Android/redroid-patches
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
a36e885cd4cea8ff139faa58eae073328c5b879b
redroid-patches/android-8.1.0_r81/system/core/0007-add-PSI-support-to-lmkd.patch
Ziyang Zhou a36e885cd4 8.1.0_r81 tweak lmkd
2024-05-27 21:07:03 +08:00

2706 lines
83 KiB
Diff
Raw Blame History

From 00074ee13c6be9a0710fb8f46dd0731bf7444f81 Mon Sep 17 00:00:00 2001
From: Ziyang Zhou <ziyang.zhou@outlook.com>
Date: Sat, 4 Nov 2023 10:24:11 +0800
Subject: [PATCH] add PSI support to lmkd
---
lmkd/Android.bp | 10 +-
lmkd/event.logtags | 38 +
lmkd/include/lmkd.h | 193 ++++
lmkd/libpsi/Android.bp | 22 +
lmkd/libpsi/OWNERS | 1 +
lmkd/libpsi/include/psi/psi.h | 68 ++
lmkd/libpsi/psi.c | 99 ++
lmkd/lmkd.c | 1745 +++++++++++++++++++++++++++------
8 files changed, 1895 insertions(+), 281 deletions(-)
create mode 100644 lmkd/event.logtags
create mode 100644 lmkd/include/lmkd.h
create mode 100644 lmkd/libpsi/Android.bp
create mode 100644 lmkd/libpsi/OWNERS
create mode 100644 lmkd/libpsi/include/psi/psi.h
create mode 100644 lmkd/libpsi/psi.c
diff --git a/lmkd/Android.bp b/lmkd/Android.bp
index 3f8a5035f..0f4789774 100644
--- a/lmkd/Android.bp
+++ b/lmkd/Android.bp
@@ -3,11 +3,15 @@ cc_binary {
srcs: ["lmkd.c"],
shared_libs: [
+ "libcutils",
"liblog",
"libprocessgroup",
- "libcutils",
+ "libpsi",
],
- cflags: ["-Werror"],
-
+ local_include_dirs: ["include"],
+ cflags: ["-Werror", "-DLMKD_TRACE_KILLS"],
init_rc: ["lmkd.rc"],
+ logtags: ["event.logtags"],
}
+
+subdirs = ["*"]
diff --git a/lmkd/event.logtags b/lmkd/event.logtags
new file mode 100644
index 000000000..065c6db6d
--- /dev/null
+++ b/lmkd/event.logtags
@@ -0,0 +1,38 @@
+# The entries in this file map a sparse set of log tag numbers to tag names.
+# This is installed on the device, in /system/etc, and parsed by logcat.
+#
+# Tag numbers are decimal integers, from 0 to 2^31. (Let's leave the
+# negative values alone for now.)
+#
+# Tag names are one or more ASCII letters and numbers or underscores, i.e.
+# "[A-Z][a-z][0-9]_". Do not include spaces or punctuation (the former
+# impacts log readability, the latter makes regex searches more annoying).
+#
+# Tag numbers and names are separated by whitespace. Blank lines and lines
+# starting with '#' are ignored.
+#
+# Optionally, after the tag names can be put a description for the value(s)
+# of the tag. Description are in the format
+# (<name>|data type[|data unit])
+# Multiple values are separated by commas.
+#
+# The data type is a number from the following values:
+# 1: int
+# 2: long
+# 3: string
+# 4: list
+#
+# The data unit is a number taken from the following list:
+# 1: Number of objects
+# 2: Number of bytes
+# 3: Number of milliseconds
+# 4: Number of allocations
+# 5: Id
+# 6: Percent
+# s: Number of seconds (monotonic time)
+# Default value for data of type int/long is 2 (bytes).
+#
+# TODO: generate ".java" and ".h" files with integer constants from this file.
+
+# for meminfo logs
+10195355 meminfo (MemFree|1),(Cached|1),(SwapCached|1),(Buffers|1),(Shmem|1),(Unevictable|1),(SwapTotal|1),(SwapFree|1),(ActiveAnon|1),(InactiveAnon|1),(ActiveFile|1),(InactiveFile|1),(SReclaimable|1),(SUnreclaim|1),(KernelStack|1),(PageTables|1),(ION_heap|1),(ION_heap_pool|1),(CmaFree|1)
diff --git a/lmkd/include/lmkd.h b/lmkd/include/lmkd.h
new file mode 100644
index 000000000..59377dd1f
--- /dev/null
+++ b/lmkd/include/lmkd.h
@@ -0,0 +1,193 @@
+/*
+ * Copyright 2018 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.
+ */
+
+#ifndef _LMKD_H_
+#define _LMKD_H_
+
+#include <arpa/inet.h>
+#include <sys/cdefs.h>
+#include <sys/types.h>
+
+__BEGIN_DECLS
+
+/*
+ * Supported LMKD commands
+ */
+enum lmk_cmd {
+ LMK_TARGET = 0, /* Associate minfree with oom_adj_score */
+ LMK_PROCPRIO, /* Register a process and set its oom_adj_score */
+ LMK_PROCREMOVE, /* Unregister a process */
+ LMK_PROCPURGE, /* Purge all registered processes */
+ LMK_GETKILLCNT, /* Get number of kills */
+};
+
+/*
+ * Max number of targets in LMK_TARGET command.
+ */
+#define MAX_TARGETS 6
+
+/*
+ * Max packet length in bytes.
+ * Longest packet is LMK_TARGET followed by MAX_TARGETS
+ * of minfree and oom_adj_score values
+ */
+#define CTRL_PACKET_MAX_SIZE (sizeof(int) * (MAX_TARGETS * 2 + 1))
+
+/* LMKD packet - first int is lmk_cmd followed by payload */
+typedef int LMKD_CTRL_PACKET[CTRL_PACKET_MAX_SIZE / sizeof(int)];
+
+/* Get LMKD packet command */
+static inline enum lmk_cmd lmkd_pack_get_cmd(LMKD_CTRL_PACKET pack) {
+ return (enum lmk_cmd)ntohl(pack[0]);
+}
+
+/* LMK_TARGET packet payload */
+struct lmk_target {
+ int minfree;
+ int oom_adj_score;
+};
+
+/*
+ * For LMK_TARGET packet get target_idx-th payload.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline void lmkd_pack_get_target(LMKD_CTRL_PACKET packet, int target_idx,
+ struct lmk_target* target) {
+ target->minfree = ntohl(packet[target_idx * 2 + 1]);
+ target->oom_adj_score = ntohl(packet[target_idx * 2 + 2]);
+}
+
+/*
+ * Prepare LMK_TARGET packet and return packet size in bytes.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline size_t lmkd_pack_set_target(LMKD_CTRL_PACKET packet, struct lmk_target* targets,
+ size_t target_cnt) {
+ int idx = 0;
+ packet[idx++] = htonl(LMK_TARGET);
+ while (target_cnt) {
+ packet[idx++] = htonl(targets->minfree);
+ packet[idx++] = htonl(targets->oom_adj_score);
+ targets++;
+ target_cnt--;
+ }
+ return idx * sizeof(int);
+}
+
+/* LMK_PROCPRIO packet payload */
+struct lmk_procprio {
+ pid_t pid;
+ uid_t uid;
+ int oomadj;
+};
+
+/*
+ * For LMK_PROCPRIO packet get its payload.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline void lmkd_pack_get_procprio(LMKD_CTRL_PACKET packet, struct lmk_procprio* params) {
+ params->pid = (pid_t)ntohl(packet[1]);
+ params->uid = (uid_t)ntohl(packet[2]);
+ params->oomadj = ntohl(packet[3]);
+}
+
+/*
+ * Prepare LMK_PROCPRIO packet and return packet size in bytes.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline size_t lmkd_pack_set_procprio(LMKD_CTRL_PACKET packet, struct lmk_procprio* params) {
+ packet[0] = htonl(LMK_PROCPRIO);
+ packet[1] = htonl(params->pid);
+ packet[2] = htonl(params->uid);
+ packet[3] = htonl(params->oomadj);
+ return 4 * sizeof(int);
+}
+
+/* LMK_PROCREMOVE packet payload */
+struct lmk_procremove {
+ pid_t pid;
+};
+
+/*
+ * For LMK_PROCREMOVE packet get its payload.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline void lmkd_pack_get_procremove(LMKD_CTRL_PACKET packet,
+ struct lmk_procremove* params) {
+ params->pid = (pid_t)ntohl(packet[1]);
+}
+
+/*
+ * Prepare LMK_PROCREMOVE packet and return packet size in bytes.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline size_t lmkd_pack_set_procremove(LMKD_CTRL_PACKET packet,
+ struct lmk_procprio* params) {
+ packet[0] = htonl(LMK_PROCREMOVE);
+ packet[1] = htonl(params->pid);
+ return 2 * sizeof(int);
+}
+
+/*
+ * Prepare LMK_PROCPURGE packet and return packet size in bytes.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline size_t lmkd_pack_set_procpurge(LMKD_CTRL_PACKET packet) {
+ packet[0] = htonl(LMK_PROCPURGE);
+ return sizeof(int);
+}
+
+/* LMK_GETKILLCNT packet payload */
+struct lmk_getkillcnt {
+ int min_oomadj;
+ int max_oomadj;
+};
+
+/*
+ * For LMK_GETKILLCNT packet get its payload.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline void lmkd_pack_get_getkillcnt(LMKD_CTRL_PACKET packet,
+ struct lmk_getkillcnt* params) {
+ params->min_oomadj = ntohl(packet[1]);
+ params->max_oomadj = ntohl(packet[2]);
+}
+
+/*
+ * Prepare LMK_GETKILLCNT packet and return packet size in bytes.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline size_t lmkd_pack_set_getkillcnt(LMKD_CTRL_PACKET packet,
+ struct lmk_getkillcnt* params) {
+ packet[0] = htonl(LMK_GETKILLCNT);
+ packet[1] = htonl(params->min_oomadj);
+ packet[2] = htonl(params->max_oomadj);
+ return 3 * sizeof(int);
+}
+
+/*
+ * Prepare LMK_GETKILLCNT reply packet and return packet size in bytes.
+ * Warning: no checks performed, caller should ensure valid parameters.
+ */
+static inline size_t lmkd_pack_set_getkillcnt_repl(LMKD_CTRL_PACKET packet, int kill_cnt) {
+ packet[0] = htonl(LMK_GETKILLCNT);
+ packet[1] = htonl(kill_cnt);
+ return 2 * sizeof(int);
+}
+
+__END_DECLS
+
+#endif /* _LMKD_H_ */
diff --git a/lmkd/libpsi/Android.bp b/lmkd/libpsi/Android.bp
new file mode 100644
index 000000000..8a970942e
--- /dev/null
+++ b/lmkd/libpsi/Android.bp
@@ -0,0 +1,22 @@
+cc_library_headers {
+ name: "libpsi_headers",
+ export_include_dirs: ["include"],
+}
+
+cc_library {
+ name: "libpsi",
+ srcs: ["psi.c"],
+ shared_libs: [
+ "liblog"
+ ],
+ header_libs: [
+ "libpsi_headers",
+ ],
+ export_header_lib_headers: [
+ "libpsi_headers",
+ ],
+ cflags: [
+ "-Wall",
+ "-Werror",
+ ],
+}
diff --git a/lmkd/libpsi/OWNERS b/lmkd/libpsi/OWNERS
new file mode 100644
index 000000000..b15bb4870
--- /dev/null
+++ b/lmkd/libpsi/OWNERS
@@ -0,0 +1 @@
+surenb@google.com
diff --git a/lmkd/libpsi/include/psi/psi.h b/lmkd/libpsi/include/psi/psi.h
new file mode 100644
index 000000000..cd49e8b60
--- /dev/null
+++ b/lmkd/libpsi/include/psi/psi.h
@@ -0,0 +1,68 @@
+/*
+ * Copyright (C) 2018 The Android Open Source Project
+ *
+ * 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.
