diff --git a/Documentation/core-api/floating-point.rst b/Documentation/core-api/floating-point.rst
new file mode 100644
index 000000000000..a8d0d4b05052
--- /dev/null
+++ b/Documentation/core-api/floating-point.rst
@@ -0,0 +1,78 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+Floating-point API
+==================
+
+Kernel code is normally prohibited from using floating-point (FP) registers or
+instructions, including the C float and double data types. This rule reduces
+system call overhead, because the kernel does not need to save and restore the
+userspace floating-point register state.
+
+However, occasionally drivers or library functions may need to include FP code.
+This is supported by isolating the functions containing FP code to a separate
+translation unit (a separate source file), and saving/restoring the FP register
+state around calls to those functions. This creates "critical sections" of
+floating-point usage.
+
+The reason for this isolation is to prevent the compiler from generating code
+touching the FP registers outside these critical sections. Compilers sometimes
+use FP registers to optimize inlined ``memcpy`` or variable assignment, as
+floating-point registers may be wider than general-purpose registers.
+
+Usability of floating-point code within the kernel is architecture-specific.
+Additionally, because a single kernel may be configured to support platforms
+both with and without a floating-point unit, FPU availability must be checked
+both at build time and at run time.
+
+Several architectures implement the generic kernel floating-point API from
+``linux/fpu.h``, as described below. Some other architectures implement their
+own unique APIs, which are documented separately.
+
+Build-time API
+--------------
+
+Floating-point code may be built if the option ``ARCH_HAS_KERNEL_FPU_SUPPORT``
+is enabled. For C code, such code must be placed in a separate file, and that
+file must have its compilation flags adjusted using the following pattern::
+
+    CFLAGS_foo.o += $(CC_FLAGS_FPU)
+    CFLAGS_REMOVE_foo.o += $(CC_FLAGS_NO_FPU)
+
+Architectures are expected to define one or both of these variables in their
+top-level Makefile as needed. For example::
+
+    CC_FLAGS_FPU := -mhard-float
+
+or::
+
+    CC_FLAGS_NO_FPU := -msoft-float
+
+Normal kernel code is assumed to use the equivalent of ``CC_FLAGS_NO_FPU``.
+
+Runtime API
+-----------
+
+The runtime API is provided in ``linux/fpu.h``. This header cannot be included
+from files implementing FP code (those with their compilation flags adjusted as
+above). Instead, it must be included when defining the FP critical sections.
+
+.. c:function:: bool kernel_fpu_available( void )
+
+        This function reports if floating-point code can be used on this CPU or
+        platform. The value returned by this function is not expected to change
+        at runtime, so it only needs to be called once, not before every
+        critical section.
+
+.. c:function:: void kernel_fpu_begin( void )
+                void kernel_fpu_end( void )
+
+        These functions create a floating-point critical section. It is only
+        valid to call ``kernel_fpu_begin()`` after a previous call to
+        ``kernel_fpu_available()`` returned ``true``. These functions are only
+        guaranteed to be callable from (preemptible or non-preemptible) process
+        context.
+
+        Preemption may be disabled inside critical sections, so their size
+        should be minimized. They are *not* required to be reentrant. If the
+        caller expects to nest critical sections, it must implement its own
+        reference counting.
diff --git a/Documentation/core-api/index.rst b/Documentation/core-api/index.rst
index 7a3a08d81f11..974beccd671f 100644
--- a/Documentation/core-api/index.rst
+++ b/Documentation/core-api/index.rst
@@ -48,6 +48,7 @@ Library functionality that is used throughout the kernel.
    errseq
    wrappers/atomic_t
    wrappers/atomic_bitops
+   floating-point
 
 Low level entry and exit
 ========================
diff --git a/Makefile b/Makefile
index b7198af9e59b..c8993b8dd605 100644
--- a/Makefile
+++ b/Makefile
@@ -981,6 +981,11 @@ KBUILD_CFLAGS	+= $(CC_FLAGS_CFI)
 export CC_FLAGS_CFI
 endif
 
+# Architectures can define flags to add/remove for floating-point support
+CC_FLAGS_FPU	+= -D_LINUX_FPU_COMPILATION_UNIT
+export CC_FLAGS_FPU
+export CC_FLAGS_NO_FPU
+
 ifneq ($(CONFIG_FUNCTION_ALIGNMENT),0)
 KBUILD_CFLAGS += -falign-functions=$(CONFIG_FUNCTION_ALIGNMENT)
 endif
diff --git a/arch/Kconfig b/arch/Kconfig
index 20c2c93d2c88..603eb47c6268 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -1480,6 +1480,12 @@ config ARCH_HAS_NONLEAF_PMD_YOUNG
 	  address translations. Page table walkers that clear the accessed bit
 	  may use this capability to reduce their search space.
 
