Oprofile分析（android oprofile性能分析）
一、内核支持：
make menuconfig
1、评测菜单中启用 Oprofile ，在 .config 文件中设置?CONFIG_PROFILING=y?和?CONFIG_OPROFILE=y
2、Kernel Feature->[]Enable hardware performance counter support for perf events不要勾选
3、在boot options->console=ttyFIQ0 androidboot.console=ttyFIQ0 init=/init profile=1
二、修改源码 将kernel/arch/arm/oprofile中的common.文件修改
/** * @file common.c * * @remark Copyright 2004 Oprofile Authors * @remark Copyright 2010 ARM Ltd. * @remark Read the file COPYING * * @author Zwane Mwaikambo * @author Will Deacon [move to perf] */ #include <linux/cpumask.h> #include <linux/err.h> #include <linux/errno.h> #include <linux/init.h> #include <linux/mutex.h> #include <linux/oprofile.h> #include <linux/perf_event.h> #include <linux/platform_device.h> #include <linux/slab.h> #include <asm/stacktrace.h> #include <linux/uaccess.h> #include <asm/perf_event.h> #include <asm/ptrace.h> #ifdef CONFIG_HW_PERF_EVENTS /* * Per performance monitor configuration as set via oprofilefs. */ struct op_counter_config { unsigned long count; unsigned long enabled; unsigned long event; unsigned long unit_mask; unsigned long kernel; unsigned long user; struct perf_event_attr attr; }; static int op_arm_enabled; static DEFINE_MUTEX(op_arm_mutex); static struct op_counter_config *counter_config; static struct perf_event **perf_events[nr_cpumask_bits]; static int perf_num_counters_bak; /* * Overflow callback for oprofile. */ static void op_overflow_handler(struct perf_event *event, int unused, struct perf_sample_data *data, struct pt_regs *regs) { int id; u32 cpu = smp_processor_id(); for (id = 0; id < perf_num_counters_bak; ++id) if (perf_events[cpu][id] == event) break; if (id != perf_num_counters_bak) oprofile_add_sample(regs, id); else pr_warning("oprofile: ignoring spurious overflow " on cpu %u\n, cpu); } /* * Called by op_arm_setup to create perf attributes to mirror the oprofile * settings in counter_config. Attributes are created as `pinned' events and * so are permanently scheduled on the PMU. */ static void op_perf_setup(void) { int i; u32 size = sizeof(struct perf_event_attr); struct perf_event_attr *attr; for (i = 0; i < perf_num_counters_bak; ++i) { attr = &counter_config[i].attr; memset(attr, 0, size); attr->type = PERF_TYPE_RAW; attr->size = size; attr->config = counter_config[i].event; attr->sample_period = counter_config[i].count; attr->pinned = 1; } } static int op_create_counter(int cpu, int event) { int ret = 0; struct perf_event *pevent; if (!counter_config[event].enabled || (perf_events[cpu][event] != NULL)) return ret; pevent = perf_event_create_kernel_counter(&counter_config[event].attr, cpu, NULL, op_overflow_handler); if (IS_ERR(pevent)) { ret = PTR_ERR(pevent); } else if (pevent->state != PERF_EVENT_STATE_ACTIVE) { pr_warning("oprofile: failed to enable event %d " on CPU %d\n, event, cpu); ret = -EBUSY; } else { perf_events[cpu][event] = pevent; } return ret; } static void op_destroy_counter(int cpu, int event) { struct perf_event *pevent = perf_events[cpu][event]; if (pevent) { perf_event_release_kernel(pevent); perf_events[cpu][event] = NULL; } } /* * Called by op_arm_start to create active perf events based on the * perviously configured attributes. */ static int op_perf_start(void) { int cpu, event, ret = 0; for_each_online_cpu(cpu) { for (event = 0; event < perf_num_counters_bak; ++event) { ret = op_create_counter(cpu, event); if (ret) goto out; } } out: return ret; } /* * Called by op_arm_stop at the end of a profiling run. */ static void op_perf_stop(void) { int cpu, event; for_each_online_cpu(cpu) for (event = 0; event < perf_num_counters_bak; ++event) op_destroy_counter(cpu, event); } char *op_name_from_perf_id(void) { enum arm_perf_pmu_ids id = armpmu_get_pmu_id(); switch (id) { case ARM_PERF_PMU_ID_XSCALE1: return "arm/xscale1"; case ARM_PERF_PMU_ID_XSCALE2: return "arm/xscale2"; case ARM_PERF_PMU_ID_V6: return "arm/armv6"; case ARM_PERF_PMU_ID_V6MP: return "arm/mpcore"; case ARM_PERF_PMU_ID_CA8: return "arm/armv7"; case ARM_PERF_PMU_ID_CA9: return "arm/armv7-ca9"; default: return NULL; } } static int op_arm_create_files(struct super_block *sb, struct dentry *root) { unsigned int i; for (i = 0; i < perf_num_counters_bak; i++) { struct dentry *dir; char buf[4]; snprintf(buf, sizeof buf, "%d", i); dir = oprofilefs_mkdir(sb, root, buf); oprofilefs_create_ulong(sb, dir, "enabled", &counter_config[i].enabled); oprofilefs_create_ulong(sb, dir, "event", &counter_config[i].event); oprofilefs_create_ulong(sb, dir, "count", &counter_config[i].count); oprofilefs_create_ulong(sb, dir, "unit_mask", &counter_config[i].unit_mask); oprofilefs_create_ulong(sb, dir, "kernel", &counter_config[i].kernel); oprofilefs_create_ulong(sb, dir, "user", &counter_config[i].user); } return 0; } static int op_arm_setup(void) { spin_lock(&oprofilefs_lock); op_perf_setup(); spin_unlock(&oprofilefs_lock); return 0; } static int op_arm_start(void) { int ret = -EBUSY; mutex_lock(&op_arm_mutex); if (!op_arm_enabled) { ret = 0; op_perf_start(); op_arm_enabled = 1; } mutex_unlock(&op_arm_mutex); return ret; } static void op_arm_stop(void) { mutex_lock(&op_arm_mutex); if (op_arm_enabled) op_perf_stop(); op_arm_enabled = 0; mutex_unlock(&op_arm_mutex); } #ifdef CONFIG_PM static int op_arm_suspend(struct platform_device *dev, pm_message_t state) { mutex_lock(&op_arm_mutex); if (op_arm_enabled) op_perf_stop(); mutex_unlock(&op_arm_mutex); return 0; } static int op_arm_resume(struct platform_device *dev) { mutex_lock(&op_arm_mutex); if (op_arm_enabled && op_perf_start()) op_arm_enabled = 0; mutex_unlock(&op_arm_mutex); return 0; } static struct platform_driver oprofile_driver = { .driver = { .name = "arm-oprofile", }, .resume = op_arm_resume, .suspend = op_arm_suspend, }; static struct platform_device *oprofile_pdev; static int __init init_driverfs(void) { int ret; ret = platform_driver_register(&oprofile_driver); if (ret) goto out; oprofile_pdev = platform_device_register_simple( oprofile_driver.driver.name, 0, NULL, 0); if (IS_ERR(oprofile_pdev)) { ret = PTR_ERR(oprofile_pdev); platform_driver_unregister(&oprofile_driver); } out: return ret; } static void exit_driverfs(void) { platform_device_unregister(oprofile_pdev); platform_driver_unregister(&oprofile_driver); } #else static int __init init_driverfs(void) { return 0; } #define exit_driverfs() do { } while (0) #endif /* CONFIG_PM */ static int report_trace(struct stackframe *frame, void *d) { unsigned int *depth = d; if (*depth) { oprofile_add_trace(frame->pc); (*depth)--; } return *depth == 0; } /* * The registers we're interested in are at the end of the variable * length saved register structure. The fp points at the end of this * structure so the address of this struct is: * (struct frame_tail *)(xxx->fp)-1 */ struct frame_tail { struct frame_tail *fp; unsigned long sp; unsigned long lr; } __attribute__((packed)); static struct frame_tail* user_backtrace(struct frame_tail *tail) { struct frame_tail buftail[2]; /* Also check accessibility of one struct frame_tail beyond */ if (!access_ok(VERIFY_READ, tail, sizeof(buftail))) return NULL; if (__copy_from_user_inatomic(buftail, tail, sizeof(buftail))) return NULL; oprofile_add_trace(buftail[0].lr); /* frame pointers should strictly progress back up the stack * (towards higher addresses) */ if (tail + 1 >= buftail[0].fp) return NULL; return buftail[0].fp-1; } static void arm_backtrace(struct pt_regs * const regs, unsigned int depth) { struct frame_tail *tail = ((struct frame_tail *) regs->ARM_fp) - 1; if (!user_mode(regs)) { struct stackframe frame; frame.fp = regs->ARM_fp; frame.sp = regs->ARM_sp; frame.lr = regs->ARM_lr; frame.pc = regs->ARM_pc; walk_stackframe(&frame, report_trace, &depth); return; } while (depth-- && tail && !((unsigned long) tail & 3)) tail = user_backtrace(tail); } int __init oprofile_arch_init(struct oprofile_operations *ops) { int cpu, ret = 0; perf_num_counters_bak = armpmu_get_max_events(); counter_config = kcalloc(perf_num_counters_bak, sizeof(struct op_counter_config), GFP_KERNEL); if (!counter_config) { pr_info("oprofile: failed to allocate %d " counters\n, perf_num_counters_bak); return -ENOMEM; } ret = init_driverfs(); if (ret) { kfree(counter_config); return ret; } for_each_possible_cpu(cpu) { perf_events[cpu] = kcalloc(perf_num_counters_bak, sizeof(struct perf_event *), GFP_KERNEL); if (!perf_events[cpu]) { pr_info("oprofile: failed to allocate %d perf events " for cpu %d\n, perf_num_counters_bak, cpu); while (--cpu >= 0) kfree(perf_events[cpu]); return -ENOMEM; } } ops->backtrace = arm_backtrace; ops->create_files = op_arm_create_files; ops->setup = op_arm_setup; ops->start = op_arm_start; ops->stop = op_arm_stop; ops->shutdown = op_arm_stop; ops->cpu_type = op_name_from_perf_id(); if (!ops->cpu_type) ret = -ENODEV; else pr_info("oprofile: using %s\n", ops->cpu_type); return ret; } void oprofile_arch_exit(void) { int cpu, id; struct perf_event *event; if (*perf_events) { exit_driverfs(); for_each_possible_cpu(cpu) { for (id = 0; id < perf_num_counters_bak; ++id) { event = perf_events[cpu][id]; if (event != NULL) perf_event_release_kernel(event); } kfree(perf_events[cpu]); } } if (counter_config) kfree(counter_config); } #else int __init oprofile_arch_init(struct oprofile_operations *ops) { pr_info("oprofile: hardware counters not available\n"); return -ENODEV; } void oprofile_arch_exit(void) {} #endif /* CONFIG_HW_PERF_EVENTS */
三、相关
kernel/sched.c--------profile_hit(SCHED_PROFILING, __builtin_return_address(0));
下面是对profile的解释：Linux -- profile内核版本2.6.18-RC7
profile只是内核的一个调试性能的工具，这个可以通过menuconfig中的Instrumentation Support->profile打开。
1. 如何使用profile：
首先确认内核支持profile，然后在内核启动时加入以下参数：profile=1或者其它参数， 新的内核支持profile=schedule 1
2. 内核启动后会创建/proc/profile文件，这个文件可以通过readprofile读取，
如readprofile -m /proc/kallsyms | sort -nr > ~/cur_profile.log,
或者readprofile -r -m /proc/kallsyms |sort -nr,