+ */
+
+#ifndef __ANDROID_PSI_H__
+#define __ANDROID_PSI_H__
+
+#include <sys/cdefs.h>
+#include <sys/types.h>
+
+__BEGIN_DECLS
+
+enum psi_stall_type {
+ PSI_SOME,
+ PSI_FULL,
+ PSI_TYPE_COUNT
+};
+
+/*
+ * Initializes psi monitor.
+ * stall_type, threshold_us and window_us are monitor parameters
+ * When successful, the function returns file descriptor that can
+ * be used with poll/epoll syscalls to wait for EPOLLPRI events.
+ * When unsuccessful, the function returns -1 and errno is set
+ * appropriately.
+ */
+int init_psi_monitor(enum psi_stall_type stall_type,
+ int threshold_us, int window_us);
+
+/*
+ * Registers psi monitor file descriptor fd on the epoll instance
+ * referred to by the file descriptor epollfd.
+ * data parameter will be associated with event's epoll_data.ptr
+ * member.
+ */
+int register_psi_monitor(int epollfd, int fd, void* data);
+
+/*
+ * Unregisters psi monitor file descriptor fd from the epoll instance
+ * referred to by the file descriptor epollfd.
+ */
+int unregister_psi_monitor(int epollfd, int fd);
+
+/*
+ * Destroys psi monitor.
+ * fd is the file descriptor returned by psi monitor initialization
+ * routine.
+ * Note that if user process exits without calling this routine
+ * kernel will destroy the monitor as its lifetime is linked to
+ * the file descriptor.
+ */
+void destroy_psi_monitor(int fd);
+
+__END_DECLS
+
+#endif // __ANDROID_PSI_H__
diff --git a/lmkd/libpsi/psi.c b/lmkd/libpsi/psi.c
new file mode 100644
index 000000000..f4d5d1878
--- /dev/null
+++ b/lmkd/libpsi/psi.c
@@ -0,0 +1,99 @@
+/*
+ * Copyright (C) 2018 The Android Open Source Project
+ *
+ * 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_TAG "libpsi"
+
+#include <errno.h>
+#include <string.h>
+#include <sys/epoll.h>
+
+#include <log/log.h>
+#include "psi/psi.h"
+
+#define PSI_MON_FILE_MEMORY "/proc/pressure/memory"
+
+static const char* stall_type_name[] = {
+ "some",
+ "full",
+};
+
+int init_psi_monitor(enum psi_stall_type stall_type,
+ int threshold_us, int window_us) {
+ int fd;
+ int res;
+ char buf[256];
+
+ fd = TEMP_FAILURE_RETRY(open(PSI_MON_FILE_MEMORY, O_WRONLY | O_CLOEXEC));
+ if (fd < 0) {
+ ALOGE("No kernel psi monitor support (errno=%d)", errno);
+ return -1;
+ }
+
+ switch (stall_type) {
+ case (PSI_SOME):
+ case (PSI_FULL):
+ res = snprintf(buf, sizeof(buf), "%s %d %d",
+ stall_type_name[stall_type], threshold_us, window_us);
+ break;
+ default:
+ ALOGE("Invalid psi stall type: %d", stall_type);
+ errno = EINVAL;
+ goto err;
+ }
+
+ if (res >= (ssize_t)sizeof(buf)) {
+ ALOGE("%s line overflow for psi stall type '%s'",
+ PSI_MON_FILE_MEMORY, stall_type_name[stall_type]);
+ errno = EINVAL;
+ goto err;
+ }
+
+ res = TEMP_FAILURE_RETRY(write(fd, buf, strlen(buf) + 1));
+ if (res < 0) {
+ ALOGE("%s write failed for psi stall type '%s'; errno=%d",
+ PSI_MON_FILE_MEMORY, stall_type_name[stall_type], errno);
+ goto err;
+ }
+
+ return fd;
+
+err:
+ close(fd);
+ return -1;
+}
+
+int register_psi_monitor(int epollfd, int fd, void* data) {
+ int res;
+ struct epoll_event epev;
+
+ epev.events = EPOLLPRI;
+ epev.data.ptr = data;
+ res = epoll_ctl(epollfd, EPOLL_CTL_ADD, fd, &epev);
+ if (res < 0) {
+ ALOGE("epoll_ctl for psi monitor failed; errno=%d", errno);
+ }
+ return res;
+}
+
+int unregister_psi_monitor(int epollfd, int fd) {
+ return epoll_ctl(epollfd, EPOLL_CTL_DEL, fd, NULL);
+}
+
+void destroy_psi_monitor(int fd) {
+ if (fd >= 0) {
+ close(fd);
+ }
+}
diff --git a/lmkd/lmkd.c b/lmkd/lmkd.c
index 5cfa2c887..3a42784bc 100644
--- a/lmkd/lmkd.c
+++ b/lmkd/lmkd.c
@@ -16,26 +16,59 @@
#define LOG_TAG "lowmemorykiller"
-#include <arpa/inet.h>
+#include <dirent.h>
#include <errno.h>
#include <inttypes.h>
+#include <pwd.h>
#include <sched.h>
#include <signal.h>
+#include <stdbool.h>
#include <stdlib.h>
#include <string.h>
#include <sys/cdefs.h>
#include <sys/epoll.h>
#include <sys/eventfd.h>
#include <sys/mman.h>
+#include <sys/resource.h>
#include <sys/socket.h>
+#include <sys/sysinfo.h>
+#include <sys/time.h>
#include <sys/types.h>
#include <time.h>
#include <unistd.h>
#include <cutils/properties.h>
+#include <cutils/sched_policy.h>
#include <cutils/sockets.h>
+#include <lmkd.h>
#include <log/log.h>
-#include <processgroup/processgroup.h>
+#include <log/log_event_list.h>
+#include <log/log_time.h>
+#include <psi/psi.h>
+#include <system/thread_defs.h>
+
+#ifdef LMKD_LOG_STATS
+#include "statslog.h"
+#endif
+
+/*
+ * Define LMKD_TRACE_KILLS to record lmkd kills in kernel traces
+ * to profile and correlate with OOM kills
+ */
+#ifdef LMKD_TRACE_KILLS
+
+#define ATRACE_TAG ATRACE_TAG_ALWAYS
+#include <cutils/trace.h>
+
+#define TRACE_KILL_START(pid) ATRACE_INT(__FUNCTION__, pid);
+#define TRACE_KILL_END() ATRACE_INT(__FUNCTION__, 0);
+
+#else /* LMKD_TRACE_KILLS */
+
+#define TRACE_KILL_START(pid) ((void)(pid))
+#define TRACE_KILL_END() ((void)0)
+
+#endif /* LMKD_TRACE_KILLS */
#ifndef __unused
#define __unused __attribute__((__unused__))
@@ -44,54 +77,120 @@
#define MEMCG_SYSFS_PATH "/dev/memcg/"
#define MEMCG_MEMORY_USAGE "/dev/memcg/memory.usage_in_bytes"
#define MEMCG_MEMORYSW_USAGE "/dev/memcg/memory.memsw.usage_in_bytes"
-#define MEMPRESSURE_WATCH_MEDIUM_LEVEL "medium"
-#define MEMPRESSURE_WATCH_CRITICAL_LEVEL "critical"
#define ZONEINFO_PATH "/proc/zoneinfo"
+#define MEMINFO_PATH "/proc/meminfo"
#define LINE_MAX 128
+/* Android Logger event logtags (see event.logtags) */
+#define MEMINFO_LOG_TAG 10195355
+
+/* gid containing AID_SYSTEM required */
#define INKERNEL_MINFREE_PATH "/sys/module/lowmemorykiller/parameters/minfree"
#define INKERNEL_ADJ_PATH "/sys/module/lowmemorykiller/parameters/adj"
#define ARRAY_SIZE(x) (sizeof(x) / sizeof(*(x)))
#define EIGHT_MEGA (1 << 23)
-enum lmk_cmd {
- LMK_TARGET,
- LMK_PROCPRIO,
- LMK_PROCREMOVE,
-};
+#define TARGET_UPDATE_MIN_INTERVAL_MS 1000
+
+#define NS_PER_MS (NS_PER_SEC / MS_PER_SEC)
+#define US_PER_MS (US_PER_SEC / MS_PER_SEC)
+
+/* Defined as ProcessList.SYSTEM_ADJ in ProcessList.java */
+#define SYSTEM_ADJ (-900)
+
+#define STRINGIFY(x) STRINGIFY_INTERNAL(x)
+#define STRINGIFY_INTERNAL(x) #x
-#define MAX_TARGETS 6
/*
- * longest is LMK_TARGET followed by MAX_TARGETS each minfree and minkillprio
- * values
+ * PSI monitor tracking window size.
+ * PSI monitor generates events at most once per window,
+ * therefore we poll memory state for the duration of
+ * PSI_WINDOW_SIZE_MS after the event happens.
