconfig.arm

@@ -3,32 +3,6 @@
 # ARM CPU Frequency scaling drivers
 #
 
-config ACPI_CPPC_CPUFREQ
-	tristate "CPUFreq driver based on the ACPI CPPC spec"
-	depends on ACPI_PROCESSOR
-	select ACPI_CPPC_LIB
-	help
-	  This adds a CPUFreq driver which uses CPPC methods
-	  as described in the ACPIv5.1 spec. CPPC stands for
-	  Collaborative Processor Performance Controls. It
-	  is based on an abstract continuous scale of CPU
-	  performance values which allows the remote power
-	  processor to flexibly optimize for power and
-	  performance. CPPC relies on power management firmware
-	  support for its operation.
-
-	  If in doubt, say N.
-
-config ACPI_CPPC_CPUFREQ_FIE
-	bool "Frequency Invariance support for CPPC cpufreq driver"
-	depends on ACPI_CPPC_CPUFREQ && GENERIC_ARCH_TOPOLOGY
-	default y
-	help
-	  This extends frequency invariance support in the CPPC cpufreq driver,
-	  by using CPPC delivered and reference performance counters.
-
-	  If in doubt, say N.
-
 config ARM_ALLWINNER_SUN50I_CPUFREQ_NVMEM
 	tristate "Allwinner nvmem based SUN50I CPUFreq driver"
 	depends on ARCH_SUNXI

drivers/cpuidle/cpuidle-riscv-sbi.c

@@ -73,26 +73,6 @@ static inline bool sbi_is_domain_state_available(void)
 	return data->available;
 }
 
-static int sbi_suspend_finisher(unsigned long suspend_type,
-				unsigned long resume_addr,
-				unsigned long opaque)
-{
-	struct sbiret ret;
-
-	ret = sbi_ecall(SBI_EXT_HSM, SBI_EXT_HSM_HART_SUSPEND,
-			suspend_type, resume_addr, opaque, 0, 0, 0);
-
-	return (ret.error) ? sbi_err_map_linux_errno(ret.error) : 0;
-}
-
-static int sbi_suspend(u32 state)
-{
-	if (state & SBI_HSM_SUSP_NON_RET_BIT)
-		return cpu_suspend(state, sbi_suspend_finisher);
-	else
-		return sbi_suspend_finisher(state, 0, 0);
-}
-
 static __cpuidle int sbi_cpuidle_enter_state(struct cpuidle_device *dev,
 					     struct cpuidle_driver *drv, int idx)
 {
@@ -100,9 +80,9 @@ static __cpuidle int sbi_cpuidle_enter_state(struct cpuidle_device *dev,
 	u32 state = states[idx];
 
 	if (state & SBI_HSM_SUSP_NON_RET_BIT)
-		return CPU_PM_CPU_IDLE_ENTER_PARAM(sbi_suspend, idx, state);
+		return CPU_PM_CPU_IDLE_ENTER_PARAM(riscv_sbi_hart_suspend, idx, state);
 	else
-		return CPU_PM_CPU_IDLE_ENTER_RETENTION_PARAM(sbi_suspend,
+		return CPU_PM_CPU_IDLE_ENTER_RETENTION_PARAM(riscv_sbi_hart_suspend,
 							     idx, state);
 }
 
@@ -133,7 +113,7 @@ static __cpuidle int __sbi_enter_domain_idle_state(struct cpuidle_device *dev,
 	else
 		state = states[idx];
 
-	ret = sbi_suspend(state) ? -1 : idx;
+	ret = riscv_sbi_hart_suspend(state) ? -1 : idx;
 
 	ct_cpuidle_exit();
 
@@ -206,17 +186,6 @@ static const struct of_device_id sbi_cpuidle_state_match[] = {
 	{ },
 };
 
