platform_system_core/libprocessgroup/processgroup.cpp
Martijn Coenen 5bb91ab553 Don't use mem cgroups for pid accounting.
Commit b82bab66 introduced the use of memory cgroups for keeping track
of forked PIDs; it basically creates a separate memory cgroup for every
process forked from zygote.

Each such memory cgroup which also have its own LRU with (in)active file
and anonymous pages. The current theory is this could potentially introduce
two problems:
1) kswapd runs longer because it has to iterate over the LRUs of all mem
cgroups, instead of over the LRUs of a single root mem cgroup;
2) the way kswapd reclaims things will be different also - I think it will
tend to bias reclaim to smaller mem cgroups, and process private pages
will end up on ZRAM swap much sooner.

Until we figure this out, fall back to the CPU accounting cgroup for keeping
track of forked PIDs. This leaves us with a single root mem cgroup again. We
can also keep userspace lmkd enabled because it only requires the root mem
cgroup.

Bug: 27381069
Change-Id: Ife397a6ac232761f2adfe6f5056582be0d1b4ff1
2016-03-18 15:35:48 +01:00

384 lines
10 KiB
C++

/*
* Copyright 2014 Google, Inc
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
//#define LOG_NDEBUG 0
#define LOG_TAG "libprocessgroup"
#include <assert.h>
#include <dirent.h>
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h>
#include <mutex>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <log/log.h>
#include <private/android_filesystem_config.h>
#include <utils/SystemClock.h>
#include <processgroup/processgroup.h>
// Uncomment line below use memory cgroups for keeping track of (forked) PIDs
// #define USE_MEMCG 1
#define MEM_CGROUP_PATH "/dev/memcg/apps"
#define MEM_CGROUP_TASKS "/dev/memcg/apps/tasks"
#define ACCT_CGROUP_PATH "/acct"
#define PROCESSGROUP_UID_PREFIX "uid_"
#define PROCESSGROUP_PID_PREFIX "pid_"
#define PROCESSGROUP_CGROUP_PROCS_FILE "/cgroup.procs"
#define PROCESSGROUP_MAX_UID_LEN 11
#define PROCESSGROUP_MAX_PID_LEN 11
#define PROCESSGROUP_MAX_PATH_LEN \
((sizeof(MEM_CGROUP_PATH) > sizeof(ACCT_CGROUP_PATH) ? \
sizeof(MEM_CGROUP_PATH) : sizeof(ACCT_CGROUP_PATH)) + \
sizeof(PROCESSGROUP_UID_PREFIX) + 1 + \
PROCESSGROUP_MAX_UID_LEN + \
sizeof(PROCESSGROUP_PID_PREFIX) + 1 + \
PROCESSGROUP_MAX_PID_LEN + \
sizeof(PROCESSGROUP_CGROUP_PROCS_FILE) + \
1)
std::once_flag init_path_flag;
struct ctx {
bool initialized;
int fd;
char buf[128];
char *buf_ptr;
size_t buf_len;
};
static const char* getCgroupRootPath() {
#ifdef USE_MEMCG
static const char* cgroup_root_path = NULL;
std::call_once(init_path_flag, [&]() {
// Check if mem cgroup is mounted, only then check for write-access to avoid
// SELinux denials
cgroup_root_path = access(MEM_CGROUP_TASKS, F_OK) || access(MEM_CGROUP_PATH, W_OK) ?
ACCT_CGROUP_PATH : MEM_CGROUP_PATH;
});
return cgroup_root_path;
#else
return ACCT_CGROUP_PATH;
#endif
}
static int convertUidToPath(char *path, size_t size, uid_t uid)
{
return snprintf(path, size, "%s/%s%d",
getCgroupRootPath(),
PROCESSGROUP_UID_PREFIX,
uid);
}