*/
-#define CTRL_PACKET_MAX (sizeof(int) * (MAX_TARGETS * 2 + 1))
+#define PSI_WINDOW_SIZE_MS 1000
+/* Polling period after initial PSI signal */
+#define PSI_POLL_PERIOD_MS 10
+/* Poll for the duration of one window after initial PSI signal */
+#define PSI_POLL_COUNT (PSI_WINDOW_SIZE_MS / PSI_POLL_PERIOD_MS)
+
+#define min(a, b) (((a) < (b)) ? (a) : (b))
+
+#define FAIL_REPORT_RLIMIT_MS 1000
/* default to old in-kernel interface if no memory pressure events */
-static int use_inkernel_interface = 1;
+static bool use_inkernel_interface = true;
static bool has_inkernel_module;
-/* memory pressure level medium event */
-static int mpevfd[2];
-#define CRITICAL_INDEX 1
-#define MEDIUM_INDEX 0
+/* memory pressure levels */
+enum vmpressure_level {
+ VMPRESS_LEVEL_LOW = 0,
+ VMPRESS_LEVEL_MEDIUM,
+ VMPRESS_LEVEL_CRITICAL,
+ VMPRESS_LEVEL_COUNT
+};
+
+static const char *level_name[] = {
+ "low",
+ "medium",
+ "critical"
+};
+
+struct {
+ int64_t min_nr_free_pages; /* recorded but not used yet */
+ int64_t max_nr_free_pages;
+} low_pressure_mem = { -1, -1 };
-static int medium_oomadj;
-static int critical_oomadj;
+struct psi_threshold {
+ enum psi_stall_type stall_type;
+ int threshold_ms;
+};
+
+static int level_oomadj[VMPRESS_LEVEL_COUNT];
+static int mpevfd[VMPRESS_LEVEL_COUNT] = { -1, -1, -1 };
static bool debug_process_killing;
static bool enable_pressure_upgrade;
static int64_t upgrade_pressure;
static int64_t downgrade_pressure;
-static bool is_go_device;
+static bool low_ram_device;
+static bool kill_heaviest_task;
+static unsigned long kill_timeout_ms;
+static bool use_minfree_levels;
+static bool per_app_memcg;
+static int swap_free_low_percentage;
+static bool use_psi_monitors = false;
+static struct psi_threshold psi_thresholds[VMPRESS_LEVEL_COUNT] = {
+ { PSI_SOME, 70 }, /* 70ms out of 1sec for partial stall */
+ { PSI_SOME, 100 }, /* 100ms out of 1sec for partial stall */
+ { PSI_FULL, 70 }, /* 70ms out of 1sec for complete stall */
+};
+
+static android_log_context ctx;
+
+/* data required to handle events */
+struct event_handler_info {
+ int data;
+ void (*handler)(int data, uint32_t events);
+};
+
+/* data required to handle socket events */
+struct sock_event_handler_info {
+ int sock;
+ struct event_handler_info handler_info;
+};
+
+/* max supported number of data connections */
+#define MAX_DATA_CONN 2
-/* control socket listen and data */
-static int ctrl_lfd;
-static int ctrl_dfd = -1;
-static int ctrl_dfd_reopened; /* did we reopen ctrl conn on this loop? */
+/* socket event handler data */
+static struct sock_event_handler_info ctrl_sock;
+static struct sock_event_handler_info data_sock[MAX_DATA_CONN];
-/* 2 memory pressure levels, 1 ctrl listen socket, 1 ctrl data socket */
-#define MAX_EPOLL_EVENTS 4
+/* vmpressure event handler data */
+static struct event_handler_info vmpressure_hinfo[VMPRESS_LEVEL_COUNT];
+
+/* 3 memory pressure levels, 1 ctrl listen socket, 2 ctrl data socket */
+#define MAX_EPOLL_EVENTS (1 + MAX_DATA_CONN + VMPRESS_LEVEL_COUNT)
static int epollfd;
static int maxevents;
@@ -103,11 +202,117 @@ static int lowmem_adj[MAX_TARGETS];
static int lowmem_minfree[MAX_TARGETS];
static int lowmem_targets_size;
-struct sysmeminfo {
- int nr_free_pages;
- int nr_file_pages;
- int nr_shmem;
- int totalreserve_pages;
+/* Fields to parse in /proc/zoneinfo */
+enum zoneinfo_field {
+ ZI_NR_FREE_PAGES = 0,
+ ZI_NR_FILE_PAGES,
+ ZI_NR_SHMEM,
+ ZI_NR_UNEVICTABLE,
+ ZI_WORKINGSET_REFAULT,
+ ZI_HIGH,
+ ZI_FIELD_COUNT
+};
+
+static const char* const zoneinfo_field_names[ZI_FIELD_COUNT] = {
+ "nr_free_pages",
+ "nr_file_pages",
+ "nr_shmem",
+ "nr_unevictable",
+ "workingset_refault",
+ "high",
+};
+
+union zoneinfo {
+ struct {
+ int64_t nr_free_pages;
+ int64_t nr_file_pages;
+ int64_t nr_shmem;
+ int64_t nr_unevictable;
+ int64_t workingset_refault;
+ int64_t high;
+ /* fields below are calculated rather than read from the file */
+ int64_t totalreserve_pages;
+ } field;
+ int64_t arr[ZI_FIELD_COUNT];
+};
+
+/* Fields to parse in /proc/meminfo */
+enum meminfo_field {
+ MI_NR_FREE_PAGES = 0,
+ MI_CACHED,
+ MI_SWAP_CACHED,
+ MI_BUFFERS,
+ MI_SHMEM,
+ MI_UNEVICTABLE,
+ MI_TOTAL_SWAP,
+ MI_FREE_SWAP,
+ MI_ACTIVE_ANON,
+ MI_INACTIVE_ANON,
+ MI_ACTIVE_FILE,
+ MI_INACTIVE_FILE,
+ MI_SRECLAIMABLE,
+ MI_SUNRECLAIM,
+ MI_KERNEL_STACK,
+ MI_PAGE_TABLES,
+ MI_ION_HELP,
+ MI_ION_HELP_POOL,
+ MI_CMA_FREE,
+ MI_FIELD_COUNT
+};
+
+static const char* const meminfo_field_names[MI_FIELD_COUNT] = {
+ "MemFree:",
+ "Cached:",
+ "SwapCached:",
+ "Buffers:",
+ "Shmem:",
+ "Unevictable:",
+ "SwapTotal:",
+ "SwapFree:",
+ "Active(anon):",
+ "Inactive(anon):",
+ "Active(file):",
+ "Inactive(file):",
+ "SReclaimable:",
+ "SUnreclaim:",
+ "KernelStack:",
+ "PageTables:",
+ "ION_heap:",
+ "ION_heap_pool:",
+ "CmaFree:",
+};
+
+union meminfo {
+ struct {
+ int64_t nr_free_pages;
+ int64_t cached;
+ int64_t swap_cached;
+ int64_t buffers;
+ int64_t shmem;
+ int64_t unevictable;
+ int64_t total_swap;
+ int64_t free_swap;
+ int64_t active_anon;
+ int64_t inactive_anon;
+ int64_t active_file;
+ int64_t inactive_file;
+ int64_t sreclaimable;
+ int64_t sunreclaimable;
+ int64_t kernel_stack;
+ int64_t page_tables;
+ int64_t ion_heap;
+ int64_t ion_heap_pool;
+ int64_t cma_free;
+ /* fields below are calculated rather than read from the file */
+ int64_t nr_file_pages;
+ } field;
+ int64_t arr[MI_FIELD_COUNT];
+};
+
+enum field_match_result {
+ NO_MATCH,
+ PARSE_FAIL,
+ PARSE_SUCCESS
};
struct adjslot_list {
@@ -123,22 +328,76 @@ struct proc {
struct proc *pidhash_next;
};
+struct reread_data {
+ const char* const filename;
+ int fd;
+};
+
+#ifdef LMKD_LOG_STATS
+static bool enable_stats_log;
+static android_log_context log_ctx;
+#endif
+
#define PIDHASH_SZ 1024
static struct proc *pidhash[PIDHASH_SZ];
#define pid_hashfn(x) ((((x) >> 8) ^ (x)) & (PIDHASH_SZ - 1))
#define ADJTOSLOT(adj) ((adj) + -OOM_SCORE_ADJ_MIN)
-static struct adjslot_list procadjslot_list[ADJTOSLOT(OOM_SCORE_ADJ_MAX) + 1];
+#define ADJTOSLOT_COUNT (ADJTOSLOT(OOM_SCORE_ADJ_MAX) + 1)
+static struct adjslot_list procadjslot_list[ADJTOSLOT_COUNT];
+
+#define MAX_DISTINCT_OOM_ADJ 32
+#define KILLCNT_INVALID_IDX 0xFF
+/*
+ * Because killcnt array is sparse a two-level indirection is used
+ * to keep the size small. killcnt_idx stores index of the element in
+ * killcnt array. Index KILLCNT_INVALID_IDX indicates an unused slot.
+ */
+static uint8_t killcnt_idx[ADJTOSLOT_COUNT];
+static uint16_t killcnt[MAX_DISTINCT_OOM_ADJ];
+static int killcnt_free_idx = 0;
+static uint32_t killcnt_total = 0;
/* PAGE_SIZE / 1024 */
static long page_k;
+static bool parse_int64(const char* str, int64_t* ret) {
+ char* endptr;
+ long long val = strtoll(str, &endptr, 10);
+ if (str == endptr || val > INT64_MAX) {
+ return false;
+ }
+ *ret = (int64_t)val;
+ return true;
+}
+
+static enum field_match_result match_field(const char* cp, const char* ap,
+ const char* const field_names[],
+ int field_count, int64_t* field,
+ int *field_idx) {
+ int64_t val;
+ int i;
+
+ for (i = 0; i < field_count; i++) {
+ if (!strcmp(cp, field_names[i])) {
+ *field_idx = i;
+ return parse_int64(ap, field) ? PARSE_SUCCESS : PARSE_FAIL;
+ }
+ }
+ return NO_MATCH;
+}
+
+/*
+ * Read file content from the beginning up to max_len bytes or EOF
+ * whichever happens first.
+ */
static ssize_t read_all(int fd, char *buf, size_t max_len)
{
ssize_t ret = 0;
+ off_t offset = 0;
while (max_len > 0) {
- ssize_t r = read(fd, buf, max_len);
+ ssize_t r = TEMP_FAILURE_RETRY(pread(fd, buf, max_len, offset));
if (r == 0) {
break;
}
@@ -147,12 +406,44 @@ static ssize_t read_all(int fd, char *buf, size_t max_len)
}
ret += r;
buf += r;
+ offset += r;
max_len -= r;
}
return ret;
}
+/*
+ * Read a new or already opened file from the beginning.
+ * If the file has not been opened yet data->fd should be set to -1.
+ * To be used with files which are read often and possibly during high
+ * memory pressure to minimize file opening which by itself requires kernel
+ * memory allocation and might result in a stall on memory stressed system.
+ */
+static int reread_file(struct reread_data *data, char *buf, size_t buf_size) {
+ ssize_t size;
+
+ if (data->fd == -1) {
+ data->fd = open(data->filename, O_RDONLY | O_CLOEXEC);
+ if (data->fd == -1) {
+ ALOGE("%s open: %s", data->filename, strerror(errno));
+ return -1;
+ }
+ }
+
+ size = read_all(data->fd, buf, buf_size - 1);
+ if (size < 0) {
+ ALOGE("%s read: %s", data->filename, strerror(errno));
+ close(data->fd);
+ data->fd = -1;
+ return -1;
+ }
+ ALOG_ASSERT((size_t)size < buf_size - 1, "%s too large", data->filename);
+ buf[size] = 0;
+
+ return 0;
+}
+
static struct proc *pid_lookup(int pid) {
struct proc *procp;
@@ -226,77 +517,118 @@ static int pid_remove(int pid) {
return 0;
}
-static void writefilestring(char *path, char *s) {
+/*
+ * Write a string to a file.
+ * Returns false if the file does not exist.
+ */
+static bool writefilestring(const char *path, const char *s,
+ bool err_if_missing) {
int fd = open(path, O_WRONLY | O_CLOEXEC);
- int len = strlen(s);
- int ret;
+ ssize_t len = strlen(s);
+ ssize_t ret;
if (fd < 0) {
- ALOGE("Error opening %s; errno=%d", path, errno);
- return;
+ if (err_if_missing) {
+ ALOGE("Error opening %s; errno=%d", path, errno);
+ }
+ return false;
}
- ret = write(fd, s, len);
+ ret = TEMP_FAILURE_RETRY(write(fd, s, len));
if (ret < 0) {
ALOGE("Error writing %s; errno=%d", path, errno);
} else if (ret < len) {
- ALOGE("Short write on %s; length=%d", path, ret);
+ ALOGE("Short write on %s; length=%zd", path, ret);
}
close(fd);
+ return true;
}
-static void cmd_procprio(int pid, int uid, int oomadj) {
+static inline long get_time_diff_ms(struct timespec *from,
+ struct timespec *to) {
+ return (to->tv_sec - from->tv_sec) * (long)MS_PER_SEC +
+ (to->tv_nsec - from->tv_nsec) / (long)NS_PER_MS;
+}
+
+static void cmd_procprio(LMKD_CTRL_PACKET packet) {
struct proc *procp;
char path[80];
char val[20];
int soft_limit_mult;
+ struct lmk_procprio params;
+ bool is_system_server;
+ struct passwd *pwdrec;
- if (oomadj < OOM_SCORE_ADJ_MIN || oomadj > OOM_SCORE_ADJ_MAX) {
- ALOGE("Invalid PROCPRIO oomadj argument %d", oomadj);
+ lmkd_pack_get_procprio(packet, &params);
+
+ if (params.oomadj < OOM_SCORE_ADJ_MIN ||
+ params.oomadj > OOM_SCORE_ADJ_MAX) {
+ ALOGE("Invalid PROCPRIO oomadj argument %d", params.oomadj);
return;
}
- snprintf(path, sizeof(path), "/proc/%d/oom_score_adj", pid);
- snprintf(val, sizeof(val), "%d", oomadj);
- writefilestring(path, val);
-
- if (use_inkernel_interface)
+ /* gid containing AID_READPROC required */
+ /* CAP_SYS_RESOURCE required */
+ /* CAP_DAC_OVERRIDE required */
+ snprintf(path, sizeof(path), "/proc/%d/oom_score_adj", params.pid);
+ snprintf(val, sizeof(val), "%d", params.oomadj);
+ if (!writefilestring(path, val, false)) {
+ ALOGW("Failed to open %s; errno=%d: process %d might have been killed",
+ path, errno, params.pid);
+ /* If this file does not exist the process is dead. */
return;
+ }
- if (oomadj >= 900) {
- soft_limit_mult = 0;
- } else if (oomadj >= 800) {
- soft_limit_mult = 0;
- } else if (oomadj >= 700) {
- soft_limit_mult = 0;
- } else if (oomadj >= 600) {
- // Launcher should be perceptible, don't kill it.
- oomadj = 200;
- soft_limit_mult = 1;
- } else if (oomadj >= 500) {
- soft_limit_mult = 0;
- } else if (oomadj >= 400) {
- soft_limit_mult = 0;
- } else if (oomadj >= 300) {
- soft_limit_mult = 1;
- } else if (oomadj >= 200) {
- soft_limit_mult = 2;
- } else if (oomadj >= 100) {
- soft_limit_mult = 10;
- } else if (oomadj >= 0) {
- soft_limit_mult = 20;
- } else {
- // Persistent processes will have a large
- // soft limit 512MB.
