LoongArch: Add hardware breakpoints/watchpoints support

Use perf framework to manage hardware instruction and data breakpoints.

LoongArch defines hardware watchpoint functions for instruction fetch
and memory load/store operations. After the software configures hardware
watchpoints, the processor hardware will monitor the access address of
the instruction fetch and load/store operation, and trigger an exception
of the watchpoint when it meets the conditions set by the watchpoint.

The hardware monitoring points for instruction fetching and load/store
operations each have a register for the overall configuration of all
monitoring points, a register for recording the status of all monitoring
points, and four registers required for configuration of each watchpoint
individually.
Signed-off-by: Qing Zhang <zhangqing@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>

arch/loongarch/Kconfig

@@ -101,6 +101,7 @@ config LOONGARCH
 	select HAVE_FUNCTION_GRAPH_TRACER
 	select HAVE_FUNCTION_TRACER
 	select HAVE_GENERIC_VDSO
+	select HAVE_HW_BREAKPOINT if PERF_EVENTS
 	select HAVE_IOREMAP_PROT
 	select HAVE_IRQ_EXIT_ON_IRQ_STACK
 	select HAVE_IRQ_TIME_ACCOUNTING

arch/loongarch/include/asm/hw_breakpoint.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2022-2023 Loongson Technology Corporation Limited
+ */
+#ifndef __ASM_HW_BREAKPOINT_H
+#define __ASM_HW_BREAKPOINT_H
+
+#include <asm/loongarch.h>
+
+#ifdef __KERNEL__
+
+/* Breakpoint */
+#define LOONGARCH_BREAKPOINT_EXECUTE		(0 << 0)
+
+/* Watchpoints */
+#define LOONGARCH_BREAKPOINT_LOAD		(1 << 0)
+#define LOONGARCH_BREAKPOINT_STORE		(1 << 1)
+
+struct arch_hw_breakpoint_ctrl {
+	u32 __reserved	: 28,
+	len		: 2,
+	type		: 2;
+};
+
+struct arch_hw_breakpoint {
+	u64 address;
+	u64 mask;
+	struct arch_hw_breakpoint_ctrl ctrl;
+};
+
+/* Lengths */
+#define LOONGARCH_BREAKPOINT_LEN_1    0b11
+#define LOONGARCH_BREAKPOINT_LEN_2    0b10
+#define LOONGARCH_BREAKPOINT_LEN_4    0b01
+#define LOONGARCH_BREAKPOINT_LEN_8    0b00
+
+/*
+ * Limits.
+ * Changing these will require modifications to the register accessors.
+ */
+#define LOONGARCH_MAX_BRP		8
+#define LOONGARCH_MAX_WRP		8
+
+/* Virtual debug register bases. */
+#define CSR_CFG_ADDR	0
+#define CSR_CFG_MASK	(CSR_CFG_ADDR + LOONGARCH_MAX_BRP)
+#define CSR_CFG_CTRL	(CSR_CFG_MASK + LOONGARCH_MAX_BRP)
+#define CSR_CFG_ASID	(CSR_CFG_CTRL + LOONGARCH_MAX_WRP)
+
+/* Debug register names. */
+#define LOONGARCH_CSR_NAME_ADDR	ADDR
+#define LOONGARCH_CSR_NAME_MASK	MASK
+#define LOONGARCH_CSR_NAME_CTRL	CTRL
+#define LOONGARCH_CSR_NAME_ASID	ASID
+
+/* Accessor macros for the debug registers. */
