Skip to content

Commit

Permalink
mem_watcher:添加对slab分配速率和总字节数的监控
Browse files Browse the repository at this point in the history
  • Loading branch information
@syxl-time
syxl-time committed Jan 10, 2025
1 parent d85b2b7 commit d0fac07
Show file tree
Hide file tree
Showing 4 changed files with 231 additions and 8 deletions.
2 changes: 1 addition & 1 deletion eBPF_Supermarket/Memory_Subsystem/mem_watcher/Makefile
View file
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ INCLUDES := -I$(OUTPUT) -I../../libbpf/include/uapi -I$(dir $(VMLINUX)) -I$(LIBB
CFLAGS := -g -Wall
ALL_LDFLAGS := $(LDFLAGS) $(EXTRA_LDFLAGS)

APPS = paf pr procstat sysstat memleak fraginfo vmasnap drsnoop oomkiller numafraginfo
APPS = paf pr procstat sysstat memleak fraginfo vmasnap drsnoop oomkiller numafraginfo slabrate

TARGETS= mem_watcher
CARGO ?= $(shell which cargo)
Expand Down
51 changes: 51 additions & 0 deletions eBPF_Supermarket/Memory_Subsystem/mem_watcher/bpf/slabrate.bpf.c
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,51 @@
#include <vmlinux.h>
#include <bpf/bpf_helpers.h>
#include <bpf/bpf_core_read.h>
#include <bpf/bpf_tracing.h>
#include "mem_watcher.h"

#define MAX_ENTRIES 10240

const volatile pid_t target_pid = 0;

static struct slabrate_info slab_zero_value = {};

struct {
__uint(type, BPF_MAP_TYPE_HASH);
__uint(max_entries, MAX_ENTRIES);
__type(key, char *);
__type(value, struct slabrate_info);
} slab_entries SEC(".maps");

static int probe_entry(struct kmem_cache *cachep)
{
__u64 pid_tgid = bpf_get_current_pid_tgid();
__u32 pid = pid_tgid >> 32;
struct slabrate_info *valuep;
const char *name = BPF_CORE_READ(cachep, name);

if (target_pid && target_pid != pid)
return 0;

valuep = bpf_map_lookup_elem(&slab_entries, &name);
if (!valuep) {
bpf_map_update_elem(&slab_entries, &name, &slab_zero_value, BPF_ANY);
valuep = bpf_map_lookup_elem(&slab_entries, &name);
if (!valuep)
return 0;
bpf_probe_read_kernel(&valuep->name, sizeof(valuep->name), name);
}

valuep->count++;
valuep->size += BPF_CORE_READ(cachep, size);

return 0;
}

SEC("kprobe/kmem_cache_alloc")
int BPF_KPROBE(kmem_cache_alloc, struct kmem_cache *cachep)
{
return probe_entry(cachep);
}

char LICENSE[] SEC("license") = "Dual BSD/GPL";
9 changes: 9 additions & 0 deletions eBPF_Supermarket/Memory_Subsystem/mem_watcher/include/mem_watcher.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -200,4 +200,13 @@ struct event {
char comm[TASK_COMM_LEN]; // 被杀死进程的命令名
};

/* slabrate.h */
#define CACHE_NAME_SIZE 32

struct slabrate_info {
char name[CACHE_NAME_SIZE];
__u64 count;
__u64 size;
};

#endif /* __MEM_WATCHER_H */
177 changes: 170 additions & 7 deletions eBPF_Supermarket/Memory_Subsystem/mem_watcher/mem_watcher.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@
#include "memleak.skel.h"
#include "vmasnap.skel.h"
#include "drsnoop.skel.h"
#include "slabrate.skel.h"

#include "mem_watcher.h"
#include "fraginfo.h"
Expand Down Expand Up @@ -122,6 +123,13 @@ struct order_entry
struct ctg_info oinfo;
};

