Merge patch series "Add support to handle misaligned accesses in S-mode"

Clément Léger <cleger@rivosinc.com> says:

Since commit 61cadb9 ("Provide new description of misaligned load/store
behavior compatible with privileged architecture.") in the RISC-V ISA
manual, it is stated that misaligned load/store might not be supported.
However, the RISC-V kernel uABI describes that misaligned accesses are
supported. In order to support that, this series adds support for S-mode
handling of misaligned accesses as well support for prctl(PR_UNALIGN).

Handling misaligned access in kernel allows for a finer grain control
of the misaligned accesses behavior, and thanks to the prctl() call,
can allow disabling misaligned access emulation to generate SIGBUS. User
space can then optimize its software by removing such access based on
SIGBUS generation.

This series is useful when using a SBI implementation that does not
handle misaligned traps as well as detecting misaligned accesses
generated by userspace application using the prctrl(PR_SET_UNALIGN)
feature.

This series can be tested using the spike simulator[1] and a modified
openSBI version[2] which allows to always delegate misaligned load/store to
S-mode. A test[3] that exercise various instructions/registers can be
executed to verify the unaligned access support.

[1] https://github.com/riscv-software-src/riscv-isa-sim
[2] https://github.com/rivosinc/opensbi/tree/dev/cleger/no_misaligned
[3] https://github.com/clementleger/unaligned_test

* b4-shazam-merge:
  riscv: add support for PR_SET_UNALIGN and PR_GET_UNALIGN
  riscv: report misaligned accesses emulation to hwprobe
  riscv: annotate check_unaligned_access_boot_cpu() with __init
  riscv: add support for sysctl unaligned_enabled control
  riscv: add floating point insn support to misaligned access emulation
  riscv: report perf event for misaligned fault
  riscv: add support for misaligned trap handling in S-mode
  riscv: remove unused functions in traps_misaligned.c

[Palmer: Evan pointed out an extra __init on
unaligned_emulation_finish(), I've squashed it in as a merge conflict.]

Link: https://lore.kernel.org/r/20231004151405.521596-1-cleger@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>

arch/riscv/Kconfig

@@ -643,6 +643,15 @@ config THREAD_SIZE_ORDER
 	  Specify the Pages of thread stack size (from 4KB to 64KB), which also
 	  affects irq stack size, which is equal to thread stack size.
 
+config RISCV_MISALIGNED
+	bool "Support misaligned load/store traps for kernel and userspace"
+	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.
+
 endmenu # "Platform type"
 
 menu "Kernel features"

arch/riscv/include/asm/cpufeature.h

@@ -33,4 +33,22 @@ extern struct riscv_isainfo hart_isa[NR_CPUS];
 int check_unaligned_access(void *unused);
 void riscv_user_isa_enable(void);
 
+#ifdef CONFIG_RISCV_MISALIGNED
+bool unaligned_ctl_available(void);
+bool check_unaligned_access_emulated(int cpu);
+void unaligned_emulation_finish(void);
+#else
+static inline bool unaligned_ctl_available(void)
+{
+	return false;
+}
+
+static inline bool check_unaligned_access_emulated(int cpu)
+{
+	return false;
+}
+
+static inline void unaligned_emulation_finish(void) {}
+#endif
+
 #endif

arch/riscv/include/asm/entry-common.h

@@ -8,4 +8,18 @@
 void handle_page_fault(struct pt_regs *regs);
 void handle_break(struct pt_regs *regs);
 
+#ifdef CONFIG_RISCV_MISALIGNED
+int handle_misaligned_load(struct pt_regs *regs);
+int handle_misaligned_store(struct pt_regs *regs);
+#else
+static inline int handle_misaligned_load(struct pt_regs *regs)
+{
+	return -1;
+}
+static inline int handle_misaligned_store(struct pt_regs *regs)
+{
+	return -1;
+}
+#endif
+
 #endif /* _ASM_RISCV_ENTRY_COMMON_H */

arch/riscv/include/asm/processor.h

@@ -8,6 +8,7 @@
 
 #include <linux/const.h>
 #include <linux/cache.h>
+#include <linux/prctl.h>
 
 #include <vdso/processor.h>
 
@@ -82,6 +83,7 @@ struct thread_struct {
 	unsigned long bad_cause;
 	unsigned long vstate_ctrl;
 	struct __riscv_v_ext_state vstate;
+	unsigned long align_ctl;
 };
 
 /* Whitelist the fstate from the task_struct for hardened usercopy */
@@ -94,6 +96,7 @@ static inline void arch_thread_struct_whitelist(unsigned long *offset,
 
 #define INIT_THREAD {					\
 	.sp = sizeof(init_stack) + (long)&init_stack,	\
+	.align_ctl = PR_UNALIGN_NOPRINT,		\
 }
 