+#define LOONGARCH_CSR_WATCH_READ(N, REG, T, VAL)			\
+do {								\
+	if (T == 0)						\
+		VAL = csr_read64(LOONGARCH_CSR_##IB##N##REG);	\
+	else							\
+		VAL = csr_read64(LOONGARCH_CSR_##DB##N##REG);	\
+} while (0)
+
+#define LOONGARCH_CSR_WATCH_WRITE(N, REG, T, VAL)			\
+do {								\
+	if (T == 0)						\
+		csr_write64(VAL, LOONGARCH_CSR_##IB##N##REG);	\
+	else							\
+		csr_write64(VAL, LOONGARCH_CSR_##DB##N##REG);	\
+} while (0)
+
+/* Exact number */
+#define CSR_FWPC_NUM		0x3f
+#define CSR_MWPC_NUM		0x3f
+
+#define CTRL_PLV_ENABLE		0x1e
+
+#define MWPnCFG3_LoadEn		8
+#define MWPnCFG3_StoreEn	9
+
+#define MWPnCFG3_Type_mask	0x3
+#define MWPnCFG3_Size_mask	0x3
+
+static inline u32 encode_ctrl_reg(struct arch_hw_breakpoint_ctrl ctrl)
+{
+	return (ctrl.len << 10) | (ctrl.type << 8);
+}
+
+static inline void decode_ctrl_reg(u32 reg, struct arch_hw_breakpoint_ctrl *ctrl)
+{
+	reg >>= 8;
+	ctrl->type = reg & MWPnCFG3_Type_mask;
+	reg >>= 2;
+	ctrl->len  = reg & MWPnCFG3_Size_mask;
+}
+
+struct task_struct;
+struct notifier_block;
+struct perf_event;
+struct perf_event_attr;
+
+extern int arch_bp_generic_fields(struct arch_hw_breakpoint_ctrl ctrl,
+				  int *gen_len, int *gen_type, int *offset);
+extern int arch_check_bp_in_kernelspace(struct arch_hw_breakpoint *hw);
+extern int hw_breakpoint_arch_parse(struct perf_event *bp,
+				    const struct perf_event_attr *attr,
+				    struct arch_hw_breakpoint *hw);
+extern int hw_breakpoint_exceptions_notify(struct notifier_block *unused,
+					   unsigned long val, void *data);
+
+extern int arch_install_hw_breakpoint(struct perf_event *bp);
+extern void arch_uninstall_hw_breakpoint(struct perf_event *bp);
+extern int hw_breakpoint_slots(int type);
+extern void hw_breakpoint_pmu_read(struct perf_event *bp);
+
+void breakpoint_handler(struct pt_regs *regs);
+void watchpoint_handler(struct pt_regs *regs);
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+extern void ptrace_hw_copy_thread(struct task_struct *task);
+extern void hw_breakpoint_thread_switch(struct task_struct *next);
+#else
+static inline void ptrace_hw_copy_thread(struct task_struct *task)
+{
+}
+static inline void hw_breakpoint_thread_switch(struct task_struct *next)
+{
+}
+#endif
+
+/* Determine number of BRP registers available. */
+static inline int get_num_brps(void)
+{
+	return csr_read64(LOONGARCH_CSR_FWPC) & CSR_FWPC_NUM;
+}
+
+/* Determine number of WRP registers available. */
+static inline int get_num_wrps(void)
+{
+	return csr_read64(LOONGARCH_CSR_MWPC) & CSR_MWPC_NUM;
+}
+
+#endif	/* __KERNEL__ */
+#endif	/* __ASM_BREAKPOINT_H */