-static bool sbi_suspend_state_is_valid(u32 state)
-{
-	if (state > SBI_HSM_SUSPEND_RET_DEFAULT &&
-	    state < SBI_HSM_SUSPEND_RET_PLATFORM)
-		return false;
-	if (state > SBI_HSM_SUSPEND_NON_RET_DEFAULT &&
-	    state < SBI_HSM_SUSPEND_NON_RET_PLATFORM)
-		return false;
-	return true;
-}
-
 static int sbi_dt_parse_state_node(struct device_node *np, u32 *state)
 {
 	int err = of_property_read_u32(np, "riscv,sbi-suspend-param", state);
@@ -226,7 +195,7 @@ static int sbi_dt_parse_state_node(struct device_node *np, u32 *state)
 		return err;
 	}
 
-	if (!sbi_suspend_state_is_valid(*state)) {
+	if (!riscv_sbi_suspend_state_is_valid(*state)) {
 		pr_warn("Invalid SBI suspend state %#x\n", *state);
 		return -EINVAL;
 	}
@@ -607,16 +576,8 @@ static int __init sbi_cpuidle_init(void)
 	int ret;
 	struct platform_device *pdev;
 
-	/*
-	 * The SBI HSM suspend function is only available when:
-	 * 1) SBI version is 0.3 or higher
-	 * 2) SBI HSM extension is available
-	 */
-	if ((sbi_spec_version < sbi_mk_version(0, 3)) ||
-	    !sbi_probe_extension(SBI_EXT_HSM)) {
-		pr_info("HSM suspend not available\n");
+	if (!riscv_sbi_hsm_is_supported())
 		return 0;
-	}
 
 	ret = platform_driver_register(&sbi_cpuidle_driver);
 	if (ret)

drivers/perf/Kconfig

@@ -96,6 +96,20 @@ config STARFIVE_STARLINK_PMU
 	   an L3 memory system. The L3 cache events are added into perf event
 	   subsystem, allowing monitoring of various L3 cache perf events.
 
+config ANDES_CUSTOM_PMU
+	bool "Andes custom PMU support"
+	depends on ARCH_RENESAS && RISCV_ALTERNATIVE && RISCV_PMU_SBI
+	default y
+	help
+	  The Andes cores implement the PMU overflow extension very
+	  similar to the standard Sscofpmf and Smcntrpmf extension.
+
+	  This will patch the overflow and pending CSRs and handle the
+	  non-standard behaviour via the regular SBI PMU driver and
+	  interface.
+
+	  If you don't know what to do here, say "Y".
+
 config ARM_PMU_ACPI
 	depends on ARM_PMU && ACPI
 	def_bool y

drivers/perf/riscv_pmu_sbi.c

@@ -19,11 +19,33 @@
 #include <linux/of.h>
 #include <linux/cpu_pm.h>
 #include <linux/sched/clock.h>
+#include <linux/soc/andes/irq.h>
 
 #include <asm/errata_list.h>
 #include <asm/sbi.h>
 #include <asm/cpufeature.h>
 
+#define ALT_SBI_PMU_OVERFLOW(__ovl)					\
+asm volatile(ALTERNATIVE_2(						\
+	"csrr %0, " __stringify(CSR_SSCOUNTOVF),			\
+	"csrr %0, " __stringify(THEAD_C9XX_CSR_SCOUNTEROF),		\
+		THEAD_VENDOR_ID, ERRATA_THEAD_PMU,			\
+		CONFIG_ERRATA_THEAD_PMU,				\
+	"csrr %0, " __stringify(ANDES_CSR_SCOUNTEROF),			\
+		0, RISCV_ISA_EXT_XANDESPMU,				\
+		CONFIG_ANDES_CUSTOM_PMU)				\
+	: "=r" (__ovl) :						\
+	: "memory")
+
+#define ALT_SBI_PMU_OVF_CLEAR_PENDING(__irq_mask)			\
+asm volatile(ALTERNATIVE(						\
+	"csrc " __stringify(CSR_IP) ", %0\n\t",				\
+	"csrc " __stringify(ANDES_CSR_SLIP) ", %0\n\t",			\
+		0, RISCV_ISA_EXT_XANDESPMU,				\
+		CONFIG_ANDES_CUSTOM_PMU)				\
+	: : "r"(__irq_mask)						\
+	: "memory")
+
 #define SYSCTL_NO_USER_ACCESS	0
 #define SYSCTL_USER_ACCESS	1
 #define SYSCTL_LEGACY		2
@@ -61,6 +83,7 @@ static int sysctl_perf_user_access __read_mostly = SYSCTL_USER_ACCESS;
 static union sbi_pmu_ctr_info *pmu_ctr_list;
 static bool riscv_pmu_use_irq;
 static unsigned int riscv_pmu_irq_num;
+static unsigned int riscv_pmu_irq_mask;
 static unsigned int riscv_pmu_irq;
 