static int convertUidPidToPath(char *path, size_t size, uid_t uid, int pid)
{
return snprintf(path, size, "%s/%s%d/%s%d",
getCgroupRootPath(),
PROCESSGROUP_UID_PREFIX,
uid,
PROCESSGROUP_PID_PREFIX,
pid);
}
static int initCtx(uid_t uid, int pid, struct ctx *ctx)
{
int ret;
char path[PROCESSGROUP_MAX_PATH_LEN] = {0};
convertUidPidToPath(path, sizeof(path), uid, pid);
strlcat(path, PROCESSGROUP_CGROUP_PROCS_FILE, sizeof(path));
int fd = open(path, O_RDONLY);
if (fd < 0) {
ret = -errno;
SLOGW("failed to open %s: %s", path, strerror(errno));
return ret;
}
ctx->fd = fd;
ctx->buf_ptr = ctx->buf;
ctx->buf_len = 0;
ctx->initialized = true;
SLOGV("Initialized context for %s", path);
return 0;
}
static int refillBuffer(struct ctx *ctx)
{
memmove(ctx->buf, ctx->buf_ptr, ctx->buf_len);
ctx->buf_ptr = ctx->buf;
ssize_t ret = read(ctx->fd, ctx->buf_ptr + ctx->buf_len,
sizeof(ctx->buf) - ctx->buf_len - 1);
if (ret < 0) {
return -errno;
} else if (ret == 0) {
return 0;
}
ctx->buf_len += ret;
ctx->buf[ctx->buf_len] = 0;
SLOGV("Read %zd to buffer: %s", ret, ctx->buf);
assert(ctx->buf_len <= sizeof(ctx->buf));
return ret;
}
static pid_t getOneAppProcess(uid_t uid, int appProcessPid, struct ctx *ctx)
{
if (!ctx->initialized) {
int ret = initCtx(uid, appProcessPid, ctx);
if (ret < 0) {
return ret;
}
}
char *eptr;
while ((eptr = (char *)memchr(ctx->buf_ptr, '\n', ctx->buf_len)) == NULL) {
int ret = refillBuffer(ctx);
if (ret == 0) {
return -ERANGE;
}
if (ret < 0) {
return ret;
}
}
*eptr = '\0';
char *pid_eptr = NULL;
errno = 0;
long pid = strtol(ctx->buf_ptr, &pid_eptr, 10);
if (errno != 0) {
return -errno;
}
if (pid_eptr != eptr) {
return -EINVAL;
}
ctx->buf_len -= (eptr - ctx->buf_ptr) + 1;
ctx->buf_ptr = eptr + 1;
return (pid_t)pid;
}
static int removeProcessGroup(uid_t uid, int pid)
{
int ret;
char path[PROCESSGROUP_MAX_PATH_LEN] = {0};
convertUidPidToPath(path, sizeof(path), uid, pid);
ret = rmdir(path);
convertUidToPath(path, sizeof(path), uid);
rmdir(path);
return ret;
}
static void removeUidProcessGroups(const char *uid_path)
{
DIR *uid = opendir(uid_path);
if (uid != NULL) {
struct dirent cur;
struct dirent *dir;
while ((readdir_r(uid, &cur, &dir) == 0) && dir) {
char path[PROCESSGROUP_MAX_PATH_LEN];
if (dir->d_type != DT_DIR) {
continue;
}
if (strncmp(dir->d_name, PROCESSGROUP_PID_PREFIX, strlen(PROCESSGROUP_PID_PREFIX))) {
continue;
}
snprintf(path, sizeof(path), "%s/%s", uid_path, dir->d_name);
SLOGV("removing %s\n", path);
rmdir(path);
}
closedir(uid);
}
}
void removeAllProcessGroups()
{
SLOGV("removeAllProcessGroups()");
const char *cgroup_root_path = getCgroupRootPath();
DIR *root = opendir(cgroup_root_path);
if (root == NULL) {
SLOGE("failed to open %s: %s", cgroup_root_path, strerror(errno));
} else {
struct dirent cur;
struct dirent *dir;
while ((readdir_r(root, &cur, &dir) == 0) && dir) {
char path[PROCESSGROUP_MAX_PATH_LEN];
if (dir->d_type != DT_DIR) {
continue;
}
if (strncmp(dir->d_name, PROCESSGROUP_UID_PREFIX, strlen(PROCESSGROUP_UID_PREFIX))) {
continue;
}
snprintf(path, sizeof(path), "%s/%s", cgroup_root_path, dir->d_name);
