hw_breakpoint 是由处理器提供专门断点寄存器来保存一个地址，是需要处理器支持的。处理器在执行过程中会不断去匹配，当匹配上后则会产生中断。

内核自带了硬件断点的样例linux-3.16\samples\hw_breakpoint\data_breakpoint.c

static void sample_hbp_handler(struct perf_event *bp,struct perf_sample_data *data,struct pt_regs *regs)
{printk(KERN_INFO "%s value is changed\n", ksym_name);dump_stack();printk(KERN_INFO "Dump stack from sample_hbp_handler\n");
}static int __init hw_break_module_init(void* addr)
{int ret;struct perf_event_attr attr;hw_breakpoint_init(&attr);attr.bp_addr = addr;attr.bp_len = HW_BREAKPOINT_LEN_4;//监控addr开始的4字节attr.bp_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R;//读写该地址都能触发sample_hbp = register_wide_hw_breakpoint(&attr, sample_hbp_handler, NULL);if (IS_ERR((void __force *)sample_hbp)) {ret = PTR_ERR((void __force *)sample_hbp);goto fail;}printk(KERN_INFO "HW Breakpoint for write installed\n");return 0;fail:printk(KERN_INFO "Breakpoint registration failed\n");return ret;

 看网上说ARM Architecture Reference Manual Supplement ARMv8.1, for ARMv8-A architecture profile这些文档的degug register里面有。文档可以在这个网站下载

https://developer.arm.com/documentation

Documentation – Arm Developer

 可以看到应该是每个cpu都注册了一个。感觉应该是在一个cpu上注册，所有cpu都会去检查（原理不清楚，上面的文档没有看明白，猜测的）

struct perf_event * __percpu *
register_wide_hw_breakpoint(struct perf_event_attr *attr,perf_overflow_handler_t triggered,void *context)
{struct perf_event * __percpu *cpu_events, *bp;
........................................for_each_online_cpu(cpu) {bp = perf_event_create_kernel_counter(attr, cpu, NULL,triggered, context);
...................................................
}

暂时不写了