+config ARCH_HAS_KERNEL_FPU_SUPPORT
+	bool
+	help
+	  Architectures that select this option can run floating-point code in
+	  the kernel, as described in Documentation/core-api/floating-point.rst.
+
 source "kernel/gcov/Kconfig"
 
 source "scripts/gcc-plugins/Kconfig"
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index 443f7566c9d1..eacbc4efe60c 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -29,6 +29,7 @@ config RISCV
 	select ARCH_HAS_KCOV
 	select ARCH_HAS_MEMBARRIER_CALLBACKS
 	select ARCH_HAS_MEMBARRIER_SYNC_CORE
+	select ARCH_HAS_KERNEL_FPU_SUPPORT if 64BIT && FPU
 	select ARCH_HAS_MMIOWB
 	select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	select ARCH_HAS_PMEM_API
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
index b43a6bb7e4dc..aaf7b9bd2275 100644
--- a/arch/riscv/Makefile
+++ b/arch/riscv/Makefile
@@ -77,6 +77,9 @@ KBUILD_CFLAGS += -march=$(shell echo $(riscv-march-y) | sed -E 's/(rv32ima|rv64i
 
 KBUILD_AFLAGS += -march=$(riscv-march-y)
 
+# For C code built with floating-point support, exclude V but keep F and D.
+CC_FLAGS_FPU  := -march=$(shell echo $(riscv-march-y) | sed -E 's/(rv32ima|rv64ima)([^v_]*)v?/\1\2/')
+
 KBUILD_CFLAGS += -mno-save-restore
 KBUILD_CFLAGS += -DCONFIG_PAGE_OFFSET=$(CONFIG_PAGE_OFFSET)
 
diff --git a/arch/riscv/include/asm/fpu.h b/arch/riscv/include/asm/fpu.h
new file mode 100644
index 000000000000..91c04c244e12
--- /dev/null
+++ b/arch/riscv/include/asm/fpu.h
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2023 SiFive
+ */
+
+#ifndef _ASM_RISCV_FPU_H
+#define _ASM_RISCV_FPU_H
+
+#include <asm/switch_to.h>
+
+#define kernel_fpu_available()	has_fpu()
+
+void kernel_fpu_begin(void);
+void kernel_fpu_end(void);
+
+#endif /* ! _ASM_RISCV_FPU_H */
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 95cf25d48405..a90f90eb0578 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -61,6 +61,7 @@ obj-$(CONFIG_MMU) += vdso.o vdso/
 
 obj-$(CONFIG_RISCV_M_MODE)	+= traps_misaligned.o
 obj-$(CONFIG_FPU)		+= fpu.o
+obj-$(CONFIG_FPU)		+= kernel_mode_fpu.o
 obj-$(CONFIG_RISCV_ISA_V)	+= vector.o
 obj-$(CONFIG_SMP)		+= smpboot.o
 obj-$(CONFIG_SMP)		+= smp.o
diff --git a/arch/riscv/kernel/kernel_mode_fpu.c b/arch/riscv/kernel/kernel_mode_fpu.c
new file mode 100644
index 000000000000..0ac8348876c4
--- /dev/null
+++ b/arch/riscv/kernel/kernel_mode_fpu.c
@@ -0,0 +1,28 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2023 SiFive
+ */
+
+#include <linux/export.h>
+#include <linux/preempt.h>
+
+#include <asm/csr.h>
+#include <asm/fpu.h>
+#include <asm/processor.h>
+#include <asm/switch_to.h>
+
+void kernel_fpu_begin(void)
+{
+	preempt_disable();
+	fstate_save(current, task_pt_regs(current));
+	csr_set(CSR_SSTATUS, SR_FS);
+}
+EXPORT_SYMBOL_GPL(kernel_fpu_begin);
+
+void kernel_fpu_end(void)
+{
+	csr_clear(CSR_SSTATUS, SR_FS);
+	fstate_restore(current, task_pt_regs(current));
+	preempt_enable();
+}
+EXPORT_SYMBOL_GPL(kernel_fpu_end);
diff --git a/arch/riscv/kernel/module-sections.c b/arch/riscv/kernel/module-sections.c
index e264e59e596e..a191ba7f8892 100644
--- a/arch/riscv/kernel/module-sections.c
+++ b/arch/riscv/kernel/module-sections.c
@@ -9,6 +9,7 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/moduleloader.h>
+#include <linux/sort.h>
 
