platform_system_core/init/service.cpp

1134 lines
36 KiB
C++
Raw Normal View History

/*
* Copyright (C) 2015 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.
*/
#include "service.h"
#include <fcntl.h>
#include <inttypes.h>
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
#include <linux/securebits.h>
#include <sched.h>
#include <sys/mount.h>
#include <sys/prctl.h>
#include <sys/resource.h>
#include <sys/stat.h>
#include <sys/system_properties.h>
#include <sys/time.h>
#include <sys/wait.h>
#include <termios.h>
#include <unistd.h>
#include <android-base/file.h>
#include <android-base/logging.h>
#include <android-base/parseint.h>
#include <android-base/properties.h>
#include <android-base/stringprintf.h>
#include <android-base/strings.h>
#include <processgroup/processgroup.h>
#include <selinux/selinux.h>
#include <system/thread_defs.h>
#include "init.h"
#include "property_service.h"
#include "util.h"
using android::base::boot_clock;
using android::base::ParseInt;
using android::base::StringPrintf;
using android::base::WriteStringToFile;
static std::string ComputeContextFromExecutable(std::string& service_name,
const std::string& service_path) {
std::string computed_context;
char* raw_con = nullptr;
char* raw_filecon = nullptr;
if (getcon(&raw_con) == -1) {
LOG(ERROR) << "could not get context while starting '" << service_name << "'";
return "";
}
std::unique_ptr<char> mycon(raw_con);
if (getfilecon(service_path.c_str(), &raw_filecon) == -1) {
LOG(ERROR) << "could not get file context while starting '" << service_name << "'";
return "";
}
std::unique_ptr<char> filecon(raw_filecon);
char* new_con = nullptr;
int rc = security_compute_create(mycon.get(), filecon.get(),
string_to_security_class("process"), &new_con);
if (rc == 0) {
computed_context = new_con;
free(new_con);
}
if (rc == 0 && computed_context == mycon.get()) {
LOG(ERROR) << "service " << service_name << " does not have a SELinux domain defined";
return "";
}
if (rc < 0) {
LOG(ERROR) << "could not get context while starting '" << service_name << "'";
return "";
}
return computed_context;
}
static void SetUpPidNamespace(const std::string& service_name) {
constexpr unsigned int kSafeFlags = MS_NODEV | MS_NOEXEC | MS_NOSUID;
// It's OK to LOG(FATAL) in this function since it's running in the first
// child process.
if (mount("", "/proc", "proc", kSafeFlags | MS_REMOUNT, "") == -1) {
PLOG(FATAL) << "couldn't remount(/proc) for " << service_name;
}
if (prctl(PR_SET_NAME, service_name.c_str()) == -1) {
PLOG(FATAL) << "couldn't set name for " << service_name;
}
pid_t child_pid = fork();
if (child_pid == -1) {
PLOG(FATAL) << "couldn't fork init inside the PID namespace for " << service_name;
}
if (child_pid > 0) {
// So that we exit with the right status.
static int init_exitstatus = 0;
signal(SIGTERM, [](int) { _exit(init_exitstatus); });
pid_t waited_pid;
int status;
while ((waited_pid = wait(&status)) > 0) {
// This loop will end when there are no processes left inside the
// PID namespace or when the init process inside the PID namespace
// gets a signal.