 #define task_pt_regs(tsk)						\
@@ -134,6 +137,12 @@ extern long riscv_v_vstate_ctrl_set_current(unsigned long arg);
 extern long riscv_v_vstate_ctrl_get_current(void);
 #endif /* CONFIG_RISCV_ISA_V */
 
+extern int get_unalign_ctl(struct task_struct *tsk, unsigned long addr);
+extern int set_unalign_ctl(struct task_struct *tsk, unsigned int val);
+
+#define GET_UNALIGN_CTL(tsk, addr)	get_unalign_ctl((tsk), (addr))
+#define SET_UNALIGN_CTL(tsk, val)	set_unalign_ctl((tsk), (val))
+
 #endif /* __ASSEMBLY__ */
 
 #endif /* _ASM_RISCV_PROCESSOR_H */

arch/riscv/kernel/Makefile

@@ -59,7 +59,7 @@ obj-y	+= patch.o
 obj-y	+= probes/
 obj-$(CONFIG_MMU) += vdso.o vdso/
 
-obj-$(CONFIG_RISCV_M_MODE)	+= traps_misaligned.o
+obj-$(CONFIG_RISCV_MISALIGNED)	+= traps_misaligned.o
 obj-$(CONFIG_FPU)		+= fpu.o
 obj-$(CONFIG_RISCV_ISA_V)	+= vector.o
 obj-$(CONFIG_SMP)		+= smpboot.o

arch/riscv/kernel/cpufeature.c

@@ -570,6 +570,9 @@ int check_unaligned_access(void *unused)
 	void *src;
 	long speed = RISCV_HWPROBE_MISALIGNED_SLOW;
 
+	if (check_unaligned_access_emulated(cpu))
+		return 0;
+
 	page = alloc_pages(GFP_NOWAIT, get_order(MISALIGNED_BUFFER_SIZE));
 	if (!page) {
 		pr_warn("Can't alloc pages to measure memcpy performance");
@@ -645,6 +648,7 @@ int check_unaligned_access(void *unused)
 
 out:
 	__free_pages(page, get_order(MISALIGNED_BUFFER_SIZE));
+	unaligned_emulation_finish();
 	return 0;
 }
 

arch/riscv/kernel/fpu.S

@@ -104,3 +104,124 @@ ENTRY(__fstate_restore)
 	csrc CSR_STATUS, t1
 	ret
 ENDPROC(__fstate_restore)
+
+#define get_f32(which) fmv.x.s a0, which; j 2f
+#define put_f32(which) fmv.s.x which, a1; j 2f
+#if __riscv_xlen == 64
+# define get_f64(which) fmv.x.d a0, which; j 2f
+# define put_f64(which) fmv.d.x which, a1; j 2f
+#else
+# define get_f64(which) fsd which, 0(a1); j 2f
+# define put_f64(which) fld which, 0(a1); j 2f
+#endif
+
+.macro fp_access_prologue
+	/*
+	 * Compute jump offset to store the correct FP register since we don't
+	 * have indirect FP register access
+	 */
+	sll t0, a0, 3
+	la t2, 1f
+	add t0, t0, t2
+	li t1, SR_FS
+	csrs CSR_STATUS, t1
+	jr t0
+1:
+.endm
+
+.macro fp_access_epilogue
+2:
+	csrc CSR_STATUS, t1
+	ret
+.endm
+
+#define fp_access_body(__access_func) \
+	__access_func(f0); \
+	__access_func(f1); \
+	__access_func(f2); \
+	__access_func(f3); \
+	__access_func(f4); \
+	__access_func(f5); \
+	__access_func(f6); \
+	__access_func(f7); \
+	__access_func(f8); \
+	__access_func(f9); \
+	__access_func(f10); \
+	__access_func(f11); \
+	__access_func(f12); \
+	__access_func(f13); \
+	__access_func(f14); \
+	__access_func(f15); \
+	__access_func(f16); \
+	__access_func(f17); \
+	__access_func(f18); \
+	__access_func(f19); \
+	__access_func(f20); \
+	__access_func(f21); \
+	__access_func(f22); \
+	__access_func(f23); \
+	__access_func(f24); \
+	__access_func(f25); \
+	__access_func(f26); \
+	__access_func(f27); \
+	__access_func(f28); \
+	__access_func(f29); \
+	__access_func(f30); \
+	__access_func(f31)
+
+
+#ifdef CONFIG_RISCV_MISALIGNED
+
+/*
+ * Disable compressed instructions set to keep a constant offset between FP
+ * load/store/move instructions
+ */
+.option norvc
+/*
+ * put_f32_reg - Set a FP register from a register containing the value
+ * a0 = FP register index to be set
+ * a1 = value to be loaded in the FP register
+ */
+SYM_FUNC_START(put_f32_reg)
+	fp_access_prologue
+	fp_access_body(put_f32)
+	fp_access_epilogue
+SYM_FUNC_END(put_f32_reg)
+
+/*
+ * get_f32_reg - Get a FP register value and return it
+ * a0 = FP register index to be retrieved
+ */
+SYM_FUNC_START(get_f32_reg)
+	fp_access_prologue
+	fp_access_body(get_f32)
+	fp_access_epilogue
+SYM_FUNC_END(get_f32_reg)
+
+/*
+ * put_f64_reg - Set a 64 bits FP register from a value or a pointer.
+ * a0 = FP register index to be set
+ * a1 = value/pointer to be loaded in the FP register (when xlen == 32 bits, we
+ * load the value to a pointer).
+ */
+SYM_FUNC_START(put_f64_reg)
+	fp_access_prologue
+	fp_access_body(put_f64)
+	fp_access_epilogue
+SYM_FUNC_END(put_f64_reg)
+
+/*
+ * put_f64_reg - Get a 64 bits FP register value and returned it or store it to
+ *	 	 a pointer.
+ * a0 = FP register index to be retrieved
+ * a1 = If xlen == 32, pointer which should be loaded with the FP register value
+ *	or unused if xlen == 64. In which case the FP register value is returned
+ *	through a0
+ */
+SYM_FUNC_START(get_f64_reg)
+	fp_access_prologue
+	fp_access_body(get_f64)
+	fp_access_epilogue
+SYM_FUNC_END(get_f64_reg)
+
+#endif /* CONFIG_RISCV_MISALIGNED */