- soft_limit_mult = 64;
+ if (use_inkernel_interface) {
+ return;
}
- snprintf(path, sizeof(path), "/dev/memcg/apps/uid_%d/pid_%d/memory.soft_limit_in_bytes", uid, pid);
- snprintf(val, sizeof(val), "%d", soft_limit_mult * EIGHT_MEGA);
- writefilestring(path, val);
+ if (per_app_memcg) {
+ if (params.oomadj >= 900) {
+ soft_limit_mult = 0;
+ } else if (params.oomadj >= 800) {
+ soft_limit_mult = 0;
+ } else if (params.oomadj >= 700) {
+ soft_limit_mult = 0;
+ } else if (params.oomadj >= 600) {
+ // Launcher should be perceptible, don't kill it.
+ params.oomadj = 200;
+ soft_limit_mult = 1;
+ } else if (params.oomadj >= 500) {
+ soft_limit_mult = 0;
+ } else if (params.oomadj >= 400) {
+ soft_limit_mult = 0;
+ } else if (params.oomadj >= 300) {
+ soft_limit_mult = 1;
+ } else if (params.oomadj >= 200) {
+ soft_limit_mult = 8;
+ } else if (params.oomadj >= 100) {
+ soft_limit_mult = 10;
+ } else if (params.oomadj >= 0) {
+ soft_limit_mult = 20;
+ } else {
+ // Persistent processes will have a large
+ // soft limit 512MB.
+ soft_limit_mult = 64;
+ }
+
+ snprintf(path, sizeof(path), MEMCG_SYSFS_PATH
+ "apps/uid_%d/pid_%d/memory.soft_limit_in_bytes",
+ params.uid, params.pid);
+ snprintf(val, sizeof(val), "%d", soft_limit_mult * EIGHT_MEGA);
+
+ /*
+ * system_server process has no memcg under /dev/memcg/apps but should be
+ * registered with lmkd. This is the best way so far to identify it.
+ */
+ is_system_server = (params.oomadj == SYSTEM_ADJ &&
+ (pwdrec = getpwnam("system")) != NULL &&
+ params.uid == pwdrec->pw_uid);
+ writefilestring(path, val, !is_system_server);
+ }
- procp = pid_lookup(pid);
+ procp = pid_lookup(params.pid);
if (!procp) {
procp = malloc(sizeof(struct proc));
if (!procp) {
@@ -304,37 +636,167 @@ static void cmd_procprio(int pid, int uid, int oomadj) {
return;
}
- procp->pid = pid;
- procp->uid = uid;
- procp->oomadj = oomadj;
+ procp->pid = params.pid;
+ procp->uid = params.uid;
+ procp->oomadj = params.oomadj;
proc_insert(procp);
} else {
proc_unslot(procp);
- procp->oomadj = oomadj;
+ procp->oomadj = params.oomadj;
proc_slot(procp);
}
}
-static void cmd_procremove(int pid) {
- if (use_inkernel_interface)
+static void cmd_procremove(LMKD_CTRL_PACKET packet) {
+ struct lmk_procremove params;
+
+ if (use_inkernel_interface) {
+ return;
+ }
+
+ lmkd_pack_get_procremove(packet, &params);
+ /*
+ * WARNING: After pid_remove() procp is freed and can't be used!
+ * Therefore placed at the end of the function.
+ */
+ pid_remove(params.pid);
+}
+
+static void cmd_procpurge() {
+ int i;
+ struct proc *procp;
+ struct proc *next;
+
+ if (use_inkernel_interface) {
return;
+ }
- pid_remove(pid);
+ for (i = 0; i <= ADJTOSLOT(OOM_SCORE_ADJ_MAX); i++) {
+ procadjslot_list[i].next = &procadjslot_list[i];
+ procadjslot_list[i].prev = &procadjslot_list[i];
+ }
+
+ for (i = 0; i < PIDHASH_SZ; i++) {
+ procp = pidhash[i];
+ while (procp) {
+ next = procp->pidhash_next;
+ free(procp);
+ procp = next;
+ }
+ }
+ memset(&pidhash[0], 0, sizeof(pidhash));
+}
+
+static void inc_killcnt(int oomadj) {
+ int slot = ADJTOSLOT(oomadj);
+ uint8_t idx = killcnt_idx[slot];
+
+ if (idx == KILLCNT_INVALID_IDX) {
+ /* index is not assigned for this oomadj */
+ if (killcnt_free_idx < MAX_DISTINCT_OOM_ADJ) {
+ killcnt_idx[slot] = killcnt_free_idx;
+ killcnt[killcnt_free_idx] = 1;
+ killcnt_free_idx++;
+ } else {
+ ALOGW("Number of distinct oomadj levels exceeds %d",
+ MAX_DISTINCT_OOM_ADJ);
+ }
+ } else {
+ /*
+ * wraparound is highly unlikely and is detectable using total
+ * counter because it has to be equal to the sum of all counters
+ */
+ killcnt[idx]++;
+ }
+ /* increment total kill counter */
+ killcnt_total++;
+}
+
+static int get_killcnt(int min_oomadj, int max_oomadj) {
+ int slot;
+ int count = 0;
+
+ if (min_oomadj > max_oomadj)
+ return 0;
+
+ /* special case to get total kill count */
+ if (min_oomadj > OOM_SCORE_ADJ_MAX)
+ return killcnt_total;
+
+ while (min_oomadj <= max_oomadj &&
+ (slot = ADJTOSLOT(min_oomadj)) < ADJTOSLOT_COUNT) {
+ uint8_t idx = killcnt_idx[slot];
+ if (idx != KILLCNT_INVALID_IDX) {
+ count += killcnt[idx];
+ }
+ min_oomadj++;
+ }
+
+ return count;
+}
+
+static int cmd_getkillcnt(LMKD_CTRL_PACKET packet) {
+ struct lmk_getkillcnt params;
+
+ if (use_inkernel_interface) {
+ /* kernel driver does not expose this information */
+ return 0;
+ }
+
+ lmkd_pack_get_getkillcnt(packet, &params);
+
+ return get_killcnt(params.min_oomadj, params.max_oomadj);
}
-static void cmd_target(int ntargets, int *params) {
+static void cmd_target(int ntargets, LMKD_CTRL_PACKET packet) {
int i;
+ struct lmk_target target;
+ char minfree_str[PROPERTY_VALUE_MAX];
+ char *pstr = minfree_str;
+ char *pend = minfree_str + sizeof(minfree_str);
+ static struct timespec last_req_tm;
+ struct timespec curr_tm;
+
+ if (ntargets < 1 || ntargets > (int)ARRAY_SIZE(lowmem_adj))
+ return;
- if (ntargets > (int)ARRAY_SIZE(lowmem_adj))
+ /*
+ * Ratelimit minfree updates to once per TARGET_UPDATE_MIN_INTERVAL_MS
+ * to prevent DoS attacks
+ */
+ if (clock_gettime(CLOCK_MONOTONIC_COARSE, &curr_tm) != 0) {
+ ALOGE("Failed to get current time");
return;
+ }
+
+ if (get_time_diff_ms(&last_req_tm, &curr_tm) <
+ TARGET_UPDATE_MIN_INTERVAL_MS) {
+ ALOGE("Ignoring frequent updated to lmkd limits");
+ return;
+ }
+
+ last_req_tm = curr_tm;
for (i = 0; i < ntargets; i++) {
- lowmem_minfree[i] = ntohl(*params++);
- lowmem_adj[i] = ntohl(*params++);
+ lmkd_pack_get_target(packet, i, &target);
+ lowmem_minfree[i] = target.minfree;
+ lowmem_adj[i] = target.oom_adj_score;
+
+ pstr += snprintf(pstr, pend - pstr, "%d:%d,", target.minfree,
+ target.oom_adj_score);
+ if (pstr >= pend) {
+ /* if no more space in the buffer then terminate the loop */
+ pstr = pend;
+ break;
+ }
}
lowmem_targets_size = ntargets;
+ /* Override the last extra comma */
+ pstr[-1] = '\0';
+ property_set("sys.lmk.minfree_levels", minfree_str);
+
if (has_inkernel_module) {
char minfreestr[128];
char killpriostr[128];
@@ -356,22 +818,29 @@ static void cmd_target(int ntargets, int *params) {
strlcat(killpriostr, val, sizeof(killpriostr));
}
- writefilestring(INKERNEL_MINFREE_PATH, minfreestr);
- writefilestring(INKERNEL_ADJ_PATH, killpriostr);
+ writefilestring(INKERNEL_MINFREE_PATH, minfreestr, true);
+ writefilestring(INKERNEL_ADJ_PATH, killpriostr, true);
}
}
-static void ctrl_data_close(void) {
- ALOGI("Closing Activity Manager data connection");
- close(ctrl_dfd);
- ctrl_dfd = -1;
+static void ctrl_data_close(int dsock_idx) {
+ struct epoll_event epev;
+
+ ALOGI("closing lmkd data connection");
+ if (epoll_ctl(epollfd, EPOLL_CTL_DEL, data_sock[dsock_idx].sock, &epev) == -1) {
+ // Log a warning and keep going
+ ALOGW("epoll_ctl for data connection socket failed; errno=%d", errno);
+ }
maxevents--;
+
+ close(data_sock[dsock_idx].sock);
+ data_sock[dsock_idx].sock = -1;
}
-static int ctrl_data_read(char *buf, size_t bufsz) {
+static int ctrl_data_read(int dsock_idx, char *buf, size_t bufsz) {
int ret = 0;
- ret = read(ctrl_dfd, buf, bufsz);
+ ret = TEMP_FAILURE_RETRY(read(data_sock[dsock_idx].sock, buf, bufsz));
if (ret == -1) {
ALOGE("control data socket read failed; errno=%d", errno);
@@ -383,39 +852,72 @@ static int ctrl_data_read(char *buf, size_t bufsz) {
return ret;
}
-static void ctrl_command_handler(void) {
- int ibuf[CTRL_PACKET_MAX / sizeof(int)];
+static int ctrl_data_write(int dsock_idx, char *buf, size_t bufsz) {
+ int ret = 0;
+
+ ret = TEMP_FAILURE_RETRY(write(data_sock[dsock_idx].sock, buf, bufsz));
+
+ if (ret == -1) {
+ ALOGE("control data socket write failed; errno=%d", errno);
+ } else if (ret == 0) {
+ ALOGE("Got EOF on control data socket");
+ ret = -1;
+ }
+
+ return ret;
+}
+
+static void ctrl_command_handler(int dsock_idx) {
+ LMKD_CTRL_PACKET packet;
int len;
- int cmd = -1;
+ enum lmk_cmd cmd;
int nargs;
int targets;
+ int kill_cnt;
- len = ctrl_data_read((char *)ibuf, CTRL_PACKET_MAX);
+ len = ctrl_data_read(dsock_idx, (char *)packet, CTRL_PACKET_MAX_SIZE);
if (len <= 0)
return;
+ if (len < (int)sizeof(int)) {
+ ALOGE("Wrong control socket read length len=%d", len);
+ return;
+ }
+
+ cmd = lmkd_pack_get_cmd(packet);
nargs = len / sizeof(int) - 1;
if (nargs < 0)
goto wronglen;
- cmd = ntohl(ibuf[0]);
-
switch(cmd) {
case LMK_TARGET:
targets = nargs / 2;
if (nargs & 0x1 || targets > (int)ARRAY_SIZE(lowmem_adj))
goto wronglen;
- cmd_target(targets, &ibuf[1]);
+ cmd_target(targets, packet);
break;
case LMK_PROCPRIO:
if (nargs != 3)
goto wronglen;
- cmd_procprio(ntohl(ibuf[1]), ntohl(ibuf[2]), ntohl(ibuf[3]));
+ cmd_procprio(packet);
break;
case LMK_PROCREMOVE:
if (nargs != 1)
goto wronglen;
- cmd_procremove(ntohl(ibuf[1]));
+ cmd_procremove(packet);
+ break;
+ case LMK_PROCPURGE:
+ if (nargs != 0)
+ goto wronglen;
+ cmd_procpurge();
+ break;
+ case LMK_GETKILLCNT:
+ if (nargs != 2)
+ goto wronglen;
+ kill_cnt = cmd_getkillcnt(packet);
+ len = lmkd_pack_set_getkillcnt_repl(packet, kill_cnt);
+ if (ctrl_data_write(dsock_idx, (char *)packet, len) != len)
+ return;
break;
default:
ALOGE("Received unknown command code %d", cmd);
@@ -428,112 +930,290 @@ wronglen:
ALOGE("Wrong control socket read length cmd=%d len=%d", cmd, len);
}
-static void ctrl_data_handler(uint32_t events) {
- if (events & EPOLLHUP) {
- ALOGI("ActivityManager disconnected");
- if (!ctrl_dfd_reopened)
- ctrl_data_close();
- } else if (events & EPOLLIN) {
- ctrl_command_handler();
+static void ctrl_data_handler(int data, uint32_t events) {
+ if (events & EPOLLIN) {
+ ctrl_command_handler(data);
}
}
-static void ctrl_connect_handler(uint32_t events __unused) {
- struct epoll_event epev;
-
- if (ctrl_dfd >= 0) {
- ctrl_data_close();
- ctrl_dfd_reopened = 1;
+static int get_free_dsock() {
+ for (int i = 0; i < MAX_DATA_CONN; i++) {
+ if (data_sock[i].sock < 0) {
+ return i;
+ }
}
+ return -1;
+}
- ctrl_dfd = accept(ctrl_lfd, NULL, NULL);
+static void ctrl_connect_handler(int data __unused, uint32_t events __unused) {
+ struct epoll_event epev;
+ int free_dscock_idx = get_free_dsock();
+
+ if (free_dscock_idx < 0) {
+ /*
+ * Number of data connections exceeded max supported. This should not
+ * happen but if it does we drop all existing connections and accept
+ * the new one. This prevents inactive connections from monopolizing
+ * data socket and if we drop ActivityManager connection it will
+ * immediately reconnect.