if (waited_pid == child_pid) {
init_exitstatus = status;
}
}
if (!WIFEXITED(init_exitstatus)) {
_exit(EXIT_FAILURE);
}
_exit(WEXITSTATUS(init_exitstatus));
}
}
static void ExpandArgs(const std::vector<std::string>& args, std::vector<char*>* strs) {
std::vector<std::string> expanded_args;
expanded_args.resize(args.size());
strs->push_back(const_cast<char*>(args[0].c_str()));
for (std::size_t i = 1; i < args.size(); ++i) {
if (!expand_props(args[i], &expanded_args[i])) {
LOG(FATAL) << args[0] << ": cannot expand '" << args[i] << "'";
}
strs->push_back(const_cast<char*>(expanded_args[i].c_str()));
}
strs->push_back(nullptr);
}
ServiceEnvironmentInfo::ServiceEnvironmentInfo() {
}
ServiceEnvironmentInfo::ServiceEnvironmentInfo(const std::string& name,
const std::string& value)
: name(name), value(value) {
}
Service::Service(const std::string& name, const std::vector<std::string>& args)
: name_(name),
classnames_({"default"}),
flags_(0),
pid_(0),
crash_count_(0),
uid_(0),
gid_(0),
namespace_flags_(0),
seclabel_(""),
init: Stop combining actions In the past, I had thought it didn't make sense to have multiple Action classes with identical triggers within ActionManager::actions_, and opted to instead combine these into a single action. In theory, it should reduce memory overhead as only one copy of the triggers needs to be stored. In practice, this ends up not being a good idea. Most importantly, given a file with the below three sections in this same order: on boot setprop a b on boot && property:true=true setprop c d on boot setprop e f Assuming that property 'true' == 'true', when the `boot` event happens, the order of the setprop commands will actually be: setprop a b setprop e f setprop c d instead of the more intuitive order of: setprop a b setprop c d setprop e f This is a mistake and this CL fixes it. It also documents this order. Secondly, with a given 'Action' now spanning multiple files, in order to keep track of which file a command is run from, the 'Command' itself needs to store this. Ironically to the original intention, this increases total ram usage. This change now only stores the file name in each 'Action' instead of each 'Command'. All in all this is a negligible trade off of ram usage. Thirdly, this requires a bunch of extra code and assumptions that don't help anything else. In particular it forces to keep property triggers sorted for easy comparison, which I'm using an std::map for currently, but that is not the best data structure to contain them. Lastly, I added the filename and line number to the 'processing action' LOG(INFO) message. Test: Boot bullhead, observe above changes Test: Boot sailfish, observe no change in boot time Change-Id: I3fbcac4ee677351314e33012c758145be82346e9
2017-04-18 22:21:54 +02:00
onrestart_(false, "<Service '" + name + "' onrestart>", 0),
ioprio_class_(IoSchedClass_NONE),
ioprio_pri_(0),
priority_(0),
oom_score_adjust_(-1000),
args_(args) {
onrestart_.InitSingleTrigger("onrestart");
}
Service::Service(const std::string& name, unsigned flags, uid_t uid, gid_t gid,
const std::vector<gid_t>& supp_gids, const CapSet& capabilities,
unsigned namespace_flags, const std::string& seclabel,
const std::vector<std::string>& args)
: name_(name),
classnames_({"default"}),
flags_(flags),
pid_(0),
crash_count_(0),
uid_(uid),
gid_(gid),
supp_gids_(supp_gids),
capabilities_(capabilities),
namespace_flags_(namespace_flags),
seclabel_(seclabel),
init: Stop combining actions In the past, I had thought it didn't make sense to have multiple Action classes with identical triggers within ActionManager::actions_, and opted to instead combine these into a single action. In theory, it should reduce memory overhead as only one copy of the triggers needs to be stored. In practice, this ends up not being a good idea. Most importantly, given a file with the below three sections in this same order: on boot setprop a b on boot && property:true=true setprop c d on boot setprop e f Assuming that property 'true' == 'true', when the `boot` event happens, the order of the setprop commands will actually be: setprop a b setprop e f setprop c d instead of the more intuitive order of: setprop a b setprop c d setprop e f This is a mistake and this CL fixes it. It also documents this order. Secondly, with a given 'Action' now spanning multiple files, in order to keep track of which file a command is run from, the 'Command' itself needs to store this. Ironically to the original intention, this increases total ram usage. This change now only stores the file name in each 'Action' instead of each 'Command'. All in all this is a negligible trade off of ram usage. Thirdly, this requires a bunch of extra code and assumptions that don't help anything else. In particular it forces to keep property triggers sorted for easy comparison, which I'm using an std::map for currently, but that is not the best data structure to contain them. Lastly, I added the filename and line number to the 'processing action' LOG(INFO) message. Test: Boot bullhead, observe above changes Test: Boot sailfish, observe no change in boot time Change-Id: I3fbcac4ee677351314e33012c758145be82346e9
2017-04-18 22:21:54 +02:00
onrestart_(false, "<Service '" + name + "' onrestart>", 0),
ioprio_class_(IoSchedClass_NONE),
ioprio_pri_(0),
priority_(0),
oom_score_adjust_(-1000),
args_(args) {
onrestart_.InitSingleTrigger("onrestart");
}
void Service::NotifyStateChange(const std::string& new_state) const {
if ((flags_ & SVC_TEMPORARY) != 0) {
// Services created by 'exec' are temporary and don't have properties tracking their state.