 unsigned long module_emit_got_entry(struct module *mod, unsigned long val)
 {
@@ -55,19 +56,27 @@ unsigned long module_emit_plt_entry(struct module *mod, unsigned long val)
 	return (unsigned long)&plt[i];
 }
 
-static int is_rela_equal(const Elf_Rela *x, const Elf_Rela *y)
+#define cmp_3way(a, b)	((a) < (b) ? -1 : (a) > (b))
+
+static int cmp_rela(const void *a, const void *b)
 {
-	return x->r_info == y->r_info && x->r_addend == y->r_addend;
+	const Elf_Rela *x = a, *y = b;
+	int i;
+
+	/* sort by type, symbol index and addend */
+	i = cmp_3way(x->r_info, y->r_info);
+	if (i == 0)
+		i = cmp_3way(x->r_addend, y->r_addend);
+	return i;
 }
 
 static bool duplicate_rela(const Elf_Rela *rela, int idx)
 {
-	int i;
-	for (i = 0; i < idx; i++) {
-		if (is_rela_equal(&rela[i], &rela[idx]))
-			return true;
-	}
-	return false;
+	/*
+	 * Entries are sorted by type, symbol index and addend. That means
+	 * that, if a duplicate entry exists, it must be in the preceding slot.
+	 */
+	return idx > 0 && cmp_rela(rela + idx, rela + idx - 1) == 0;
 }
 
 static void count_max_entries(Elf_Rela *relas, int num,
@@ -87,11 +96,33 @@ static void count_max_entries(Elf_Rela *relas, int num,
 	}
 }
 
+static bool rela_needs_plt_got(const Elf_Rela *rela)
+{
+	unsigned int type = ELF_R_TYPE(rela->r_info);
+
+	return type == R_RISCV_CALL_PLT || type == R_RISCV_GOT_HI20;
+}
+
+/* Copy PLT and GOT relas to the scratch array. */
+static unsigned int partition_plt_got_relas(const Elf_Rela *relas, Elf_Rela *scratch,
+					    unsigned int num_rela)
+{
+	int j = 0;
+
+	for (int i = 0; i < num_rela; i++)
+		if (rela_needs_plt_got(&relas[i]))
+			scratch[j++] = relas[i];
+
+	return j;
+}
+
 int module_frob_arch_sections(Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
 			      char *secstrings, struct module *mod)
 {
 	unsigned int num_plts = 0;
 	unsigned int num_gots = 0;
+	Elf_Rela *scratch = NULL;
+	size_t scratch_size = 0;
 	int i;
 
 	/*
@@ -132,9 +163,26 @@ int module_frob_arch_sections(Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
 		if (!(dst_sec->sh_flags & SHF_EXECINSTR))
 			continue;
 
-		count_max_entries(relas, num_rela, &num_plts, &num_gots);
+		/*
+		 * apply_relocate_add() relies on HI20 and LO12 relocation pairs being
+		 * close together, so sort a copy of the section to avoid interfering.
+		 */
+		if (sechdrs[i].sh_size > scratch_size) {
+			scratch = kvrealloc(scratch, scratch_size, sechdrs[i].sh_size, GFP_KERNEL);
+			if (!scratch)
+				return -ENOMEM;
+			scratch_size = sechdrs[i].sh_size;
+		}
+
+		/* sort relocations requiring a PLT or GOT entry so duplicates are adjacent */
+		num_rela = partition_plt_got_relas(relas, scratch, num_rela);
+		sort(scratch, num_rela, sizeof(Elf_Rela), cmp_rela, NULL);
+		count_max_entries(scratch, num_rela, &num_plts, &num_gots);
 	}
 