 /* Cache the available counters in a bitmask */
@@ -694,7 +717,7 @@ static irqreturn_t pmu_sbi_ovf_handler(int irq, void *dev)
 
 	event = cpu_hw_evt->events[fidx];
 	if (!event) {
-		csr_clear(CSR_SIP, BIT(riscv_pmu_irq_num));
+		ALT_SBI_PMU_OVF_CLEAR_PENDING(riscv_pmu_irq_mask);
 		return IRQ_NONE;
 	}
 
@@ -708,7 +731,7 @@ static irqreturn_t pmu_sbi_ovf_handler(int irq, void *dev)
 	 * Overflow interrupt pending bit should only be cleared after stopping
 	 * all the counters to avoid any race condition.
 	 */
-	csr_clear(CSR_SIP, BIT(riscv_pmu_irq_num));
+	ALT_SBI_PMU_OVF_CLEAR_PENDING(riscv_pmu_irq_mask);
 
 	/* No overflow bit is set */
 	if (!overflow)
@@ -780,8 +803,7 @@ static int pmu_sbi_starting_cpu(unsigned int cpu, struct hlist_node *node)
 
 	if (riscv_pmu_use_irq) {
 		cpu_hw_evt->irq = riscv_pmu_irq;
-		csr_clear(CSR_IP, BIT(riscv_pmu_irq_num));
-		csr_set(CSR_IE, BIT(riscv_pmu_irq_num));
+		ALT_SBI_PMU_OVF_CLEAR_PENDING(riscv_pmu_irq_mask);
 		enable_percpu_irq(riscv_pmu_irq, IRQ_TYPE_NONE);
 	}
 
@@ -792,7 +814,6 @@ static int pmu_sbi_dying_cpu(unsigned int cpu, struct hlist_node *node)
 {
 	if (riscv_pmu_use_irq) {
 		disable_percpu_irq(riscv_pmu_irq);
-		csr_clear(CSR_IE, BIT(riscv_pmu_irq_num));
 	}
 
 	/* Disable all counters access for user mode now */
@@ -816,8 +837,14 @@ static int pmu_sbi_setup_irqs(struct riscv_pmu *pmu, struct platform_device *pde
 		   riscv_cached_mimpid(0) == 0) {
 		riscv_pmu_irq_num = THEAD_C9XX_RV_IRQ_PMU;
 		riscv_pmu_use_irq = true;
+	} else if (riscv_isa_extension_available(NULL, XANDESPMU) &&
+		   IS_ENABLED(CONFIG_ANDES_CUSTOM_PMU)) {
+		riscv_pmu_irq_num = ANDES_SLI_CAUSE_BASE + ANDES_RV_IRQ_PMOVI;
+		riscv_pmu_use_irq = true;
 	}
 
+	riscv_pmu_irq_mask = BIT(riscv_pmu_irq_num % BITS_PER_LONG);
+
 	if (!riscv_pmu_use_irq)
 		return -EOPNOTSUPP;
 

include/asm-generic/bitops/__ffs.h

@@ -5,12 +5,12 @@
 #include <asm/types.h>
 
 /**
- * __ffs - find first bit in word.
+ * generic___ffs - find first bit in word.
  * @word: The word to search
  *
  * Undefined if no bit exists, so code should check against 0 first.
  */
-static __always_inline unsigned long __ffs(unsigned long word)
+static __always_inline unsigned long generic___ffs(unsigned long word)
 {
 	int num = 0;
 