return;
}
std::string prop_name = StringPrintf("init.svc.%s", name_.c_str());
property_set(prop_name.c_str(), new_state.c_str());
if (new_state == "running") {
uint64_t start_ns = time_started_.time_since_epoch().count();
Replace the "coldboot" timeout with a property. Also rename init's existing boot-time related properties so they're all "ro.*" properties. Example result: # Three properties showing when init started... [ro.boottime.init]: [5294587604] # ...how long it waited for ueventd... [ro.boottime.init.cold_boot_wait]: [646956470] # ...and how long SELinux initialization took... [ro.boottime.init.selinux]: [45742921] # Plus one property for each service, showing when it first started. [ro.boottime.InputEventFind]: [10278767840] [ro.boottime.adbd]: [8359267180] [ro.boottime.atfwd]: [10338554773] [ro.boottime.audioserver]: [10298157478] [ro.boottime.bootanim]: [9323670089] [ro.boottime.cameraserver]: [10299402321] [ro.boottime.cnd]: [10335931856] [ro.boottime.debuggerd]: [7001352774] [ro.boottime.debuggerd64]: [7002261785] [ro.boottime.drm]: [10301082113] [ro.boottime.fingerprintd]: [10331443314] [ro.boottime.flash-nanohub-fw]: [6995265534] [ro.boottime.gatekeeperd]: [10340355242] [ro.boottime.healthd]: [7856893380] [ro.boottime.hwservicemanager]: [7856051088] [ro.boottime.imscmservice]: [10290530758] [ro.boottime.imsdatadaemon]: [10358136702] [ro.boottime.imsqmidaemon]: [10289084872] [ro.boottime.installd]: [10303296020] [ro.boottime.irsc_util]: [10279807632] [ro.boottime.keystore]: [10305034093] [ro.boottime.lmkd]: [7863506714] [ro.boottime.loc_launcher]: [10324525241] [ro.boottime.logd]: [6526221633] [ro.boottime.logd-reinit]: [7850662702] [ro.boottime.mcfg-sh]: [10337268315] [ro.boottime.media]: [10312152687] [ro.boottime.mediacodec]: [10306852530] [ro.boottime.mediadrm]: [10308707999] [ro.boottime.mediaextractor]: [10310681177] [ro.boottime.msm_irqbalance]: [7862451974] [ro.boottime.netd]: [10313523104] [ro.boottime.netmgrd]: [10285009351] [ro.boottime.oem_qmi_server]: [10293329092] [ro.boottime.per_mgr]: [7857915776] [ro.boottime.per_proxy]: [8335121605] [ro.boottime.perfd]: [10283443101] [ro.boottime.qcamerasvr]: [10329644772] [ro.boottime.qmuxd]: [10282346643] [ro.boottime.qseecomd]: [6855708593] [ro.boottime.qti]: [10286196851] [ro.boottime.ril-daemon]: [10314933677] [ro.boottime.rmt_storage]: [7859105047] [ro.boottime.servicemanager]: [7864555881] [ro.boottime.ss_ramdump]: [8337634938] [ro.boottime.ssr_setup]: [8336268324] [ro.boottime.surfaceflinger]: [7866921402] [ro.boottime.thermal-engine]: [10281249924] [ro.boottime.time_daemon]: [10322006542] [ro.boottime.ueventd]: [5618663938] [ro.boottime.vold]: [7003493920] [ro.boottime.wificond]: [10316641073] [ro.boottime.wpa_supplicant]: [18959816881] [ro.boottime.zygote]: [10295295029] [ro.boottime.zygote_secondary]: [10296637269] Bug: http://b/31800756 Test: boots Change-Id: I094cce0c1bab9406d950ca94212689dc2e15dba5
2016-11-29 20:20:58 +01:00
property_set(StringPrintf("ro.boottime.%s", name_.c_str()).c_str(),
StringPrintf("%" PRIu64, start_ns).c_str());
}
}
void Service::KillProcessGroup(int signal) {
LOG(INFO) << "Sending signal " << signal
<< " to service '" << name_
<< "' (pid " << pid_ << ") process group...";
int r;
if (signal == SIGTERM) {
r = killProcessGroupOnce(uid_, pid_, signal);
} else {
r = killProcessGroup(uid_, pid_, signal);
}
if (r == -1) {
LOG(ERROR) << "killProcessGroup(" << uid_ << ", " << pid_ << ", " << signal << ") failed";
}
if (kill(-pid_, signal) == -1) {
PLOG(ERROR) << "kill(" << pid_ << ", " << signal << ") failed";
}
}
void Service::SetProcessAttributes() {
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
// Keep capabilites on uid change.