+	if (scratch)
+		kvfree(scratch);
+
 	mod->arch.plt.shdr->sh_type = SHT_NOBITS;
 	mod->arch.plt.shdr->sh_flags = SHF_EXECINSTR | SHF_ALLOC;
 	mod->arch.plt.shdr->sh_addralign = L1_CACHE_BYTES;
diff --git a/drivers/gpu/drm/amd/display/Kconfig b/drivers/gpu/drm/amd/display/Kconfig
index 901d1961b739..5fcd4f778dc3 100644
--- a/drivers/gpu/drm/amd/display/Kconfig
+++ b/drivers/gpu/drm/amd/display/Kconfig
@@ -8,7 +8,7 @@ config DRM_AMD_DC
 	depends on BROKEN || !CC_IS_CLANG || ARM64 || RISCV || SPARC64 || X86_64
 	select SND_HDA_COMPONENT if SND_HDA_CORE
 	# !CC_IS_CLANG: https://github.com/ClangBuiltLinux/linux/issues/1752
-	select DRM_AMD_DC_FP if (X86 || LOONGARCH || (PPC64 && ALTIVEC) || (ARM64 && KERNEL_MODE_NEON && !CC_IS_CLANG))
+	select DRM_AMD_DC_FP if ARCH_HAS_KERNEL_FPU_SUPPORT && (!ARM64 || !CC_IS_CLANG)
 	help
 	  Choose this option if you want to use the new display engine
 	  support for AMDGPU. This adds required support for Vega and
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c b/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c
index 172aa10a8800..e46f8ce41d87 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c
@@ -26,16 +26,7 @@
 
 #include "dc_trace.h"
 
-#if defined(CONFIG_X86)
-#include <asm/fpu/api.h>
-#elif defined(CONFIG_PPC64)
-#include <asm/switch_to.h>
-#include <asm/cputable.h>
-#elif defined(CONFIG_ARM64)
-#include <asm/neon.h>
-#elif defined(CONFIG_LOONGARCH)
-#include <asm/fpu.h>
-#endif
+#include <linux/fpu.h>
 
 /**
  * DOC: DC FPU manipulation overview
@@ -60,11 +51,9 @@ static DEFINE_PER_CPU(int, fpu_recursion_depth);
  */
 inline void dc_assert_fp_enabled(void)
 {
-	int *pcpu, depth = 0;
+	int depth;
 
-	pcpu = get_cpu_ptr(&fpu_recursion_depth);
-	depth = *pcpu;
-	put_cpu_ptr(&fpu_recursion_depth);
+	depth = __this_cpu_read(fpu_recursion_depth);
 
 	ASSERT(depth >= 1);
 }
@@ -84,33 +73,17 @@ inline void dc_assert_fp_enabled(void)
  */
 void dc_fpu_begin(const char *function_name, const int line)
 {
-	int *pcpu;
+	int depth;
 
-	pcpu = get_cpu_ptr(&fpu_recursion_depth);
-	*pcpu += 1;
-
-	if (*pcpu == 1) {
-#if defined(CONFIG_X86) || defined(CONFIG_LOONGARCH)
-		migrate_disable();
+	WARN_ON_ONCE(!in_task());
+	preempt_disable();
+	depth = __this_cpu_inc_return(fpu_recursion_depth);
+	if (depth == 1) {
+		BUG_ON(!kernel_fpu_available());
 		kernel_fpu_begin();
-#elif defined(CONFIG_PPC64)
-		if (cpu_has_feature(CPU_FTR_VSX_COMP)) {
-			preempt_disable();
-			enable_kernel_vsx();
-		} else if (cpu_has_feature(CPU_FTR_ALTIVEC_COMP)) {
-			preempt_disable();
-			enable_kernel_altivec();
-		} else if (!cpu_has_feature(CPU_FTR_FPU_UNAVAILABLE)) {
-			preempt_disable();
-			enable_kernel_fp();
-		}
-#elif defined(CONFIG_ARM64)
-		kernel_neon_begin();
-#endif
 	}
 