arch/loongarch/include/asm/loongarch.h

@@ -970,42 +970,42 @@ static __always_inline void iocsr_write64(u64 val, u32 reg)
 
 #define LOONGARCH_CSR_DB0ADDR		0x310	/* data breakpoint 0 address */
 #define LOONGARCH_CSR_DB0MASK		0x311	/* data breakpoint 0 mask */
-#define LOONGARCH_CSR_DB0CTL		0x312	/* data breakpoint 0 control */
+#define LOONGARCH_CSR_DB0CTRL		0x312	/* data breakpoint 0 control */
 #define LOONGARCH_CSR_DB0ASID		0x313	/* data breakpoint 0 asid */
 
 #define LOONGARCH_CSR_DB1ADDR		0x318	/* data breakpoint 1 address */
 #define LOONGARCH_CSR_DB1MASK		0x319	/* data breakpoint 1 mask */
-#define LOONGARCH_CSR_DB1CTL		0x31a	/* data breakpoint 1 control */
+#define LOONGARCH_CSR_DB1CTRL		0x31a	/* data breakpoint 1 control */
 #define LOONGARCH_CSR_DB1ASID		0x31b	/* data breakpoint 1 asid */
 
 #define LOONGARCH_CSR_DB2ADDR		0x320	/* data breakpoint 2 address */
 #define LOONGARCH_CSR_DB2MASK		0x321	/* data breakpoint 2 mask */
-#define LOONGARCH_CSR_DB2CTL		0x322	/* data breakpoint 2 control */
+#define LOONGARCH_CSR_DB2CTRL		0x322	/* data breakpoint 2 control */
 #define LOONGARCH_CSR_DB2ASID		0x323	/* data breakpoint 2 asid */
 
 #define LOONGARCH_CSR_DB3ADDR		0x328	/* data breakpoint 3 address */
 #define LOONGARCH_CSR_DB3MASK		0x329	/* data breakpoint 3 mask */
-#define LOONGARCH_CSR_DB3CTL		0x32a	/* data breakpoint 3 control */
+#define LOONGARCH_CSR_DB3CTRL		0x32a	/* data breakpoint 3 control */
 #define LOONGARCH_CSR_DB3ASID		0x32b	/* data breakpoint 3 asid */
 
 #define LOONGARCH_CSR_DB4ADDR		0x330	/* data breakpoint 4 address */
 #define LOONGARCH_CSR_DB4MASK		0x331	/* data breakpoint 4 maks */
-#define LOONGARCH_CSR_DB4CTL		0x332	/* data breakpoint 4 control */
+#define LOONGARCH_CSR_DB4CTRL		0x332	/* data breakpoint 4 control */
 #define LOONGARCH_CSR_DB4ASID		0x333	/* data breakpoint 4 asid */
 
 #define LOONGARCH_CSR_DB5ADDR		0x338	/* data breakpoint 5 address */
 #define LOONGARCH_CSR_DB5MASK		0x339	/* data breakpoint 5 mask */
-#define LOONGARCH_CSR_DB5CTL		0x33a	/* data breakpoint 5 control */
+#define LOONGARCH_CSR_DB5CTRL		0x33a	/* data breakpoint 5 control */
 #define LOONGARCH_CSR_DB5ASID		0x33b	/* data breakpoint 5 asid */
 
 #define LOONGARCH_CSR_DB6ADDR		0x340	/* data breakpoint 6 address */
 #define LOONGARCH_CSR_DB6MASK		0x341	/* data breakpoint 6 mask */
-#define LOONGARCH_CSR_DB6CTL		0x342	/* data breakpoint 6 control */
+#define LOONGARCH_CSR_DB6CTRL		0x342	/* data breakpoint 6 control */
 #define LOONGARCH_CSR_DB6ASID		0x343	/* data breakpoint 6 asid */
 
 #define LOONGARCH_CSR_DB7ADDR		0x348	/* data breakpoint 7 address */
 #define LOONGARCH_CSR_DB7MASK		0x349	/* data breakpoint 7 mask */
-#define LOONGARCH_CSR_DB7CTL		0x34a	/* data breakpoint 7 control */
+#define LOONGARCH_CSR_DB7CTRL		0x34a	/* data breakpoint 7 control */
 #define LOONGARCH_CSR_DB7ASID		0x34b	/* data breakpoint 7 asid */
 
 #define LOONGARCH_CSR_FWPC		0x380	/* instruction breakpoint config */
@@ -1013,42 +1013,42 @@ static __always_inline void iocsr_write64(u64 val, u32 reg)
 
 #define LOONGARCH_CSR_IB0ADDR		0x390	/* inst breakpoint 0 address */
 #define LOONGARCH_CSR_IB0MASK		0x391	/* inst breakpoint 0 mask */
-#define LOONGARCH_CSR_IB0CTL		0x392	/* inst breakpoint 0 control */
+#define LOONGARCH_CSR_IB0CTRL		0x392	/* inst breakpoint 0 control */
 #define LOONGARCH_CSR_IB0ASID		0x393	/* inst breakpoint 0 asid */
 