if (capabilities_.any() && uid_) {
if (prctl(PR_SET_SECUREBITS, SECBIT_KEEP_CAPS | SECBIT_KEEP_CAPS_LOCKED) != 0) {
PLOG(FATAL) << "prtcl(PR_SET_KEEPCAPS) failed for " << name_;
}
}
// TODO: work out why this fails for `console` then upgrade to FATAL.
if (setpgid(0, getpid()) == -1) PLOG(ERROR) << "setpgid failed for " << name_;
if (gid_) {
if (setgid(gid_) != 0) {
PLOG(FATAL) << "setgid failed for " << name_;
}
}
if (setgroups(supp_gids_.size(), &supp_gids_[0]) != 0) {
PLOG(FATAL) << "setgroups failed for " << name_;
}
if (uid_) {
if (setuid(uid_) != 0) {
PLOG(FATAL) << "setuid failed for " << name_;
}
}
if (!seclabel_.empty()) {
if (setexeccon(seclabel_.c_str()) < 0) {
PLOG(FATAL) << "cannot setexeccon('" << seclabel_ << "') for " << name_;
}
}
if (priority_ != 0) {
if (setpriority(PRIO_PROCESS, 0, priority_) != 0) {
PLOG(FATAL) << "setpriority failed for " << name_;
}
}
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
if (capabilities_.any()) {
if (!SetCapsForExec(capabilities_)) {
LOG(FATAL) << "cannot set capabilities for " << name_;
}
}
}
void Service::Reap() {
if (!(flags_ & SVC_ONESHOT) || (flags_ & SVC_RESTART)) {
KillProcessGroup(SIGKILL);
}
// Remove any descriptor resources we may have created.
std::for_each(descriptors_.begin(), descriptors_.end(),
std::bind(&DescriptorInfo::Clean, std::placeholders::_1));
if (flags_ & SVC_EXEC) {
LOG(INFO) << "SVC_EXEC pid " << pid_ << " finished...";
}
if (flags_ & SVC_TEMPORARY) {
return;
}
pid_ = 0;
flags_ &= (~SVC_RUNNING);
// Oneshot processes go into the disabled state on exit,
// except when manually restarted.
if ((flags_ & SVC_ONESHOT) && !(flags_ & SVC_RESTART)) {
flags_ |= SVC_DISABLED;
}
// Disabled and reset processes do not get restarted automatically.
if (flags_ & (SVC_DISABLED | SVC_RESET)) {
NotifyStateChange("stopped");
return;
}
// If we crash > 4 times in 4 minutes, reboot into recovery.