-	TRACE_DCN_FPU(true, function_name, line, *pcpu);
-	put_cpu_ptr(&fpu_recursion_depth);
+	TRACE_DCN_FPU(true, function_name, line, depth);
 }
 
 /**
@@ -125,30 +98,15 @@ void dc_fpu_begin(const char *function_name, const int line)
  */
 void dc_fpu_end(const char *function_name, const int line)
 {
-	int *pcpu;
+	int depth;
 
-	pcpu = get_cpu_ptr(&fpu_recursion_depth);
-	*pcpu -= 1;
-	if (*pcpu <= 0) {
-#if defined(CONFIG_X86) || defined(CONFIG_LOONGARCH)
+	depth = __this_cpu_dec_return(fpu_recursion_depth);
+	if (depth == 0) {
 		kernel_fpu_end();
-		migrate_enable();
-#elif defined(CONFIG_PPC64)
-		if (cpu_has_feature(CPU_FTR_VSX_COMP)) {
-			disable_kernel_vsx();
-			preempt_enable();
-		} else if (cpu_has_feature(CPU_FTR_ALTIVEC_COMP)) {
-			disable_kernel_altivec();
-			preempt_enable();
-		} else if (!cpu_has_feature(CPU_FTR_FPU_UNAVAILABLE)) {
-			disable_kernel_fp();
-			preempt_enable();
-		}
-#elif defined(CONFIG_ARM64)
-		kernel_neon_end();
-#endif
+	} else {
+		WARN_ON_ONCE(depth < 0);
 	}
 
-	TRACE_DCN_FPU(false, function_name, line, *pcpu);
-	put_cpu_ptr(&fpu_recursion_depth);
+	TRACE_DCN_FPU(false, function_name, line, depth);
+	preempt_enable();
 }
diff --git a/drivers/gpu/drm/amd/display/dc/dml/Makefile b/drivers/gpu/drm/amd/display/dc/dml/Makefile
index 0ba9a7997d56..2d04de698dd6 100644
--- a/drivers/gpu/drm/amd/display/dc/dml/Makefile
+++ b/drivers/gpu/drm/amd/display/dc/dml/Makefile
@@ -25,40 +25,8 @@
 # It provides the general basic services required by other DAL
 # subcomponents.
 
-ifdef CONFIG_X86
-dml_ccflags-$(CONFIG_CC_IS_GCC) := -mhard-float
-dml_ccflags := $(dml_ccflags-y) -msse
-endif
-
-ifdef CONFIG_PPC64
-dml_ccflags := -mhard-float -maltivec
-endif
-
-ifdef CONFIG_ARM64
-dml_rcflags := -mgeneral-regs-only
-endif
-
-ifdef CONFIG_LOONGARCH
-dml_ccflags := -mfpu=64
-dml_rcflags := -msoft-float
-endif
-
-ifdef CONFIG_CC_IS_GCC
-ifneq ($(call gcc-min-version, 70100),y)
-IS_OLD_GCC = 1
-endif
-endif
-
-ifdef CONFIG_X86
-ifdef IS_OLD_GCC
-# Stack alignment mismatch, proceed with caution.
-# GCC < 7.1 cannot compile code using `double` and -mpreferred-stack-boundary=3
-# (8B stack alignment).
-dml_ccflags += -mpreferred-stack-boundary=4
-else
-dml_ccflags += -msse2
-endif
-endif
+dml_ccflags := $(CC_FLAGS_FPU)
+dml_rcflags := $(CC_FLAGS_NO_FPU)
 
 ifneq ($(CONFIG_FRAME_WARN),0)
 ifeq ($(filter y,$(CONFIG_KASAN)$(CONFIG_KCSAN)),y)
diff --git a/include/linux/fpu.h b/include/linux/fpu.h
new file mode 100644
index 000000000000..2fb63e22913b
--- /dev/null
+++ b/include/linux/fpu.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef _LINUX_FPU_H
+#define _LINUX_FPU_H
+
+#ifdef _LINUX_FPU_COMPILATION_UNIT
+#error FP code must be compiled separately. See Documentation/core-api/floating-point.rst.
+#endif
+
+#include <asm/fpu.h>
+
+#endif