 #define LOONGARCH_CSR_IB1ADDR		0x398	/* inst breakpoint 1 address */
 #define LOONGARCH_CSR_IB1MASK		0x399	/* inst breakpoint 1 mask */
-#define LOONGARCH_CSR_IB1CTL		0x39a	/* inst breakpoint 1 control */
+#define LOONGARCH_CSR_IB1CTRL		0x39a	/* inst breakpoint 1 control */
 #define LOONGARCH_CSR_IB1ASID		0x39b	/* inst breakpoint 1 asid */
 
 #define LOONGARCH_CSR_IB2ADDR		0x3a0	/* inst breakpoint 2 address */
 #define LOONGARCH_CSR_IB2MASK		0x3a1	/* inst breakpoint 2 mask */
-#define LOONGARCH_CSR_IB2CTL		0x3a2	/* inst breakpoint 2 control */
+#define LOONGARCH_CSR_IB2CTRL		0x3a2	/* inst breakpoint 2 control */
 #define LOONGARCH_CSR_IB2ASID		0x3a3	/* inst breakpoint 2 asid */
 
 #define LOONGARCH_CSR_IB3ADDR		0x3a8	/* inst breakpoint 3 address */
 #define LOONGARCH_CSR_IB3MASK		0x3a9	/* breakpoint 3 mask */
-#define LOONGARCH_CSR_IB3CTL		0x3aa	/* inst breakpoint 3 control */
+#define LOONGARCH_CSR_IB3CTRL		0x3aa	/* inst breakpoint 3 control */
 #define LOONGARCH_CSR_IB3ASID		0x3ab	/* inst breakpoint 3 asid */
 
 #define LOONGARCH_CSR_IB4ADDR		0x3b0	/* inst breakpoint 4 address */
 #define LOONGARCH_CSR_IB4MASK		0x3b1	/* inst breakpoint 4 mask */
-#define LOONGARCH_CSR_IB4CTL		0x3b2	/* inst breakpoint 4 control */
+#define LOONGARCH_CSR_IB4CTRL		0x3b2	/* inst breakpoint 4 control */
 #define LOONGARCH_CSR_IB4ASID		0x3b3	/* inst breakpoint 4 asid */
 
 #define LOONGARCH_CSR_IB5ADDR		0x3b8	/* inst breakpoint 5 address */
 #define LOONGARCH_CSR_IB5MASK		0x3b9	/* inst breakpoint 5 mask */
-#define LOONGARCH_CSR_IB5CTL		0x3ba	/* inst breakpoint 5 control */
+#define LOONGARCH_CSR_IB5CTRL		0x3ba	/* inst breakpoint 5 control */
 #define LOONGARCH_CSR_IB5ASID		0x3bb	/* inst breakpoint 5 asid */
 
 #define LOONGARCH_CSR_IB6ADDR		0x3c0	/* inst breakpoint 6 address */
 #define LOONGARCH_CSR_IB6MASK		0x3c1	/* inst breakpoint 6 mask */
-#define LOONGARCH_CSR_IB6CTL		0x3c2	/* inst breakpoint 6 control */
+#define LOONGARCH_CSR_IB6CTRL		0x3c2	/* inst breakpoint 6 control */
 #define LOONGARCH_CSR_IB6ASID		0x3c3	/* inst breakpoint 6 asid */
 
 #define LOONGARCH_CSR_IB7ADDR		0x3c8	/* inst breakpoint 7 address */
 #define LOONGARCH_CSR_IB7MASK		0x3c9	/* inst breakpoint 7 mask */
-#define LOONGARCH_CSR_IB7CTL		0x3ca	/* inst breakpoint 7 control */
+#define LOONGARCH_CSR_IB7CTRL		0x3ca	/* inst breakpoint 7 control */
 #define LOONGARCH_CSR_IB7ASID		0x3cb	/* inst breakpoint 7 asid */
 
 #define LOONGARCH_CSR_DEBUG		0x500	/* debug config */

arch/loongarch/include/asm/processor.h

@@ -11,6 +11,7 @@
 
 #include <asm/cpu.h>
 #include <asm/cpu-info.h>
+#include <asm/hw_breakpoint.h>
 #include <asm/loongarch.h>
 #include <asm/vdso/processor.h>
 #include <uapi/asm/ptrace.h>
@@ -127,10 +128,14 @@ struct thread_struct {
 	struct loongarch_vdso_info *vdso;
 