+ */
+ for (int i = 0; i < MAX_DATA_CONN; i++) {
+ ctrl_data_close(i);
+ }
+ free_dscock_idx = 0;
+ }
- if (ctrl_dfd < 0) {
+ data_sock[free_dscock_idx].sock = accept(ctrl_sock.sock, NULL, NULL);
+ if (data_sock[free_dscock_idx].sock < 0) {
ALOGE("lmkd control socket accept failed; errno=%d", errno);
return;
}
- ALOGI("ActivityManager connected");
- maxevents++;
+ ALOGI("lmkd data connection established");
+ /* use data to store data connection idx */
+ data_sock[free_dscock_idx].handler_info.data = free_dscock_idx;
+ data_sock[free_dscock_idx].handler_info.handler = ctrl_data_handler;
epev.events = EPOLLIN;
- epev.data.ptr = (void *)ctrl_data_handler;
- if (epoll_ctl(epollfd, EPOLL_CTL_ADD, ctrl_dfd, &epev) == -1) {
+ epev.data.ptr = (void *)&(data_sock[free_dscock_idx].handler_info);
+ if (epoll_ctl(epollfd, EPOLL_CTL_ADD, data_sock[free_dscock_idx].sock, &epev) == -1) {
ALOGE("epoll_ctl for data connection socket failed; errno=%d", errno);
- ctrl_data_close();
+ ctrl_data_close(free_dscock_idx);
+ return;
+ }
+ maxevents++;
+}
+
+#ifdef LMKD_LOG_STATS
+static void memory_stat_parse_line(char* line, struct memory_stat* mem_st) {
+ char key[LINE_MAX + 1];
+ int64_t value;
+
+ sscanf(line, "%" STRINGIFY(LINE_MAX) "s %" SCNd64 "", key, &value);
+
+ if (strcmp(key, "total_") < 0) {
return;
}
+
+ if (!strcmp(key, "total_pgfault"))
+ mem_st->pgfault = value;
+ else if (!strcmp(key, "total_pgmajfault"))
+ mem_st->pgmajfault = value;
+ else if (!strcmp(key, "total_rss"))
+ mem_st->rss_in_bytes = value;
+ else if (!strcmp(key, "total_cache"))
+ mem_st->cache_in_bytes = value;
+ else if (!strcmp(key, "total_swap"))
+ mem_st->swap_in_bytes = value;
+}
+
+static int memory_stat_from_cgroup(struct memory_stat* mem_st, int pid, uid_t uid) {
+ FILE *fp;
+ char buf[PATH_MAX];
+
+ snprintf(buf, sizeof(buf), MEMCG_PROCESS_MEMORY_STAT_PATH, uid, pid);
+
+ fp = fopen(buf, "r");
+
+ if (fp == NULL) {
+ ALOGE("%s open failed: %s", buf, strerror(errno));
+ return -1;
+ }
+
+ while (fgets(buf, PAGE_SIZE, fp) != NULL) {
+ memory_stat_parse_line(buf, mem_st);
+ }
+ fclose(fp);
+
+ return 0;
+}
+
+static int memory_stat_from_procfs(struct memory_stat* mem_st, int pid) {
+ char path[PATH_MAX];
+ char buffer[PROC_STAT_BUFFER_SIZE];
+ int fd, ret;
+
+ snprintf(path, sizeof(path), PROC_STAT_FILE_PATH, pid);
+ if ((fd = open(path, O_RDONLY | O_CLOEXEC)) < 0) {
+ ALOGE("%s open failed: %s", path, strerror(errno));
+ return -1;
+ }
+
+ ret = read(fd, buffer, sizeof(buffer));
+ if (ret < 0) {
+ ALOGE("%s read failed: %s", path, strerror(errno));
+ close(fd);
+ return -1;
+ }
+ close(fd);
+
+ // field 10 is pgfault
+ // field 12 is pgmajfault
+ // field 22 is starttime
+ // field 24 is rss_in_pages
+ int64_t pgfault = 0, pgmajfault = 0, starttime = 0, rss_in_pages = 0;
+ if (sscanf(buffer,
+ "%*u %*s %*s %*d %*d %*d %*d %*d %*d %" SCNd64 " %*d "
+ "%" SCNd64 " %*d %*u %*u %*d %*d %*d %*d %*d %*d "
+ "%" SCNd64 " %*d %" SCNd64 "",
+ &pgfault, &pgmajfault, &starttime, &rss_in_pages) != 4) {
+ return -1;
+ }
+ mem_st->pgfault = pgfault;
+ mem_st->pgmajfault = pgmajfault;
+ mem_st->rss_in_bytes = (rss_in_pages * PAGE_SIZE);
+ mem_st->process_start_time_ns = starttime * (NS_PER_SEC / sysconf(_SC_CLK_TCK));
+ return 0;
}
+#endif
-static int zoneinfo_parse_protection(char *cp) {
- int max = 0;
- int zoneval;
+/* /prop/zoneinfo parsing routines */
+static int64_t zoneinfo_parse_protection(char *cp) {
+ int64_t max = 0;
+ long long zoneval;
char *save_ptr;
- for (cp = strtok_r(cp, "(), ", &save_ptr); cp; cp = strtok_r(NULL, "), ", &save_ptr)) {
- zoneval = strtol(cp, &cp, 0);
- if (zoneval > max)
- max = zoneval;
+ for (cp = strtok_r(cp, "(), ", &save_ptr); cp;
+ cp = strtok_r(NULL, "), ", &save_ptr)) {
+ zoneval = strtoll(cp, &cp, 0);
+ if (zoneval > max) {
+ max = (zoneval > INT64_MAX) ? INT64_MAX : zoneval;
+ }
}
return max;
}
-static void zoneinfo_parse_line(char *line, struct sysmeminfo *mip) {
+static bool zoneinfo_parse_line(char *line, union zoneinfo *zi) {
char *cp = line;
char *ap;
char *save_ptr;
+ int64_t val;
+ int field_idx;
cp = strtok_r(line, " ", &save_ptr);
- if (!cp)
- return;
+ if (!cp) {
+ return true;
+ }
- ap = strtok_r(NULL, " ", &save_ptr);
- if (!ap)
- return;
+ if (!strcmp(cp, "protection:")) {
+ ap = strtok_r(NULL, ")", &save_ptr);
+ } else {
+ ap = strtok_r(NULL, " ", &save_ptr);
+ }
- if (!strcmp(cp, "nr_free_pages"))
- mip->nr_free_pages += strtol(ap, NULL, 0);
- else if (!strcmp(cp, "nr_file_pages"))
- mip->nr_file_pages += strtol(ap, NULL, 0);
- else if (!strcmp(cp, "nr_shmem"))
- mip->nr_shmem += strtol(ap, NULL, 0);
- else if (!strcmp(cp, "high"))
- mip->totalreserve_pages += strtol(ap, NULL, 0);
- else if (!strcmp(cp, "protection:"))
- mip->totalreserve_pages += zoneinfo_parse_protection(ap);
+ if (!ap) {
+ return true;
+ }
+
+ switch (match_field(cp, ap, zoneinfo_field_names,
+ ZI_FIELD_COUNT, &val, &field_idx)) {
+ case (PARSE_SUCCESS):
+ zi->arr[field_idx] += val;
+ break;
+ case (NO_MATCH):
+ if (!strcmp(cp, "protection:")) {
+ zi->field.totalreserve_pages +=
+ zoneinfo_parse_protection(ap);
+ }
+ break;
+ case (PARSE_FAIL):
+ default:
+ return false;
+ }
+ return true;
}
-static int zoneinfo_parse(struct sysmeminfo *mip) {
- int fd;
- ssize_t size;
+static int zoneinfo_parse(union zoneinfo *zi) {
+ static struct reread_data file_data = {
+ .filename = ZONEINFO_PATH,
+ .fd = -1,
+ };
char buf[PAGE_SIZE];
char *save_ptr;
char *line;
- memset(mip, 0, sizeof(struct sysmeminfo));
+ memset(zi, 0, sizeof(union zoneinfo));
- fd = open(ZONEINFO_PATH, O_RDONLY | O_CLOEXEC);
- if (fd == -1) {
- ALOGE("%s open: errno=%d", ZONEINFO_PATH, errno);
+ if (reread_file(&file_data, buf, sizeof(buf)) < 0) {
return -1;
}
- size = read_all(fd, buf, sizeof(buf) - 1);
- if (size < 0) {
- ALOGE("%s read: errno=%d", ZONEINFO_PATH, errno);
- close(fd);
+ for (line = strtok_r(buf, "\n", &save_ptr); line;
+ line = strtok_r(NULL, "\n", &save_ptr)) {
+ if (!zoneinfo_parse_line(line, zi)) {
+ ALOGE("%s parse error", file_data.filename);
+ return -1;
+ }
+ }
+ zi->field.totalreserve_pages += zi->field.high;
+
+ return 0;
+}
+
+/* /prop/meminfo parsing routines */
+static bool meminfo_parse_line(char *line, union meminfo *mi) {
+ char *cp = line;
+ char *ap;
+ char *save_ptr;
+ int64_t val;
+ int field_idx;
+ enum field_match_result match_res;
+
+ cp = strtok_r(line, " ", &save_ptr);
+ if (!cp) {
+ return false;
+ }
+
+ ap = strtok_r(NULL, " ", &save_ptr);
+ if (!ap) {
+ return false;
+ }
+
+ match_res = match_field(cp, ap, meminfo_field_names, MI_FIELD_COUNT,
+ &val, &field_idx);
+ if (match_res == PARSE_SUCCESS) {
+ mi->arr[field_idx] = val / page_k;
+ }
+ return (match_res != PARSE_FAIL);
+}
+
+static int meminfo_parse(union meminfo *mi) {
+ static struct reread_data file_data = {
+ .filename = MEMINFO_PATH,
+ .fd = -1,
+ };
+ char buf[PAGE_SIZE];
+ char *save_ptr;
+ char *line;
+
+ memset(mi, 0, sizeof(union meminfo));
+
+ if (reread_file(&file_data, buf, sizeof(buf)) < 0) {
return -1;
}
- ALOG_ASSERT((size_t)size < sizeof(buf) - 1, "/proc/zoneinfo too large");
- buf[size] = 0;
- for (line = strtok_r(buf, "\n", &save_ptr); line; line = strtok_r(NULL, "\n", &save_ptr))
- zoneinfo_parse_line(line, mip);
+ for (line = strtok_r(buf, "\n", &save_ptr); line;
+ line = strtok_r(NULL, "\n", &save_ptr)) {
+ if (!meminfo_parse_line(line, mi)) {
+ ALOGE("%s parse error", file_data.filename);
+ return -1;
+ }
+ }
+ mi->field.nr_file_pages = mi->field.cached + mi->field.swap_cached +
+ mi->field.buffers;
- close(fd);
return 0;
}
+static void meminfo_log(union meminfo *mi) {
+ android_log_write_list_begin(ctx);
+ for (int field_idx = 0; field_idx < MI_FIELD_COUNT; field_idx++) {
+ android_log_write_int32(ctx, (int32_t)min(mi->arr[field_idx] * page_k, INT32_MAX));
+ }
+ android_log_write_list_end(ctx);
+ android_log_write_list(ctx, LOG_ID_EVENTS);
+}
+
static int proc_get_size(int pid) {
char path[PATH_MAX];
char line[LINE_MAX];
@@ -542,6 +1222,7 @@ static int proc_get_size(int pid) {
int total;
ssize_t ret;
+ /* gid containing AID_READPROC required */
snprintf(path, PATH_MAX, "/proc/%d/statm", pid);