@@ -41,4 +41,8 @@ static __always_inline unsigned long __ffs(unsigned long word)
 	return num;
 }
 
+#ifndef __HAVE_ARCH___FFS
+#define __ffs(word) generic___ffs(word)
+#endif
+
 #endif /* _ASM_GENERIC_BITOPS___FFS_H_ */

include/asm-generic/bitops/__fls.h

@@ -5,12 +5,12 @@
 #include <asm/types.h>
 
 /**
- * __fls - find last (most-significant) set bit in a long word
+ * generic___fls - find last (most-significant) set bit in a long word
  * @word: the word to search
  *
  * Undefined if no set bit exists, so code should check against 0 first.
  */
-static __always_inline unsigned long __fls(unsigned long word)
+static __always_inline unsigned long generic___fls(unsigned long word)
 {
 	int num = BITS_PER_LONG - 1;
 
@@ -41,4 +41,8 @@ static __always_inline unsigned long __fls(unsigned long word)
 	return num;
 }
 
+#ifndef __HAVE_ARCH___FLS
+#define __fls(word) generic___fls(word)
+#endif
+
 #endif /* _ASM_GENERIC_BITOPS___FLS_H_ */

include/asm-generic/bitops/ffs.h

@@ -3,14 +3,14 @@
 #define _ASM_GENERIC_BITOPS_FFS_H_
 
 /**
- * ffs - find first bit set
+ * generic_ffs - find first bit set
  * @x: the word to search
  *
  * This is defined the same way as
  * the libc and compiler builtin ffs routines, therefore
  * differs in spirit from ffz (man ffs).
  */
-static inline int ffs(int x)
+static inline int generic_ffs(int x)
 {
 	int r = 1;
 
@@ -39,4 +39,8 @@ static inline int ffs(int x)
 	return r;
 }
 
+#ifndef __HAVE_ARCH_FFS
+#define ffs(x) generic_ffs(x)
+#endif
+
 #endif /* _ASM_GENERIC_BITOPS_FFS_H_ */

include/asm-generic/bitops/fls.h

@@ -3,14 +3,14 @@
 #define _ASM_GENERIC_BITOPS_FLS_H_
 
 /**
- * fls - find last (most-significant) bit set
+ * generic_fls - find last (most-significant) bit set
  * @x: the word to search
  *
  * This is defined the same way as ffs.
  * Note fls(0) = 0, fls(1) = 1, fls(0x80000000) = 32.
  */
 
-static __always_inline int fls(unsigned int x)
+static __always_inline int generic_fls(unsigned int x)
 {
 	int r = 32;
 
@@ -39,4 +39,8 @@ static __always_inline int fls(unsigned int x)
 	return r;
 }
 
+#ifndef __HAVE_ARCH_FLS
+#define fls(x) generic_fls(x)
+#endif
+
 #endif /* _ASM_GENERIC_BITOPS_FLS_H_ */

include/linux/mm.h

@@ -87,7 +87,7 @@ extern int sysctl_legacy_va_layout;
 
 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
 extern const int mmap_rnd_bits_min;
-extern const int mmap_rnd_bits_max;
+extern int mmap_rnd_bits_max __ro_after_init;
 extern int mmap_rnd_bits __read_mostly;
 #endif
 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS

include/linux/sync_core.h

@@ -17,5 +17,19 @@ static inline void sync_core_before_usermode(void)
 }
 #endif
 
-#endif /* _LINUX_SYNC_CORE_H */
+#ifdef CONFIG_ARCH_HAS_PREPARE_SYNC_CORE_CMD
+#include <asm/sync_core.h>
+#else
+/*
+ * This is a dummy prepare_sync_core_cmd() implementation that can be used on
+ * all architectures which provide unconditional core serializing instructions
+ * in switch_mm().
+ * If your architecture doesn't provide such core serializing instructions in
+ * switch_mm(), you may need to write your own functions.
+ */
+static inline void prepare_sync_core_cmd(struct mm_struct *mm)
+{
+}
+#endif
 