 	/*
-	 * FPU & vector registers, must be at last because
-	 * they are conditionally copied at fork().
+	 * FPU & vector registers, must be at the last of inherited
+	 * context because they are conditionally copied at fork().
 	 */
 	struct loongarch_fpu fpu FPU_ALIGN;
+
+	/* Hardware breakpoints pinned to this task. */
+	struct perf_event *hbp_break[LOONGARCH_MAX_BRP];
+	struct perf_event *hbp_watch[LOONGARCH_MAX_WRP];
 };
 
 #define thread_saved_ra(tsk)	(tsk->thread.sched_ra)
@@ -172,6 +177,8 @@ struct thread_struct {
 		.fcc		= 0,				\
 		.fpr		= {{{0,},},},			\
 	},							\
+	.hbp_break		= {0},				\
+	.hbp_watch		= {0},				\
 }
 
 struct task_struct;
@@ -184,10 +191,6 @@ extern unsigned long		boot_option_idle_override;
  */
 extern void start_thread(struct pt_regs *regs, unsigned long pc, unsigned long sp);
 
-static inline void flush_thread(void)
-{
-}
-
 unsigned long __get_wchan(struct task_struct *p);
 
 #define __KSTK_TOS(tsk) ((unsigned long)task_stack_page(tsk) + \

arch/loongarch/include/asm/switch_to.h

@@ -34,6 +34,7 @@ extern asmlinkage struct task_struct *__switch_to(struct task_struct *prev,
 #define switch_to(prev, next, last)						\
 do {										\
 	lose_fpu_inatomic(1, prev);						\
+	hw_breakpoint_thread_switch(next);					\
 	(last) = __switch_to(prev, next, task_thread_info(next),		\
 		 __builtin_return_address(0), __builtin_frame_address(0));	\
 } while (0)

arch/loongarch/kernel/Makefile

@@ -50,5 +50,6 @@ obj-$(CONFIG_UNWINDER_GUESS)	+= unwind_guess.o
 obj-$(CONFIG_UNWINDER_PROLOGUE) += unwind_prologue.o
 
 obj-$(CONFIG_PERF_EVENTS)	+= perf_event.o perf_regs.o
+obj-$(CONFIG_HAVE_HW_BREAKPOINT)	+= hw_breakpoint.o
 
 CPPFLAGS_vmlinux.lds		:= $(KBUILD_CFLAGS)

arch/loongarch/kernel/process.c

@@ -18,6 +18,7 @@
 #include <linux/sched/debug.h>
 #include <linux/sched/task.h>
 #include <linux/sched/task_stack.h>
+#include <linux/hw_breakpoint.h>
 #include <linux/mm.h>
 #include <linux/stddef.h>
 #include <linux/unistd.h>
@@ -96,6 +97,11 @@ void start_thread(struct pt_regs *regs, unsigned long pc, unsigned long sp)
 	regs->regs[3] = sp;
 }
 
+void flush_thread(void)
+{
+	flush_ptrace_hw_breakpoint(current);
+}
+
 void exit_thread(struct task_struct *tsk)
 {
 }
@@ -181,6 +187,7 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
 		childregs->regs[2] = tls;
 
 out:
+	ptrace_hw_copy_thread(p);
 	clear_tsk_thread_flag(p, TIF_USEDFPU);
 	clear_tsk_thread_flag(p, TIF_USEDSIMD);
 	clear_tsk_thread_flag(p, TIF_LSX_CTX_LIVE);

arch/loongarch/kernel/traps.c

@@ -511,7 +511,17 @@ asmlinkage void noinstr do_bp(struct pt_regs *regs)
 
 asmlinkage void noinstr do_watch(struct pt_regs *regs)
 {
+	irqentry_state_t state = irqentry_enter(regs);
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	breakpoint_handler(regs);
+	watchpoint_handler(regs);
+	force_sig(SIGTRAP);
+#else
 	pr_warn("Hardware watch point handler not implemented!\n");
+#endif
+
+	irqentry_exit(regs, state);
 }
 
 asmlinkage void noinstr do_ri(struct pt_regs *regs)