removeUidProcessGroups(path);
SLOGV("removing %s\n", path);
rmdir(path);
}
closedir(root);
}
}
static int killProcessGroupOnce(uid_t uid, int initialPid, int signal)
{
int processes = 0;
struct ctx ctx;
pid_t pid;
ctx.initialized = false;
while ((pid = getOneAppProcess(uid, initialPid, &ctx)) >= 0) {
processes++;
if (pid == 0) {
// Should never happen... but if it does, trying to kill this
// will boomerang right back and kill us! Let's not let that happen.
SLOGW("Yikes, we've been told to kill pid 0! How about we don't do that.");
continue;
}
if (pid != initialPid) {
// We want to be noisy about killing processes so we can understand
// what is going on in the log; however, don't be noisy about the base
// process, since that it something we always kill, and we have already
// logged elsewhere about killing it.
SLOGI("Killing pid %d in uid %d as part of process group %d", pid, uid, initialPid);
}
int ret = kill(pid, signal);
if (ret == -1) {
SLOGW("failed to kill pid %d: %s", pid, strerror(errno));
}
}
if (ctx.initialized) {
close(ctx.fd);
}
return processes;
}
int killProcessGroup(uid_t uid, int initialPid, int signal)
{
int processes;
const int sleep_us = 5 * 1000; // 5ms
int64_t startTime = android::uptimeMillis();
int retry = 40;
while ((processes = killProcessGroupOnce(uid, initialPid, signal)) > 0) {
SLOGV("killed %d processes for processgroup %d\n", processes, initialPid);
if (retry > 0) {
usleep(sleep_us);
--retry;
} else {
SLOGE("failed to kill %d processes for processgroup %d\n",
processes, initialPid);
break;
}
}
SLOGV("Killed process group uid %d pid %d in %" PRId64 "ms, %d procs remain", uid, initialPid,
android::uptimeMillis()-startTime, processes);
if (processes == 0) {
return removeProcessGroup(uid, initialPid);
} else {
return -1;
}
}
static int mkdirAndChown(const char *path, mode_t mode, uid_t uid, gid_t gid)
{
int ret;
ret = mkdir(path, mode);
if (ret < 0 && errno != EEXIST) {
return -errno;
}
ret = chown(path, uid, gid);
if (ret < 0) {
ret = -errno;
rmdir(path);
return ret;
}
return 0;
}
int createProcessGroup(uid_t uid, int initialPid)
{
char path[PROCESSGROUP_MAX_PATH_LEN] = {0};
int ret;
convertUidToPath(path, sizeof(path), uid);
ret = mkdirAndChown(path, 0750, AID_SYSTEM, AID_SYSTEM);
if (ret < 0) {
SLOGE("failed to make and chown %s: %s", path, strerror(-ret));
return ret;
}
convertUidPidToPath(path, sizeof(path), uid, initialPid);
ret = mkdirAndChown(path, 0750, AID_SYSTEM, AID_SYSTEM);
if (ret < 0) {
SLOGE("failed to make and chown %s: %s", path, strerror(-ret));
return ret;
}
strlcat(path, PROCESSGROUP_CGROUP_PROCS_FILE, sizeof(path));
int fd = open(path, O_WRONLY);
if (fd < 0) {
ret = -errno;
SLOGE("failed to open %s: %s", path, strerror(errno));
return ret;
}
char pid[PROCESSGROUP_MAX_PID_LEN + 1] = {0};
int len = snprintf(pid, sizeof(pid), "%d", initialPid);
ret = write(fd, pid, len);
if (ret < 0) {
ret = -errno;
SLOGE("failed to write '%s' to %s: %s", pid, path, strerror(errno));
} else {
ret = 0;
}
close(fd);
return ret;
}