#define OUTPUT_ROWS_LIMIT 10240

static pid_t target_pid = 0;
static bool clear_screen = true;
static int output_rows = 20;
static int count = 99999999;

int compare_entries(const void *a, const void *b)
{
struct order_entry *entryA = (struct order_entry *)a;
Expand Down Expand Up @@ -204,7 +212,6 @@ static volatile bool exiting = false;
} \
}

// 为 oomkiller 使用的宏，指定 map_name
#define LOAD_AND_ATTACH_SKELETON_WITH_MAP(skel, event, map_name) \
do \
{ \
Expand All @@ -230,7 +237,6 @@ static volatile bool exiting = false;
} \
} while (0)

// 保留原有逻辑的宏
#define LOAD_AND_ATTACH_SKELETON(skel, event) \
do \
{ \
Expand Down Expand Up @@ -273,7 +279,8 @@ static struct env
int interval; // 打印间隔，单位为秒
int duration; // 运行时长，单位为秒
bool part2; // 是否启用系统内存状态报告的扩展部分
bool oomkiller; // 是否启用oomkiller事件处理
bool oomkiller; // 是否启用oomkiller事件处理
bool slabrate;

long choose_pid; // 选择的进程号
bool rss; // 是否打印进程页面信息
Expand All @@ -292,8 +299,9 @@ static struct env
.print_time = false, // 默认不打印地址申请时间
.rss = false, // 默认不打印进程页面信息
.part2 = false, // 默认关闭系统内存状态报告的扩展部分
.oomkiller = false, // 默认关闭oomkiller事件处理
.choose_pid = 0, // 默认不选择特定进程
.oomkiller = false, // 默认关闭oomkiller事件处理
.slabrate = false,
.choose_pid = 0, // 默认不选择特定进程
.interval = 1, // 默认打印间隔为1秒
.duration = 10, // 默认持续运行10秒
};
Expand Down Expand Up @@ -343,7 +351,9 @@ static const struct argp_option opts[] = {
{"oomkiller", 'o', 0, 0, "print oomkiller (内存不足时被杀死的进程信息)"},
{0, 0, 0, 0, "numafraginfo:", 16},
{"numafraginfo", 'N', 0, 0, "print numafraginfo"},


{0, 0, 0, 0, "slabrate:", 17},
{"slabrate", 'e', 0, 0, "print slabrate"},

{NULL, 'h', NULL, OPTION_HIDDEN, "show the full help"},
{0},
Expand Down Expand Up @@ -409,6 +419,9 @@ static error_t parse_arg(int key, char *arg, struct argp_state *state)
case 'N':
env.numafraginfo = true;
break;
case 'e':
env.slabrate = true;
break;
default:
return ARGP_ERR_UNKNOWN;
}
Expand Down Expand Up @@ -449,6 +462,7 @@ static int process_numafraginfo(struct numafraginfo_bpf *skel_numafraginfo);
static int process_vmasnap(struct vmasnap_bpf *skel_vmasnap);
static int process_drsnoop(struct drsnoop_bpf *skel_drsnoop);
static int process_oomkiller(struct oomkiller_bpf *skel_oomkiller); // 新增的oomkiller处理函数原型
static int process_slabrate(struct slabrate_bpf *skel_slabrate);
static int handle_event_oomkiller(void *ctx, void *data, size_t data_sz); // 新增的oomkiller事件处理函数
static __u64 adjust_time_to_program_start_time(__u64 first_query_time);
static int update_addr_times(struct memleak_bpf *skel_memleak);
Expand All @@ -469,7 +483,8 @@ int main(int argc, char **argv)
struct numafraginfo_bpf *skel_numafraginfo;
struct vmasnap_bpf *skel_vmasnap;
struct oomkiller_bpf *skel_oomkiller;
struct drsnoop_bpf *skel_drsnoop;
struct drsnoop_bpf *skel_drsnoop;
struct slabrate_bpf *skel_slabrate;