或者readprofile -r && sleep 1 && readprofile -m /proc/kallsyms |sort -nr >~/cur_profile.log
3. 读取/proc/profile可获得哪些内容？
根据启动配置profile＝？的不同，获取的内容不同：
如果配置成profile=? 可以获得每个函数执行次数，用来调试函数性能很有用
如果设置成profile=schedule ?可以获得每个函数调用schedule的次数，用来调试schedule很有用
四、总结一下：
opcontrol –init 加载模块， mout /dev/oprofile 创建必需的文件和目录
opcontrol --no-vmlinux 或者 opcontrol --vmlinux=/boot/vmlinux-`uname -r` 决定是否对 kernel 进行 profiling
opcontrol --list-events //命令可以查看此结构中支持的事件
opcontrol --event=CPU_CLK_UNHALTED:5000 --event=DATA_CACHE_MISSES:1000 --event=INSTRUCTION_CACHE_MISSES:1000 --event=MEMORY_REQUESTS:1000
opcontrol --reset 清楚当前会话中的数据
opcontrol --start 开始 profiling
./wls 运行应用程序， oprofile 会对它进行 profiling 通过opcontrol --event=L2_CACHE_MISS:500 --event=L1_DTLB_MISS_AND_L2_DTLB_HIT:500
opcontrol --dump 把收集到的数据写入文件 --event=.........命令来进行事件的设置；
opcontrol --stop 停止 profiling 此命令--event参数必须依次给出，无论有多少个，不可分行，切记
opcotrol -h 关闭守护进程 oprofiled
opcontrol --shutdown 停止 oprofiled
opcontrol --deinit 卸载模块
常用的是 3→7 这几个过程，得到性能数据之后，可以使用 opreport, opstack, opgprof, opannotate几个工具进行分析，我常用的是
opreport, opannotate 进行分析。
五、示例与常见问题解诀
Command format :$:tager board, #:Host
1.Extract opf.tar.gz and busybox.tar.gz ($: tar -zxvf ***.tar.gz)
2.Use adb push opf to /data/opf and busybox to /data/busybox
(#: adb push opf /data/opf and #: adb push busybox /data/busybox)
3.Run $:export PATH=$PATH:/data/opf
$: is command to begin
4.Run $:opcontrol --init
Remark: if target board have oprofile,please delete the files(path:/system/xbin)
Question:
grep: /etc/mtab: No such file or directory
grep: /etc/mtab: No such file or directory
Kernel support not available, missing opcontrol --init as root
Answer:
a.$:mount -o remount rw /
b.$:mount -o rw,remount -t yaffs2 /dev/block/mmcblk0p9 /system
c.$:touch /system/etc/mtab or touch /etc/mtab
d.$:echo nodev /dev/oprofile oprofilefs rw 0 0 > /system/etc/mtab or /etc/mtab
e.make sure mtab the nodev $:cat /system/etc/mtab
5.$:opcontrol --init
Question:
cat: can't open '/dev/oprofile/cpu_type': No such file or directory
Unable to open cpu_type file for reading
Make sure you have done opcontrol --init
cpu_type 'unset' is not valid
you should upgrade oprofile or force the use of timer mode
Answer:
a.$:mkdir /dev/oprofile (if not exist)
b.mount -t oprofilefs nodev /dev/oprofile
6.$:opcontrol --session-dir=/data/first --no-vmlinux --callgraph=5
7.$:opcontrol --start
Question:
oprofile: could not open unit mask description file /home/henry/workspace/opf/pc/build/oprofile/share/oprofile//arm/armv7-ca9/unit_masks
Using default event: CPU_CYCLES:100000:0:1:1
oprofile: could not open unit mask description file /home/henry/workspace/opf/pc/build/oprofile/share/oprofile//arm/armv7-ca9/unit_masks
Answer:
a.To target board running $:mkdir -p /home/henry/workspace/opf/pc/build/oprofile
b.Extract share.tar.gz
c.Use ADB push share (#:adb push share /home/henry/workspace/opf/pc/build/oprofile/share)
8.Run App to profile
9.$:opcontrol --dump
10.$:opcontrol --stop
11.$:opreport --session-dir=/data/first -l /data/test/test
$:opreport --session-dir=/data/first -l >log
$:opreport --session-dir=/data/first -l /data/data/Service/lib/libXXX.so >logXXX
$:opreport --session-dir=/data/first -l /data/data/Service/lib/libBBB.so >logBBB
12.$:opcontrol --reset
13.opcontrol --shutdown
Remark: $:opcontrol --shutdown and $:opcontrol --init only execution once