+#endif /* _LINUX_SYNC_CORE_H */

init/Kconfig

@@ -1986,6 +1986,9 @@ source "kernel/Kconfig.locks"
 config ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	bool
 
+config ARCH_HAS_PREPARE_SYNC_CORE_CMD
+	bool
+
 config ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
 	bool
 

kernel/sched/core.c

@@ -6647,7 +6647,9 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 	 *     if (signal_pending_state())	    if (p->state & @state)
 	 *
 	 * Also, the membarrier system call requires a full memory barrier
-	 * after coming from user-space, before storing to rq->curr.
+	 * after coming from user-space, before storing to rq->curr; this
+	 * barrier matches a full barrier in the proximity of the membarrier
+	 * system call exit.
 	 */
 	rq_lock(rq, &rf);
 	smp_mb__after_spinlock();
@@ -6718,12 +6720,20 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 		 *
 		 * Here are the schemes providing that barrier on the
 		 * various architectures:
-		 * - mm ? switch_mm() : mmdrop() for x86, s390, sparc, PowerPC.
-		 *   switch_mm() rely on membarrier_arch_switch_mm() on PowerPC.
+		 * - mm ? switch_mm() : mmdrop() for x86, s390, sparc, PowerPC,
+		 *   RISC-V.  switch_mm() relies on membarrier_arch_switch_mm()
+		 *   on PowerPC and on RISC-V.
 		 * - finish_lock_switch() for weakly-ordered
 		 *   architectures where spin_unlock is a full barrier,
 		 * - switch_to() for arm64 (weakly-ordered, spin_unlock
 		 *   is a RELEASE barrier),
+		 *
+		 * The barrier matches a full barrier in the proximity of
+		 * the membarrier system call entry.
+		 *
+		 * On RISC-V, this barrier pairing is also needed for the
+		 * SYNC_CORE command when switching between processes, cf.
+		 * the inline comments in membarrier_arch_switch_mm().
 		 */
 		++*switch_count;
 

kernel/sched/membarrier.c

@@ -254,7 +254,7 @@ static int membarrier_global_expedited(void)
 		return 0;
 
 	/*
-	 * Matches memory barriers around rq->curr modification in
+	 * Matches memory barriers after rq->curr modification in
 	 * scheduler.
 	 */
 	smp_mb();	/* system call entry is not a mb. */
@@ -304,7 +304,7 @@ static int membarrier_global_expedited(void)
 
 	/*
 	 * Memory barrier on the caller thread _after_ we finished
-	 * waiting for the last IPI. Matches memory barriers around
+	 * waiting for the last IPI. Matches memory barriers before
 	 * rq->curr modification in scheduler.
 	 */
 	smp_mb();	/* exit from system call is not a mb */
@@ -324,6 +324,7 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 		      MEMBARRIER_STATE_PRIVATE_EXPEDITED_SYNC_CORE_READY))
 			return -EPERM;
 		ipi_func = ipi_sync_core;
+		prepare_sync_core_cmd(mm);
 	} else if (flags == MEMBARRIER_FLAG_RSEQ) {
 		if (!IS_ENABLED(CONFIG_RSEQ))
 			return -EINVAL;
@@ -343,8 +344,12 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 		return 0;
 
 	/*
-	 * Matches memory barriers around rq->curr modification in
+	 * Matches memory barriers after rq->curr modification in
 	 * scheduler.
+	 *
+	 * On RISC-V, this barrier pairing is also needed for the
+	 * SYNC_CORE command when switching between processes, cf.
+	 * the inline comments in membarrier_arch_switch_mm().
 	 */
 	smp_mb();	/* system call entry is not a mb. */
 