fd = open(path, O_RDONLY | O_CLOEXEC);
if (fd == -1)
@@ -565,6 +1246,7 @@ static char *proc_get_name(int pid) {
char *cp;
ssize_t ret;
+ /* gid containing AID_READPROC required */
snprintf(path, PATH_MAX, "/proc/%d/cmdline", pid);
fd = open(path, O_RDONLY | O_CLOEXEC);
if (fd == -1)
@@ -586,87 +1268,207 @@ static struct proc *proc_adj_lru(int oomadj) {
return (struct proc *)adjslot_tail(&procadjslot_list[ADJTOSLOT(oomadj)]);
}
+static struct proc *proc_get_heaviest(int oomadj) {
+ struct adjslot_list *head = &procadjslot_list[ADJTOSLOT(oomadj)];
+ struct adjslot_list *curr = head->next;
+ struct proc *maxprocp = NULL;
+ int maxsize = 0;
+ while (curr != head) {
+ int pid = ((struct proc *)curr)->pid;
+ int tasksize = proc_get_size(pid);
+ if (tasksize <= 0) {
+ struct adjslot_list *next = curr->next;
+ pid_remove(pid);
+ curr = next;
+ } else {
+ if (tasksize > maxsize) {
+ maxsize = tasksize;
+ maxprocp = (struct proc *)curr;
+ }
+ curr = curr->next;
+ }
+ }
+ return maxprocp;
+}
+
+static void set_process_group_and_prio(int pid, SchedPolicy sp, int prio) {
+ DIR* d;
+ char proc_path[PATH_MAX];
+ struct dirent* de;
+
+ snprintf(proc_path, sizeof(proc_path), "/proc/%d/task", pid);
+ if (!(d = opendir(proc_path))) {
+ ALOGW("Failed to open %s; errno=%d: process pid(%d) might have died", proc_path, errno,
+ pid);
+ return;
+ }
+
+ while ((de = readdir(d))) {
+ int t_pid;
+
+ if (de->d_name[0] == '.') continue;
+ t_pid = atoi(de->d_name);
+
+ if (!t_pid) {
+ ALOGW("Failed to get t_pid for '%s' of pid(%d)", de->d_name, pid);
+ continue;
+ }
+
+ if (setpriority(PRIO_PROCESS, t_pid, prio) && errno != ESRCH) {
+ ALOGW("Unable to raise priority of killing t_pid (%d): errno=%d", t_pid, errno);
+ }
+
+ if (set_cpuset_policy(t_pid, sp)) {
+ ALOGW("Failed to set_cpuset_policy on pid(%d) t_pid(%d) to %d", pid, t_pid, (int)sp);
+ continue;
+ }
+ }
+ closedir(d);
+}
+
+static int last_killed_pid = -1;
+
/* Kill one process specified by procp. Returns the size of the process killed */
-static int kill_one_process(struct proc* procp, int min_score_adj, bool is_critical) {
+static int kill_one_process(struct proc* procp, int min_oom_score) {
int pid = procp->pid;
uid_t uid = procp->uid;
char *taskname;
int tasksize;
int r;
+ int result = -1;
+
+#ifdef LMKD_LOG_STATS
+ struct memory_stat mem_st = {};
+ int memory_stat_parse_result = -1;
+#else
+ /* To prevent unused parameter warning */
+ (void)(min_oom_score);
+#endif
taskname = proc_get_name(pid);
if (!taskname) {
- pid_remove(pid);
- return -1;
+ goto out;
}
tasksize = proc_get_size(pid);
if (tasksize <= 0) {
- pid_remove(pid);
- return -1;
+ goto out;
+ }
+
+#ifdef LMKD_LOG_STATS
+ if (enable_stats_log) {
+ if (per_app_memcg) {
+ memory_stat_parse_result = memory_stat_from_cgroup(&mem_st, pid, uid);
+ } else {
+ memory_stat_parse_result = memory_stat_from_procfs(&mem_st, pid);
+ }
}
+#endif
+
+ TRACE_KILL_START(pid);
- ALOGI(
- "Killing '%s' (%d), uid %d, adj %d\n"
- " to free %ldkB because system is under %s memory pressure oom_adj %d\n",
- taskname, pid, uid, procp->oomadj, tasksize * page_k, is_critical ? "critical" : "medium",
- min_score_adj);
+ /* CAP_KILL required */
r = kill(pid, SIGKILL);
- pid_remove(pid);
+
+ set_process_group_and_prio(pid, SP_FOREGROUND, ANDROID_PRIORITY_HIGHEST);
+
+ inc_killcnt(procp->oomadj);
+ ALOGE("Kill '%s' (%d), uid %d, oom_adj %d to free %ldkB", taskname, pid, uid, procp->oomadj,
+ tasksize * page_k);
+
+ TRACE_KILL_END();
+
+ last_killed_pid = pid;
if (r) {
- ALOGE("kill(%d): errno=%d", procp->pid, errno);
- return -1;
+ ALOGE("kill(%d): errno=%d", pid, errno);
+ goto out;
} else {
- return tasksize;
+#ifdef LMKD_LOG_STATS
+ if (memory_stat_parse_result == 0) {
+ stats_write_lmk_kill_occurred(log_ctx, LMK_KILL_OCCURRED, uid, taskname,
+ procp->oomadj, mem_st.pgfault, mem_st.pgmajfault, mem_st.rss_in_bytes,
+ mem_st.cache_in_bytes, mem_st.swap_in_bytes, mem_st.process_start_time_ns,
+ min_oom_score);
+ } else if (enable_stats_log) {
+ stats_write_lmk_kill_occurred(log_ctx, LMK_KILL_OCCURRED, uid, taskname, procp->oomadj,
+ -1, -1, tasksize * BYTES_IN_KILOBYTE, -1, -1, -1,
+ min_oom_score);
+ }
+#endif
+ result = tasksize;
}
+
+out:
+ /*
+ * WARNING: After pid_remove() procp is freed and can't be used!
+ * Therefore placed at the end of the function.
+ */
+ pid_remove(pid);
+ return result;
}
/*
- * Find a process to kill based on the current (possibly estimated) free memory
- * and cached memory sizes. Returns the size of the killed processes.
+ * Find one process to kill at or above the given oom_adj level.
+ * Returns size of the killed process.
*/
-static int find_and_kill_process(bool is_critical) {
+static int find_and_kill_process(int min_score_adj) {
int i;
int killed_size = 0;
- int min_score_adj = is_critical ? critical_oomadj : medium_oomadj;
+
+#ifdef LMKD_LOG_STATS
+ bool lmk_state_change_start = false;
+#endif
for (i = OOM_SCORE_ADJ_MAX; i >= min_score_adj; i--) {
struct proc *procp;
-retry:
- procp = proc_adj_lru(i);
-
- if (procp) {
- killed_size = kill_one_process(procp, min_score_adj, is_critical);
- if (killed_size < 0) {
- goto retry;
- } else {
- return killed_size;
+ while (true) {
+ procp = kill_heaviest_task ?
+ proc_get_heaviest(i) : proc_adj_lru(i);
+
+ if (!procp)
+ break;
+
+ killed_size = kill_one_process(procp, min_score_adj);
+ if (killed_size >= 0) {
+#ifdef LMKD_LOG_STATS
+ if (enable_stats_log && !lmk_state_change_start) {
+ lmk_state_change_start = true;
+ stats_write_lmk_state_changed(log_ctx, LMK_STATE_CHANGED,
+ LMK_STATE_CHANGE_START);
+ }
+#endif
+ break;
}
}
+ if (killed_size) {
+ break;
+ }
}
- return 0;
+#ifdef LMKD_LOG_STATS
+ if (enable_stats_log && lmk_state_change_start) {
+ stats_write_lmk_state_changed(log_ctx, LMK_STATE_CHANGED, LMK_STATE_CHANGE_STOP);
+ }
+#endif
+
+ return killed_size;
}
-static int64_t get_memory_usage(const char* path) {
+static int64_t get_memory_usage(struct reread_data *file_data) {
int ret;
int64_t mem_usage;
char buf[32];
- int fd = open(path, O_RDONLY | O_CLOEXEC);
- if (fd == -1) {
- ALOGE("%s open: errno=%d", path, errno);
+
+ if (reread_file(file_data, buf, sizeof(buf)) < 0) {
return -1;
}
- ret = read_all(fd, buf, sizeof(buf) - 1);
- close(fd);
- if (ret < 0) {
- ALOGE("%s error: errno=%d", path, errno);
+ if (!parse_int64(buf, &mem_usage)) {
+ ALOGE("%s parse error", file_data->filename);
return -1;
}
- sscanf(buf, "%" SCNd64, &mem_usage);
if (mem_usage == 0) {
ALOGE("No memory!");
return -1;
@@ -674,76 +1476,337 @@ static int64_t get_memory_usage(const char* path) {
return mem_usage;
}
-static void mp_event_common(bool is_critical) {
+void record_low_pressure_levels(union meminfo *mi) {
+ if (low_pressure_mem.min_nr_free_pages == -1 ||
+ low_pressure_mem.min_nr_free_pages > mi->field.nr_free_pages) {
+ if (debug_process_killing) {
+ ALOGI("Low pressure min memory update from %" PRId64 " to %" PRId64,
+ low_pressure_mem.min_nr_free_pages, mi->field.nr_free_pages);
+ }
+ low_pressure_mem.min_nr_free_pages = mi->field.nr_free_pages;
+ }
+ /*
+ * Free memory at low vmpressure events occasionally gets spikes,
+ * possibly a stale low vmpressure event with memory already
+ * freed up (no memory pressure should have been reported).
+ * Ignore large jumps in max_nr_free_pages that would mess up our stats.
+ */
+ if (low_pressure_mem.max_nr_free_pages == -1 ||
+ (low_pressure_mem.max_nr_free_pages < mi->field.nr_free_pages &&
+ mi->field.nr_free_pages - low_pressure_mem.max_nr_free_pages <
+ low_pressure_mem.max_nr_free_pages * 0.1)) {
+ if (debug_process_killing) {
+ ALOGI("Low pressure max memory update from %" PRId64 " to %" PRId64,
+ low_pressure_mem.max_nr_free_pages, mi->field.nr_free_pages);
+ }
+ low_pressure_mem.max_nr_free_pages = mi->field.nr_free_pages;
+ }
+}
+
+enum vmpressure_level upgrade_level(enum vmpressure_level level) {
+ return (enum vmpressure_level)((level < VMPRESS_LEVEL_CRITICAL) ?
+ level + 1 : level);
+}
+
+enum vmpressure_level downgrade_level(enum vmpressure_level level) {
+ return (enum vmpressure_level)((level > VMPRESS_LEVEL_LOW) ?