arch/riscv/kernel/process.c

@@ -25,6 +25,7 @@
 #include <asm/thread_info.h>
 #include <asm/cpuidle.h>
 #include <asm/vector.h>
+#include <asm/cpufeature.h>
 
 register unsigned long gp_in_global __asm__("gp");
 
@@ -41,6 +42,23 @@ void arch_cpu_idle(void)
 	cpu_do_idle();
 }
 
+int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
+{
+	if (!unaligned_ctl_available())
+		return -EINVAL;
+
+	tsk->thread.align_ctl = val;
+	return 0;
+}
+
+int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
+{
+	if (!unaligned_ctl_available())
+		return -EINVAL;
+
+	return put_user(tsk->thread.align_ctl, (unsigned long __user *)adr);
+}
+
 void __show_regs(struct pt_regs *regs)
 {
 	show_regs_print_info(KERN_DEFAULT);

arch/riscv/kernel/smpboot.c

@@ -248,7 +248,6 @@ asmlinkage __visible void smp_callin(void)
 	riscv_ipi_enable();
 
 	numa_add_cpu(curr_cpuid);
-	set_cpu_online(curr_cpuid, 1);
 
 	/*
 	 * Boot-time misaligned access speed measurements are done in parallel
@@ -258,6 +257,7 @@ asmlinkage __visible void smp_callin(void)
 	    (per_cpu(misaligned_access_speed, curr_cpuid) == RISCV_HWPROBE_MISALIGNED_UNKNOWN)) {
 		check_unaligned_access(NULL);
 	}
+	set_cpu_online(curr_cpuid, 1);
 
 	if (has_vector()) {
 		if (riscv_v_setup_vsize())

arch/riscv/kernel/traps.c

@@ -179,14 +179,6 @@ asmlinkage __visible __trap_section void do_trap_insn_illegal(struct pt_regs *re
 
 DO_ERROR_INFO(do_trap_load_fault,
 	SIGSEGV, SEGV_ACCERR, "load access fault");
-#ifndef CONFIG_RISCV_M_MODE
-DO_ERROR_INFO(do_trap_load_misaligned,
-	SIGBUS, BUS_ADRALN, "Oops - load address misaligned");
-DO_ERROR_INFO(do_trap_store_misaligned,
-	SIGBUS, BUS_ADRALN, "Oops - store (or AMO) address misaligned");
-#else
-int handle_misaligned_load(struct pt_regs *regs);
-int handle_misaligned_store(struct pt_regs *regs);
 
 asmlinkage __visible __trap_section void do_trap_load_misaligned(struct pt_regs *regs)
 {
@@ -229,7 +221,6 @@ asmlinkage __visible __trap_section void do_trap_store_misaligned(struct pt_regs
 		irqentry_nmi_exit(regs, state);
 	}
 }
-#endif
 DO_ERROR_INFO(do_trap_store_fault,
 	SIGSEGV, SEGV_ACCERR, "store (or AMO) access fault");
 DO_ERROR_INFO(do_trap_ecall_s,