@@ -420,7 +425,7 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 
 	/*
 	 * Memory barrier on the caller thread _after_ we finished
-	 * waiting for the last IPI. Matches memory barriers around
+	 * waiting for the last IPI. Matches memory barriers before
 	 * rq->curr modification in scheduler.
 	 */
 	smp_mb();	/* exit from system call is not a mb */

mm/mmap.c

@@ -64,7 +64,7 @@
 
 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
 const int mmap_rnd_bits_min = CONFIG_ARCH_MMAP_RND_BITS_MIN;
-const int mmap_rnd_bits_max = CONFIG_ARCH_MMAP_RND_BITS_MAX;
+int mmap_rnd_bits_max __ro_after_init = CONFIG_ARCH_MMAP_RND_BITS_MAX;
 int mmap_rnd_bits __read_mostly = CONFIG_ARCH_MMAP_RND_BITS;
 #endif
 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS

tools/perf/pmu-events/arch/riscv/andes/ax45/firmware.json

@@ -0,0 +1,68 @@
+[
+  {
+    "ArchStdEvent": "FW_MISALIGNED_LOAD"
+  },
+  {
+    "ArchStdEvent": "FW_MISALIGNED_STORE"
+  },
+  {
+    "ArchStdEvent": "FW_ACCESS_LOAD"
+  },
+  {
+    "ArchStdEvent": "FW_ACCESS_STORE"
+  },
+  {
+    "ArchStdEvent": "FW_ILLEGAL_INSN"
+  },
+  {
+    "ArchStdEvent": "FW_SET_TIMER"
+  },
+  {
+    "ArchStdEvent": "FW_IPI_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_IPI_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_FENCE_I_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_FENCE_I_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_SFENCE_VMA_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_SFENCE_VMA_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_SFENCE_VMA_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_SFENCE_VMA_ASID_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_GVMA_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_GVMA_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_GVMA_VMID_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_GVMA_VMID_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_VVMA_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_VVMA_RECEIVED"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_VVMA_ASID_SENT"
+  },
+  {
+    "ArchStdEvent": "FW_HFENCE_VVMA_ASID_RECEIVED"
+  }
+]