+ level - 1 : level);
+}
+
+static bool is_kill_pending(void) {
+ char buf[24];
+
+ if (last_killed_pid < 0) {
+ return false;
+ }
+
+ snprintf(buf, sizeof(buf), "/proc/%d/", last_killed_pid);
+ if (access(buf, F_OK) == 0) {
+ return true;
+ }
+
+ // reset last killed PID because there's nothing pending
+ last_killed_pid = -1;
+ return false;
+}
+
+static void mp_event_common(int data, uint32_t events __unused) {
int ret;
unsigned long long evcount;
- int index = is_critical ? CRITICAL_INDEX : MEDIUM_INDEX;
int64_t mem_usage, memsw_usage;
int64_t mem_pressure;
+ enum vmpressure_level lvl;
+ union meminfo mi;
+ union zoneinfo zi;
+ struct timespec curr_tm;
+ static struct timespec last_kill_tm;
+ static unsigned long kill_skip_count = 0;
+ enum vmpressure_level level = (enum vmpressure_level)data;
+ long other_free = 0, other_file = 0;
+ int min_score_adj;
+ int minfree = 0;
+ static struct reread_data mem_usage_file_data = {
+ .filename = MEMCG_MEMORY_USAGE,
+ .fd = -1,
+ };
+ static struct reread_data memsw_usage_file_data = {
+ .filename = MEMCG_MEMORYSW_USAGE,
+ .fd = -1,
+ };
+
+ if (debug_process_killing) {
+ ALOGI("%s memory pressure event is triggered", level_name[level]);
+ }
+
+ if (!use_psi_monitors) {
+ /*
+ * Check all event counters from low to critical
+ * and upgrade to the highest priority one. By reading
+ * eventfd we also reset the event counters.
+ */
+ for (lvl = VMPRESS_LEVEL_LOW; lvl < VMPRESS_LEVEL_COUNT; lvl++) {
+ if (mpevfd[lvl] != -1 &&
+ TEMP_FAILURE_RETRY(read(mpevfd[lvl],
+ &evcount, sizeof(evcount))) > 0 &&
+ evcount > 0 && lvl > level) {
+ level = lvl;
+ }
+ }
+ }
+
+ if (clock_gettime(CLOCK_MONOTONIC_COARSE, &curr_tm) != 0) {
+ ALOGE("Failed to get current time");
+ return;
+ }
+
+ if (kill_timeout_ms) {
+ // If we're within the timeout, see if there's pending reclaim work
+ // from the last killed process. If there is (as evidenced by
+ // /proc/<pid> continuing to exist), skip killing for now.
+ if ((get_time_diff_ms(&last_kill_tm, &curr_tm) < (long) kill_timeout_ms) &&
+ (low_ram_device || is_kill_pending())) {
+ kill_skip_count++;
+ return;
+ }
+ }
+
+ if (kill_skip_count > 0) {
+ ALOGI("%lu memory pressure events were skipped after a kill!",
+ kill_skip_count);
+ kill_skip_count = 0;
+ }
+
+ if (meminfo_parse(&mi) < 0 || zoneinfo_parse(&zi) < 0) {
+ ALOGE("Failed to get free memory!");
+ return;
+ }
+
+ if (use_minfree_levels) {
+ int i;
+
+ other_free = mi.field.nr_free_pages - zi.field.totalreserve_pages;
+ if (mi.field.nr_file_pages > (mi.field.shmem + mi.field.unevictable + mi.field.swap_cached)) {
+ other_file = (mi.field.nr_file_pages - mi.field.shmem -
+ mi.field.unevictable - mi.field.swap_cached);
+ } else {
+ other_file = 0;
+ }
+
+ min_score_adj = OOM_SCORE_ADJ_MAX + 1;
+ for (i = 0; i < lowmem_targets_size; i++) {
+ minfree = lowmem_minfree[i];
+ if (other_free < minfree && other_file < minfree) {
+ min_score_adj = lowmem_adj[i];
+ break;
+ }
+ }
+
+ if (min_score_adj == OOM_SCORE_ADJ_MAX + 1) {
+ if (debug_process_killing) {
+ ALOGI("Ignore %s memory pressure event "
+ "(free memory=%ldkB, cache=%ldkB, limit=%ldkB)",
+ level_name[level], other_free * page_k, other_file * page_k,
+ (long)lowmem_minfree[lowmem_targets_size - 1] * page_k);
+ }
+ return;
+ }
- ret = read(mpevfd[index], &evcount, sizeof(evcount));
- if (ret < 0)
- ALOGE("Error reading memory pressure event fd; errno=%d",
- errno);
+ goto do_kill;
+ }
+
+ if (level == VMPRESS_LEVEL_LOW) {
+ record_low_pressure_levels(&mi);
+ }
- mem_usage = get_memory_usage(MEMCG_MEMORY_USAGE);
- memsw_usage = get_memory_usage(MEMCG_MEMORYSW_USAGE);
- if (memsw_usage < 0 || mem_usage < 0) {
- find_and_kill_process(is_critical);
+ if (level_oomadj[level] > OOM_SCORE_ADJ_MAX) {
+ /* Do not monitor this pressure level */
return;
}
+ if ((mem_usage = get_memory_usage(&mem_usage_file_data)) < 0) {
+ goto do_kill;
+ }
+ if ((memsw_usage = get_memory_usage(&memsw_usage_file_data)) < 0) {
+ goto do_kill;
+ }
+
// Calculate percent for swappinness.
mem_pressure = (mem_usage * 100) / memsw_usage;
- if (enable_pressure_upgrade && !is_critical) {
+ if (enable_pressure_upgrade && level != VMPRESS_LEVEL_CRITICAL) {
// We are swapping too much.
if (mem_pressure < upgrade_pressure) {
- ALOGI("Event upgraded to critical.");
- is_critical = true;
+ level = upgrade_level(level);
+ if (debug_process_killing) {
+ ALOGI("Event upgraded to %s", level_name[level]);
+ }
}
}
- // If the pressure is larger than downgrade_pressure lmk will not
- // kill any process, since enough memory is available.
- if (mem_pressure > downgrade_pressure) {
- if (debug_process_killing) {
- ALOGI("Ignore %s memory pressure", is_critical ? "critical" : "medium");
- }
- return;
- } else if (is_critical && mem_pressure > upgrade_pressure) {
- if (debug_process_killing) {
- ALOGI("Downgrade critical memory pressure");
+ // If we still have enough swap space available, check if we want to
+ // ignore/downgrade pressure events.
+ if (mi.field.free_swap >=
+ mi.field.total_swap * swap_free_low_percentage / 100) {
+ // If the pressure is larger than downgrade_pressure lmk will not
+ // kill any process, since enough memory is available.
+ if (mem_pressure > downgrade_pressure) {
+ if (debug_process_killing) {
+ ALOGI("Ignore %s memory pressure", level_name[level]);
+ }
+ return;
+ } else if (level == VMPRESS_LEVEL_CRITICAL && mem_pressure > upgrade_pressure) {
+ if (debug_process_killing) {
+ ALOGI("Downgrade critical memory pressure");
+ }
+ // Downgrade event, since enough memory available.
+ level = downgrade_level(level);
}
- // Downgrade event to medium, since enough memory available.
- is_critical = false;
}
- if (find_and_kill_process(is_critical) == 0) {
- if (debug_process_killing) {
- ALOGI("Nothing to kill");
+do_kill:
+ if (low_ram_device) {
+ /* For Go devices kill only one task */
+ if (find_and_kill_process(level_oomadj[level]) == 0) {
+ if (debug_process_killing) {
+ ALOGI("Nothing to kill");
+ }
+ } else {
+ meminfo_log(&mi);
+ }
+ } else {
+ int pages_freed;
+ static struct timespec last_report_tm;
+ static unsigned long report_skip_count = 0;
+
+ if (!use_minfree_levels) {
+ /* Free up enough memory to downgrate the memory pressure to low level */
+ if (mi.field.nr_free_pages >= low_pressure_mem.max_nr_free_pages) {
+ if (debug_process_killing) {
+ ALOGI("Ignoring pressure since more memory is "
+ "available (%" PRId64 ") than watermark (%" PRId64 ")",
+ mi.field.nr_free_pages, low_pressure_mem.max_nr_free_pages);
+ }
+ return;
+ }
+ min_score_adj = level_oomadj[level];
+ }
+
+ pages_freed = find_and_kill_process(min_score_adj);
+
+ if (pages_freed == 0) {
+ /* Rate limit kill reports when nothing was reclaimed */
+ if (get_time_diff_ms(&last_report_tm, &curr_tm) < FAIL_REPORT_RLIMIT_MS) {
+ report_skip_count++;
+ return;
+ }
+ } else {
+ /* If we killed anything, update the last killed timestamp. */
+ last_kill_tm = curr_tm;
}
+
+ /* Log meminfo whenever we kill or when report rate limit allows */
+ meminfo_log(&mi);
+
+ if (use_minfree_levels) {
+ ALOGI("Reclaimed %ldkB, cache(%ldkB) and "
+ "free(%" PRId64 "kB)-reserved(%" PRId64 "kB) below min(%ldkB) for oom_adj %d",
+ pages_freed * page_k,
+ other_file * page_k, mi.field.nr_free_pages * page_k,
+ zi.field.totalreserve_pages * page_k,
+ minfree * page_k, min_score_adj);
+ } else {
+ ALOGI("Reclaimed %ldkB at oom_adj %d",
+ pages_freed * page_k, min_score_adj);
+ }
+
+ if (report_skip_count > 0) {
+ ALOGI("Suppressed %lu failed kill reports", report_skip_count);
+ report_skip_count = 0;
+ }
+
+ last_report_tm = curr_tm;
+ }
+}
+
+static bool init_mp_psi(enum vmpressure_level level) {
+ int fd = init_psi_monitor(psi_thresholds[level].stall_type,
+ psi_thresholds[level].threshold_ms * US_PER_MS,
+ PSI_WINDOW_SIZE_MS * US_PER_MS);
+
+ if (fd < 0) {
+ return false;
}
+
+ vmpressure_hinfo[level].handler = mp_event_common;
+ vmpressure_hinfo[level].data = level;
+ if (register_psi_monitor(epollfd, fd, &vmpressure_hinfo[level]) < 0) {
+ destroy_psi_monitor(fd);
+ return false;
+ }
+ maxevents++;
+ mpevfd[level] = fd;
+
+ return true;
}
-static void mp_event(uint32_t events __unused) {
- mp_event_common(false);
+static void destroy_mp_psi(enum vmpressure_level level) {
+ int fd = mpevfd[level];
+
+ if (unregister_psi_monitor(epollfd, fd) < 0) {
+ ALOGE("Failed to unregister psi monitor for %s memory pressure; errno=%d",
+ level_name[level], errno);
+ }
+ destroy_psi_monitor(fd);
+ mpevfd[level] = -1;
}
-static void mp_event_critical(uint32_t events __unused) {
- mp_event_common(true);
+static bool init_psi_monitors() {
+ if (!init_mp_psi(VMPRESS_LEVEL_LOW)) {
+ return false;
+ }
+ if (!init_mp_psi(VMPRESS_LEVEL_MEDIUM)) {
+ destroy_mp_psi(VMPRESS_LEVEL_LOW);
+ return false;
+ }
+ if (!init_mp_psi(VMPRESS_LEVEL_CRITICAL)) {
+ destroy_mp_psi(VMPRESS_LEVEL_MEDIUM);
+ destroy_mp_psi(VMPRESS_LEVEL_LOW);
+ return false;
+ }
+ return true;
}
-static int init_mp_common(char *levelstr, void *event_handler, bool is_critical)
-{
+static bool init_mp_common(enum vmpressure_level level) {
int mpfd;
int evfd;
int evctlfd;
char buf[256];
struct epoll_event epev;
int ret;
- int mpevfd_index = is_critical ? CRITICAL_INDEX : MEDIUM_INDEX;
+ int level_idx = (int)level;
+ const char *levelstr = level_name[level_idx];