boot_clock::time_point now = boot_clock::now();
if ((flags_ & SVC_CRITICAL) && !(flags_ & SVC_RESTART)) {
if (now < time_crashed_ + 4min) {
if (++crash_count_ > 4) {
Replace the "coldboot" timeout with a property. Also rename init's existing boot-time related properties so they're all "ro.*" properties. Example result: # Three properties showing when init started... [ro.boottime.init]: [5294587604] # ...how long it waited for ueventd... [ro.boottime.init.cold_boot_wait]: [646956470] # ...and how long SELinux initialization took... [ro.boottime.init.selinux]: [45742921] # Plus one property for each service, showing when it first started. [ro.boottime.InputEventFind]: [10278767840] [ro.boottime.adbd]: [8359267180] [ro.boottime.atfwd]: [10338554773] [ro.boottime.audioserver]: [10298157478] [ro.boottime.bootanim]: [9323670089] [ro.boottime.cameraserver]: [10299402321] [ro.boottime.cnd]: [10335931856] [ro.boottime.debuggerd]: [7001352774] [ro.boottime.debuggerd64]: [7002261785] [ro.boottime.drm]: [10301082113] [ro.boottime.fingerprintd]: [10331443314] [ro.boottime.flash-nanohub-fw]: [6995265534] [ro.boottime.gatekeeperd]: [10340355242] [ro.boottime.healthd]: [7856893380] [ro.boottime.hwservicemanager]: [7856051088] [ro.boottime.imscmservice]: [10290530758] [ro.boottime.imsdatadaemon]: [10358136702] [ro.boottime.imsqmidaemon]: [10289084872] [ro.boottime.installd]: [10303296020] [ro.boottime.irsc_util]: [10279807632] [ro.boottime.keystore]: [10305034093] [ro.boottime.lmkd]: [7863506714] [ro.boottime.loc_launcher]: [10324525241] [ro.boottime.logd]: [6526221633] [ro.boottime.logd-reinit]: [7850662702] [ro.boottime.mcfg-sh]: [10337268315] [ro.boottime.media]: [10312152687] [ro.boottime.mediacodec]: [10306852530] [ro.boottime.mediadrm]: [10308707999] [ro.boottime.mediaextractor]: [10310681177] [ro.boottime.msm_irqbalance]: [7862451974] [ro.boottime.netd]: [10313523104] [ro.boottime.netmgrd]: [10285009351] [ro.boottime.oem_qmi_server]: [10293329092] [ro.boottime.per_mgr]: [7857915776] [ro.boottime.per_proxy]: [8335121605] [ro.boottime.perfd]: [10283443101] [ro.boottime.qcamerasvr]: [10329644772] [ro.boottime.qmuxd]: [10282346643] [ro.boottime.qseecomd]: [6855708593] [ro.boottime.qti]: [10286196851] [ro.boottime.ril-daemon]: [10314933677] [ro.boottime.rmt_storage]: [7859105047] [ro.boottime.servicemanager]: [7864555881] [ro.boottime.ss_ramdump]: [8337634938] [ro.boottime.ssr_setup]: [8336268324] [ro.boottime.surfaceflinger]: [7866921402] [ro.boottime.thermal-engine]: [10281249924] [ro.boottime.time_daemon]: [10322006542] [ro.boottime.ueventd]: [5618663938] [ro.boottime.vold]: [7003493920] [ro.boottime.wificond]: [10316641073] [ro.boottime.wpa_supplicant]: [18959816881] [ro.boottime.zygote]: [10295295029] [ro.boottime.zygote_secondary]: [10296637269] Bug: http://b/31800756 Test: boots Change-Id: I094cce0c1bab9406d950ca94212689dc2e15dba5
2016-11-29 20:20:58 +01:00
LOG(ERROR) << "critical process '" << name_ << "' exited 4 times in 4 minutes";
panic();
}
} else {
time_crashed_ = now;
crash_count_ = 1;
}
}
flags_ &= (~SVC_RESTART);
flags_ |= SVC_RESTARTING;
// Execute all onrestart commands for this service.