tools/perf/pmu-events/arch/riscv/andes/ax45/instructions.json

@@ -0,0 +1,127 @@
+[
+	{
+		"EventCode": "0x10",
+		"EventName": "cycle_count",
+		"BriefDescription": "Cycle count"
+	},
+	{
+		"EventCode": "0x20",
+		"EventName": "inst_count",
+		"BriefDescription": "Retired instruction count"
+	},
+	{
+		"EventCode": "0x30",
+		"EventName": "int_load_inst",
+		"BriefDescription": "Integer load instruction count"
+	},
+	{
+		"EventCode": "0x40",
+		"EventName": "int_store_inst",
+		"BriefDescription": "Integer store instruction count"
+	},
+	{
+		"EventCode": "0x50",
+		"EventName": "atomic_inst",
+		"BriefDescription": "Atomic instruction count"
+	},
+	{
+		"EventCode": "0x60",
+		"EventName": "sys_inst",
+		"BriefDescription": "System instruction count"
+	},
+	{
+		"EventCode": "0x70",
+		"EventName": "int_compute_inst",
+		"BriefDescription": "Integer computational instruction count"
+	},
+	{
+		"EventCode": "0x80",
+		"EventName": "condition_br",
+		"BriefDescription": "Conditional branch instruction count"
+	},
+	{
+		"EventCode": "0x90",
+		"EventName": "taken_condition_br",
+		"BriefDescription": "Taken conditional branch instruction count"
+	},
+	{
+		"EventCode": "0xA0",
+		"EventName": "jal_inst",
+		"BriefDescription": "JAL instruction count"
+	},
+	{
+		"EventCode": "0xB0",
+		"EventName": "jalr_inst",
+		"BriefDescription": "JALR instruction count"
+	},
+	{
+		"EventCode": "0xC0",
+		"EventName": "ret_inst",
+		"BriefDescription": "Return instruction count"
+	},
+	{
+		"EventCode": "0xD0",
+		"EventName": "control_trans_inst",
+		"BriefDescription": "Control transfer instruction count"
+	},
+	{
+		"EventCode": "0xE0",
+		"EventName": "ex9_inst",
+		"BriefDescription": "EXEC.IT instruction count"
+	},
+	{
+		"EventCode": "0xF0",
+		"EventName": "int_mul_inst",
+		"BriefDescription": "Integer multiplication instruction count"
+	},
+	{
+		"EventCode": "0x100",
+		"EventName": "int_div_rem_inst",
+		"BriefDescription": "Integer division/remainder instruction count"
+	},
+	{
+		"EventCode": "0x110",
+		"EventName": "float_load_inst",
+		"BriefDescription": "Floating-point load instruction count"
+	},
+	{
+		"EventCode": "0x120",
+		"EventName": "float_store_inst",
+		"BriefDescription": "Floating-point store instruction count"
+	},
+	{
+		"EventCode": "0x130",
+		"EventName": "float_add_sub_inst",
+		"BriefDescription": "Floating-point addition/subtraction instruction count"
+	},
+	{
+		"EventCode": "0x140",
+		"EventName": "float_mul_inst",
+		"BriefDescription": "Floating-point multiplication instruction count"
+	},
+	{
+		"EventCode": "0x150",
+		"EventName": "float_fused_muladd_inst",
+		"BriefDescription": "Floating-point fused multiply-add instruction count"
+	},
+	{
+		"EventCode": "0x160",
+		"EventName": "float_div_sqrt_inst",
+		"BriefDescription": "Floating-point division or square-root instruction count"
+	},
+	{
+		"EventCode": "0x170",
+		"EventName": "other_float_inst",
+		"BriefDescription": "Other floating-point instruction count"
+	},
+	{
+		"EventCode": "0x180",
+		"EventName": "int_mul_add_sub_inst",
+		"BriefDescription": "Integer multiplication and add/sub instruction count"
+	},
+	{
+		"EventCode": "0x190",
+		"EventName": "retired_ops",
+		"BriefDescription": "Retired operation count"
+	}
+]

tools/perf/pmu-events/arch/riscv/andes/ax45/memory.json

@@ -0,0 +1,57 @@
+[
+	{
+		"EventCode": "0x01",
+		"EventName": "ilm_access",
+		"BriefDescription": "ILM access"
+	},
+	{
+		"EventCode": "0x11",
+		"EventName": "dlm_access",
+		"BriefDescription": "DLM access"
+	},
+	{
+		"EventCode": "0x21",
+		"EventName": "icache_access",
+		"BriefDescription": "ICACHE access"
+	},
+	{
+		"EventCode": "0x31",
+		"EventName": "icache_miss",
+		"BriefDescription": "ICACHE miss"
+	},
+	{
+		"EventCode": "0x41",
+		"EventName": "dcache_access",
+		"BriefDescription": "DCACHE access"
+	},
+	{
+		"EventCode": "0x51",
+		"EventName": "dcache_miss",
+		"BriefDescription": "DCACHE miss"
+	},
+	{
+		"EventCode": "0x61",
+		"EventName": "dcache_load_access",
+		"BriefDescription": "DCACHE load access"
+	},
+	{
+		"EventCode": "0x71",
+		"EventName": "dcache_load_miss",
+		"BriefDescription": "DCACHE load miss"
+	},
+	{
+		"EventCode": "0x81",
+		"EventName": "dcache_store_access",
+		"BriefDescription": "DCACHE store access"
+	},
+	{
+		"EventCode": "0x91",
+		"EventName": "dcache_store_miss",
+		"BriefDescription": "DCACHE store miss"
+	},
+	{
+		"EventCode": "0xA1",
+		"EventName": "dcache_wb",
+		"BriefDescription": "DCACHE writeback"
+	}
+]