+ /* gid containing AID_SYSTEM required */
mpfd = open(MEMCG_SYSFS_PATH "memory.pressure_level", O_RDONLY | O_CLOEXEC);
if (mpfd < 0) {
ALOGI("No kernel memory.pressure_level support (errno=%d)", errno);
@@ -768,7 +1831,7 @@ static int init_mp_common(char *levelstr, void *event_handler, bool is_critical)
goto err;
}
- ret = write(evctlfd, buf, strlen(buf) + 1);
+ ret = TEMP_FAILURE_RETRY(write(evctlfd, buf, strlen(buf) + 1));
if (ret == -1) {
ALOGE("cgroup.event_control write failed for level %s; errno=%d",
levelstr, errno);
@@ -776,15 +1839,19 @@ static int init_mp_common(char *levelstr, void *event_handler, bool is_critical)
}
epev.events = EPOLLIN;
- epev.data.ptr = event_handler;
+ /* use data to store event level */
+ vmpressure_hinfo[level_idx].data = level_idx;
+ vmpressure_hinfo[level_idx].handler = mp_event_common;
+ epev.data.ptr = (void *)&vmpressure_hinfo[level_idx];
ret = epoll_ctl(epollfd, EPOLL_CTL_ADD, evfd, &epev);
if (ret == -1) {
ALOGE("epoll_ctl for level %s failed; errno=%d", levelstr, errno);
goto err;
}
maxevents++;
- mpevfd[mpevfd_index] = evfd;
- return 0;
+ mpevfd[level] = evfd;
+ close(evctlfd);
+ return true;
err:
close(evfd);
@@ -793,17 +1860,7 @@ err_eventfd:
err_open_evctlfd:
close(mpfd);
err_open_mpfd:
- return -1;
-}
-
-static int init_mp_medium()
-{
- return init_mp_common(MEMPRESSURE_WATCH_MEDIUM_LEVEL, (void *)&mp_event, false);
-}
-
-static int init_mp_critical()
-{
- return init_mp_common(MEMPRESSURE_WATCH_CRITICAL_LEVEL, (void *)&mp_event_critical, true);
+ return false;
}
static int init(void) {
@@ -822,36 +1879,54 @@ static int init(void) {
return -1;
}
- ctrl_lfd = android_get_control_socket("lmkd");
- if (ctrl_lfd < 0) {
+ // mark data connections as not connected
+ for (int i = 0; i < MAX_DATA_CONN; i++) {
+ data_sock[i].sock = -1;
+ }
+
+ ctrl_sock.sock = android_get_control_socket("lmkd");
+ if (ctrl_sock.sock < 0) {
ALOGE("get lmkd control socket failed");
return -1;
}
- ret = listen(ctrl_lfd, 1);
+ ret = listen(ctrl_sock.sock, MAX_DATA_CONN);
if (ret < 0) {
ALOGE("lmkd control socket listen failed (errno=%d)", errno);
return -1;
}
epev.events = EPOLLIN;
- epev.data.ptr = (void *)ctrl_connect_handler;
- if (epoll_ctl(epollfd, EPOLL_CTL_ADD, ctrl_lfd, &epev) == -1) {
+ ctrl_sock.handler_info.handler = ctrl_connect_handler;
+ epev.data.ptr = (void *)&(ctrl_sock.handler_info);
+ if (epoll_ctl(epollfd, EPOLL_CTL_ADD, ctrl_sock.sock, &epev) == -1) {
ALOGE("epoll_ctl for lmkd control socket failed (errno=%d)", errno);
return -1;
}
maxevents++;
has_inkernel_module = !access(INKERNEL_MINFREE_PATH, W_OK);
- use_inkernel_interface = has_inkernel_module && !is_go_device;
+ use_inkernel_interface = has_inkernel_module;
if (use_inkernel_interface) {
ALOGI("Using in-kernel low memory killer interface");
} else {
- ret = init_mp_medium();
- ret |= init_mp_critical();
- if (ret)
+ /* Try to use psi monitor first if kernel has it */
+ use_psi_monitors = property_get_bool("ro.lmk.use_psi", true) &&
+ init_psi_monitors();
+ /* Fall back to vmpressure */
+ if (!use_psi_monitors &&
+ (!init_mp_common(VMPRESS_LEVEL_LOW) ||
+ !init_mp_common(VMPRESS_LEVEL_MEDIUM) ||
+ !init_mp_common(VMPRESS_LEVEL_CRITICAL))) {
ALOGE("Kernel does not support memory pressure events or in-kernel low memory killer");
+ return -1;
+ }
+ if (use_psi_monitors) {
+ ALOGI("Using psi monitors for memory pressure detection");
+ } else {
+ ALOGI("Using vmpressure for memory pressure detection");
+ }
}
for (i = 0; i <= ADJTOSLOT(OOM_SCORE_ADJ_MAX); i++) {
@@ -859,17 +1934,44 @@ static int init(void) {
procadjslot_list[i].prev = &procadjslot_list[i];
}
+ memset(killcnt_idx, KILLCNT_INVALID_IDX, sizeof(killcnt_idx));
+
return 0;
}
static void mainloop(void) {
+ struct event_handler_info* handler_info;
+ struct event_handler_info* poll_handler = NULL;
+ struct timespec last_report_tm, curr_tm;
+ struct epoll_event *evt;
+ long delay = -1;
+ int polling = 0;
+
while (1) {
struct epoll_event events[maxevents];
int nevents;
int i;
- ctrl_dfd_reopened = 0;
- nevents = epoll_wait(epollfd, events, maxevents, -1);
+ if (polling) {
+ /* Calculate next timeout */
+ clock_gettime(CLOCK_MONOTONIC_COARSE, &curr_tm);
+ delay = get_time_diff_ms(&last_report_tm, &curr_tm);
+ delay = (delay < PSI_POLL_PERIOD_MS) ?
+ PSI_POLL_PERIOD_MS - delay : PSI_POLL_PERIOD_MS;
+
+ /* Wait for events until the next polling timeout */
+ nevents = epoll_wait(epollfd, events, maxevents, delay);
+
+ clock_gettime(CLOCK_MONOTONIC_COARSE, &curr_tm);
+ if (get_time_diff_ms(&last_report_tm, &curr_tm) >= PSI_POLL_PERIOD_MS) {
+ polling--;
+ poll_handler->handler(poll_handler->data, 0);
+ last_report_tm = curr_tm;
+ }
+ } else {
+ /* Wait for events with no timeout */
+ nevents = epoll_wait(epollfd, events, maxevents, -1);
+ }
if (nevents == -1) {
if (errno == EINTR)
@@ -878,11 +1980,44 @@ static void mainloop(void) {
continue;
}
- for (i = 0; i < nevents; ++i) {
- if (events[i].events & EPOLLERR)
+ /*
+ * First pass to see if any data socket connections were dropped.
+ * Dropped connection should be handled before any other events
+ * to deallocate data connection and correctly handle cases when
+ * connection gets dropped and reestablished in the same epoll cycle.
+ * In such cases it's essential to handle connection closures first.
+ */
+ for (i = 0, evt = &events[0]; i < nevents; ++i, evt++) {
+ if ((evt->events & EPOLLHUP) && evt->data.ptr) {
+ ALOGI("lmkd data connection dropped");
+ handler_info = (struct event_handler_info*)evt->data.ptr;
+ ctrl_data_close(handler_info->data);
+ }
+ }
+
+ /* Second pass to handle all other events */
+ for (i = 0, evt = &events[0]; i < nevents; ++i, evt++) {
+ if (evt->events & EPOLLERR)
ALOGD("EPOLLERR on event #%d", i);
- if (events[i].data.ptr)
- (*(void (*)(uint32_t))events[i].data.ptr)(events[i].events);
+ if (evt->events & EPOLLHUP) {
+ /* This case was handled in the first pass */
+ continue;
+ }
+ if (evt->data.ptr) {
+ handler_info = (struct event_handler_info*)evt->data.ptr;
+ handler_info->handler(handler_info->data, evt->events);
+
+ if (use_psi_monitors && handler_info->handler == mp_event_common) {
+ /*
+ * Poll for the duration of PSI_WINDOW_SIZE_MS after the
+ * initial PSI event because psi events are rate-limited
+ * at one per sec.
+ */
+ polling = PSI_POLL_COUNT;
+ poll_handler = handler_info;
+ clock_gettime(CLOCK_MONOTONIC_COARSE, &last_report_tm);
+ }
+ }
}
}
}
@@ -892,18 +2027,72 @@ int main(int argc __unused, char **argv __unused) {
.sched_priority = 1,
};
- medium_oomadj = property_get_int32("ro.lmk.medium", 800);
- critical_oomadj = property_get_int32("ro.lmk.critical", 0);
+ /* By default disable low level vmpressure events */
+ level_oomadj[VMPRESS_LEVEL_LOW] =
+ property_get_int32("ro.lmk.low", OOM_SCORE_ADJ_MAX + 1);
+ level_oomadj[VMPRESS_LEVEL_MEDIUM] =
+ property_get_int32("ro.lmk.medium", 800);
+ level_oomadj[VMPRESS_LEVEL_CRITICAL] =
+ property_get_int32("ro.lmk.critical", 0);
debug_process_killing = property_get_bool("ro.lmk.debug", false);
- enable_pressure_upgrade = property_get_bool("ro.lmk.critical_upgrade", false);
- upgrade_pressure = (int64_t)property_get_int32("ro.lmk.upgrade_pressure", 50);
- downgrade_pressure = (int64_t)property_get_int32("ro.lmk.downgrade_pressure", 60);
- is_go_device = property_get_bool("ro.config.low_ram", false);
-
- mlockall(MCL_FUTURE);
- sched_setscheduler(0, SCHED_FIFO, &param);
- if (!init())
+
+ /* By default disable upgrade/downgrade logic */
+ enable_pressure_upgrade =
+ property_get_bool("ro.lmk.critical_upgrade", false);
+ upgrade_pressure =
+ (int64_t)property_get_int32("ro.lmk.upgrade_pressure", 100);
+ downgrade_pressure =
+ (int64_t)property_get_int32("ro.lmk.downgrade_pressure", 100);
+ kill_heaviest_task =
+ property_get_bool("ro.lmk.kill_heaviest_task", false);
+ low_ram_device = property_get_bool("ro.config.low_ram", false);
+ kill_timeout_ms =
+ (unsigned long)property_get_int32("ro.lmk.kill_timeout_ms", 0);
+ use_minfree_levels =
+ property_get_bool("ro.lmk.use_minfree_levels", false);
+ per_app_memcg =
+ property_get_bool("ro.config.per_app_memcg", low_ram_device);
+ swap_free_low_percentage =
+ property_get_int32("ro.lmk.swap_free_low_percentage", 10);
+
+ ctx = create_android_logger(MEMINFO_LOG_TAG);
+
+#ifdef LMKD_LOG_STATS
+ statslog_init(&log_ctx, &enable_stats_log);
+#endif
+
+ if (!init()) {
+ if (!use_inkernel_interface) {
+ /*
+ * MCL_ONFAULT pins pages as they fault instead of loading
+ * everything immediately all at once. (Which would be bad,
+ * because as of this writing, we have a lot of mapped pages we
+ * never use.) Old kernels will see MCL_ONFAULT and fail with
+ * EINVAL; we ignore this failure.
+ *
+ * N.B. read the man page for mlockall. MCL_CURRENT | MCL_ONFAULT
+ * pins ⊆ MCL_CURRENT, converging to just MCL_CURRENT as we fault
+ * in pages.
+ */
+ /* CAP_IPC_LOCK required */
+ if (mlockall(MCL_CURRENT | MCL_FUTURE | MCL_ONFAULT) && (errno != EINVAL)) {
+ ALOGW("mlockall failed %s", strerror(errno));
+ }
+
+ /* CAP_NICE required */
+ if (sched_setscheduler(0, SCHED_FIFO, &param)) {
+ ALOGW("set SCHED_FIFO failed %s", strerror(errno));
+ }
+ }
+
mainloop();
+ }
+
+#ifdef LMKD_LOG_STATS
+ statslog_destroy(&log_ctx);
+#endif
+
+ android_log_destroy(&ctx);
ALOGI("exiting");
return 0;
--
2.45.1
Reference in New Issue View Git Blame Copy Permalink