onrestart_.ExecuteAllCommands();
NotifyStateChange("restarting");
return;
}
void Service::DumpState() const {
LOG(INFO) << "service " << name_;
LOG(INFO) << " class '" << android::base::Join(classnames_, " ") << "'";
LOG(INFO) << " exec "<< android::base::Join(args_, " ");
std::for_each(descriptors_.begin(), descriptors_.end(),
[] (const auto& info) { LOG(INFO) << *info; });
}
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
bool Service::ParseCapabilities(const std::vector<std::string>& args, std::string* err) {
capabilities_ = 0;
if (!CapAmbientSupported()) {
*err = "capabilities requested but the kernel does not support ambient capabilities";
return false;
}
unsigned int last_valid_cap = GetLastValidCap();
if (last_valid_cap >= capabilities_.size()) {
LOG(WARNING) << "last valid run-time capability is larger than CAP_LAST_CAP";
}
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
for (size_t i = 1; i < args.size(); i++) {
const std::string& arg = args[i];
int res = LookupCap(arg);
if (res < 0) {
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
*err = StringPrintf("invalid capability '%s'", arg.c_str());
return false;
}
unsigned int cap = static_cast<unsigned int>(res); // |res| is >= 0.
if (cap > last_valid_cap) {
*err = StringPrintf("capability '%s' not supported by the kernel", arg.c_str());
return false;
}
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
capabilities_[cap] = true;
}
return true;
}
bool Service::ParseClass(const std::vector<std::string>& args, std::string* err) {
classnames_ = std::set<std::string>(args.begin() + 1, args.end());
return true;
}
bool Service::ParseConsole(const std::vector<std::string>& args, std::string* err) {
flags_ |= SVC_CONSOLE;
console_ = args.size() > 1 ? "/dev/" + args[1] : "";
return true;
}
bool Service::ParseCritical(const std::vector<std::string>& args, std::string* err) {
flags_ |= SVC_CRITICAL;
return true;
}
bool Service::ParseDisabled(const std::vector<std::string>& args, std::string* err) {
flags_ |= SVC_DISABLED;
flags_ |= SVC_RC_DISABLED;
return true;
}
bool Service::ParseGroup(const std::vector<std::string>& args, std::string* err) {
gid_ = decode_uid(args[1].c_str());
for (std::size_t n = 2; n < args.size(); n++) {
supp_gids_.emplace_back(decode_uid(args[n].c_str()));
}
return true;
}
bool Service::ParsePriority(const std::vector<std::string>& args, std::string* err) {
priority_ = 0;
if (!ParseInt(args[1], &priority_,
static_cast<int>(ANDROID_PRIORITY_HIGHEST), // highest is negative
static_cast<int>(ANDROID_PRIORITY_LOWEST))) {
*err = StringPrintf("process priority value must be range %d - %d",
ANDROID_PRIORITY_HIGHEST, ANDROID_PRIORITY_LOWEST);
return false;
}
return true;
}
bool Service::ParseIoprio(const std::vector<std::string>& args, std::string* err) {
if (!ParseInt(args[2], &ioprio_pri_, 0, 7)) {
*err = "priority value must be range 0 - 7";
return false;
}
if (args[1] == "rt") {
ioprio_class_ = IoSchedClass_RT;
} else if (args[1] == "be") {
ioprio_class_ = IoSchedClass_BE;
} else if (args[1] == "idle") {
ioprio_class_ = IoSchedClass_IDLE;
} else {
*err = "ioprio option usage: ioprio <rt|be|idle> <0-7>";
return false;
}
return true;
}
bool Service::ParseKeycodes(const std::vector<std::string>& args, std::string* err) {
for (std::size_t i = 1; i < args.size(); i++) {
int code;
if (ParseInt(args[i], &code)) {
keycodes_.emplace_back(code);
} else {
LOG(WARNING) << "ignoring invalid keycode: " << args[i];
}
}
return true;
}
bool Service::ParseOneshot(const std::vector<std::string>& args, std::string* err) {
flags_ |= SVC_ONESHOT;
return true;
}
bool Service::ParseOnrestart(const std::vector<std::string>& args, std::string* err) {
std::vector<std::string> str_args(args.begin() + 1, args.end());