err = argp_parse(&argp, argc, argv, 0, NULL, NULL);
if (err)
Expand Down Expand Up @@ -538,6 +553,51 @@ int main(int argc, char **argv)
{
PROCESS_SKEL(skel_drsnoop, drsnoop);
}
else if (env.slabrate)
{
PROCESS_SKEL(skel_slabrate, slabrate);
}
else
{
fprintf(stderr, "No valid option was given\n");
return 1;
}

if (env.paf)
{
err = process_paf(skel_paf);
}
else if (env.pr)
{
err = process_pr(skel_pr);
}
else if (env.procstat)
{
err = process_procstat(skel_procstat);
}
else if (env.fraginfo)
{
err = process_fraginfo(skel_fraginfo);
}
else if(env.numafraginfo){
err = process_numafraginfo(skel_numafraginfo);
}
else if (env.vmasnap)
{
err = process_vmasnap(skel_vmasnap);
}
else if (env.sysstat)
{
err = process_sysstat(skel_sysstat);
}
else if (env.memleak)
{
err = process_memleak(skel_memleak, env);
}
else if (env.oomkiller) // 处理 oomkiller
{
err = process_oomkiller(skel_oomkiller); // 使用处理 oom
}

return 0;
}
Expand Down Expand Up @@ -1706,3 +1766,106 @@ static int process_drsnoop(struct drsnoop_bpf *skel_drsnoop) {

return err < 0 ? -err : 0;
}

static int sort_column(const void *obj1, const void *obj2)
{
struct slabrate_info *s1 = (struct slabrate_info *)obj1;
struct slabrate_info *s2 = (struct slabrate_info *)obj2;

return s2->size - s1->size;
}

static int print_stat(struct slabrate_bpf *obj)
{
char *key, **prev_key = NULL;
static struct slabrate_info values[OUTPUT_ROWS_LIMIT];
int i, err = 0, rows = 0;
int fd = bpf_map__fd(obj->maps.slab_entries);

printf("%-32s %6s %10s\n", "CACHE", "ALLOCS", "BYTES");

while (1) {
err = bpf_map_get_next_key(fd, prev_key, &key);
if (err) {
if (errno == ENOENT) {
err = 0;
break;
}
return err;
}
err = bpf_map_lookup_elem(fd, &key, &values[rows++]);
if (err) {
return err;
}
prev_key = &key;
}

qsort(values, rows, sizeof(struct slabrate_info), sort_column);
rows = rows < output_rows ? rows : output_rows;
for (i = 0; i < rows; i++)
printf("%-32s %6lld %10lld\n",
values[i].name, values[i].count, values[i].size);

printf("\n");
prev_key = NULL;

while (1) {
err = bpf_map_get_next_key(fd, prev_key, &key);
if (err) {
if (errno == ENOENT) {
err = 0;
break;
}
return err;
}
err = bpf_map_delete_elem(fd, &key);
if (err) {
return err;
}
prev_key = &key;
}
return err;
}

static int process_slabrate(struct slabrate_bpf *skel_slabrate) {
int err;

bpf_program__set_autoload(skel_slabrate->progs.kmem_cache_alloc, true);

skel_slabrate->rodata->target_pid = target_pid;

err = slabrate_bpf__load(skel_slabrate);
if (err) {
fprintf(stderr, "Failed to load and verify BPF skeleton\n");
return 1;
}

err = slabrate_bpf__attach(skel_slabrate);
if (err) {
fprintf(stderr, "Failed to attach BPF skeleton\n");
goto slabrate_cleanup;
}

while (1) {
sleep(1);

if (clear_screen) {
err = system("clear");
if (err)
goto slabrate_cleanup;
}

err = print_stat(skel_slabrate);
if (err)
goto slabrate_cleanup;

count--;
if (exiting || !count)
goto slabrate_cleanup;
}

slabrate_cleanup:
slabrate_bpf__destroy(skel_slabrate);

return err != 0;
}

0 comments on commit d0fac07

Please sign in to comment.