init: Stop combining actions In the past, I had thought it didn't make sense to have multiple Action classes with identical triggers within ActionManager::actions_, and opted to instead combine these into a single action. In theory, it should reduce memory overhead as only one copy of the triggers needs to be stored. In practice, this ends up not being a good idea. Most importantly, given a file with the below three sections in this same order: on boot setprop a b on boot && property:true=true setprop c d on boot setprop e f Assuming that property 'true' == 'true', when the `boot` event happens, the order of the setprop commands will actually be: setprop a b setprop e f setprop c d instead of the more intuitive order of: setprop a b setprop c d setprop e f This is a mistake and this CL fixes it. It also documents this order. Secondly, with a given 'Action' now spanning multiple files, in order to keep track of which file a command is run from, the 'Command' itself needs to store this. Ironically to the original intention, this increases total ram usage. This change now only stores the file name in each 'Action' instead of each 'Command'. All in all this is a negligible trade off of ram usage. Thirdly, this requires a bunch of extra code and assumptions that don't help anything else. In particular it forces to keep property triggers sorted for easy comparison, which I'm using an std::map for currently, but that is not the best data structure to contain them. Lastly, I added the filename and line number to the 'processing action' LOG(INFO) message. Test: Boot bullhead, observe above changes Test: Boot sailfish, observe no change in boot time Change-Id: I3fbcac4ee677351314e33012c758145be82346e9
2017-04-18 22:21:54 +02:00
int line = onrestart_.NumCommands() + 1;
onrestart_.AddCommand(str_args, line, err);
return true;
}
bool Service::ParseNamespace(const std::vector<std::string>& args, std::string* err) {
for (size_t i = 1; i < args.size(); i++) {
if (args[i] == "pid") {
namespace_flags_ |= CLONE_NEWPID;
// PID namespaces require mount namespaces.
namespace_flags_ |= CLONE_NEWNS;
} else if (args[i] == "mnt") {
namespace_flags_ |= CLONE_NEWNS;
} else {
*err = "namespace must be 'pid' or 'mnt'";
return false;
}
}
return true;
}
bool Service::ParseOomScoreAdjust(const std::vector<std::string>& args, std::string* err) {
if (!ParseInt(args[1], &oom_score_adjust_, -1000, 1000)) {
*err = "oom_score_adjust value must be in range -1000 - +1000";
return false;
}
return true;
}
bool Service::ParseSeclabel(const std::vector<std::string>& args, std::string* err) {
seclabel_ = args[1];
return true;
}
bool Service::ParseSetenv(const std::vector<std::string>& args, std::string* err) {
envvars_.emplace_back(args[1], args[2]);
return true;
}
template <typename T>
bool Service::AddDescriptor(const std::vector<std::string>& args, std::string* err) {
int perm = args.size() > 3 ? std::strtoul(args[3].c_str(), 0, 8) : -1;
uid_t uid = args.size() > 4 ? decode_uid(args[4].c_str()) : 0;
gid_t gid = args.size() > 5 ? decode_uid(args[5].c_str()) : 0;
std::string context = args.size() > 6 ? args[6] : "";
auto descriptor = std::make_unique<T>(args[1], args[2], uid, gid, perm, context);
auto old =
std::find_if(descriptors_.begin(), descriptors_.end(),
[&descriptor] (const auto& other) { return descriptor.get() == other.get(); });
if (old != descriptors_.end()) {
*err = "duplicate descriptor " + args[1] + " " + args[2];
return false;
}
descriptors_.emplace_back(std::move(descriptor));
return true;
}
// name type perm [ uid gid context ]
bool Service::ParseSocket(const std::vector<std::string>& args, std::string* err) {
if (args[2] != "dgram" && args[2] != "stream" && args[2] != "seqpacket") {
*err = "socket type must be 'dgram', 'stream' or 'seqpacket'";
return false;
}
return AddDescriptor<SocketInfo>(args, err);
}
// name type perm [ uid gid context ]
bool Service::ParseFile(const std::vector<std::string>& args, std::string* err) {
if (args[2] != "r" && args[2] != "w" && args[2] != "rw") {
*err = "file type must be 'r', 'w' or 'rw'";
return false;
}
if ((args[1][0] != '/') || (args[1].find("../") != std::string::npos)) {
*err = "file name must not be relative";
return false;
}
return AddDescriptor<FileInfo>(args, err);
}
bool Service::ParseUser(const std::vector<std::string>& args, std::string* err) {
uid_ = decode_uid(args[1].c_str());
return true;
}
bool Service::ParseWritepid(const std::vector<std::string>& args, std::string* err) {
writepid_files_.assign(args.begin() + 1, args.end());
return true;
}
class Service::OptionParserMap : public KeywordMap<OptionParser> {
public:
OptionParserMap() {
}
private:
Map& map() const override;
};
Service::OptionParserMap::Map& Service::OptionParserMap::map() const {
constexpr std::size_t kMax = std::numeric_limits<std::size_t>::max();
// clang-format off
static const Map option_parsers = {
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
{"capabilities",
{1, kMax, &Service::ParseCapabilities}},
{"class", {1, kMax, &Service::ParseClass}},
{"console", {0, 1, &Service::ParseConsole}},
{"critical", {0, 0, &Service::ParseCritical}},
{"disabled", {0, 0, &Service::ParseDisabled}},
{"group", {1, NR_SVC_SUPP_GIDS + 1, &Service::ParseGroup}},
{"ioprio", {2, 2, &Service::ParseIoprio}},
{"priority", {1, 1, &Service::ParsePriority}},
{"keycodes", {1, kMax, &Service::ParseKeycodes}},
{"oneshot", {0, 0, &Service::ParseOneshot}},
{"onrestart", {1, kMax, &Service::ParseOnrestart}},
{"oom_score_adjust",
{1, 1, &Service::ParseOomScoreAdjust}},
{"namespace", {1, 2, &Service::ParseNamespace}},
{"seclabel", {1, 1, &Service::ParseSeclabel}},
{"setenv", {2, 2, &Service::ParseSetenv}},
{"socket", {3, 6, &Service::ParseSocket}},
{"file", {2, 2, &Service::ParseFile}},
{"user", {1, 1, &Service::ParseUser}},
{"writepid", {1, kMax, &Service::ParseWritepid}},
};
// clang-format on
return option_parsers;
}
bool Service::ParseLine(const std::vector<std::string>& args, std::string* err) {
if (args.empty()) {
*err = "option needed, but not provided";
return false;
}
static const OptionParserMap parser_map;
auto parser = parser_map.FindFunction(args[0], args.size() - 1, err);
if (!parser) {
return false;
}
return (this->*parser)(args, err);
}
bool Service::ExecStart(std::unique_ptr<Timer>* exec_waiter) {
flags_ |= SVC_EXEC | SVC_ONESHOT;
exec_waiter->reset(new Timer);
if (!Start()) {
exec_waiter->reset();
return false;
}
return true;
}
bool Service::Start() {
// Starting a service removes it from the disabled or reset state and
// immediately takes it out of the restarting state if it was in there.
flags_ &= (~(SVC_DISABLED|SVC_RESTARTING|SVC_RESET|SVC_RESTART|SVC_DISABLED_START));
// Running processes require no additional work --- if they're in the
// process of exiting, we've ensured that they will immediately restart
// on exit, unless they are ONESHOT.
if (flags_ & SVC_RUNNING) {
return false;
}
bool needs_console = (flags_ & SVC_CONSOLE);
if (needs_console) {
if (console_.empty()) {
console_ = default_console;
}
// Make sure that open call succeeds to ensure a console driver is
// properly registered for the device node
int console_fd = open(console_.c_str(), O_RDWR | O_CLOEXEC);
if (console_fd < 0) {
PLOG(ERROR) << "service '" << name_ << "' couldn't open console '" << console_ << "'";
flags_ |= SVC_DISABLED;
return false;
}
close(console_fd);
}
struct stat sb;
if (stat(args_[0].c_str(), &sb) == -1) {
PLOG(ERROR) << "cannot find '" << args_[0] << "', disabling '" << name_ << "'";
flags_ |= SVC_DISABLED;
return false;
}
std::string scon;
if (!seclabel_.empty()) {
scon = seclabel_;
} else {
LOG(INFO) << "computing context for service '" << name_ << "'";
scon = ComputeContextFromExecutable(name_, args_[0]);
if (scon == "") {
return false;
}
}
LOG(INFO) << "starting service '" << name_ << "'...";
pid_t pid = -1;
if (namespace_flags_) {
pid = clone(nullptr, nullptr, namespace_flags_ | SIGCHLD, nullptr);
} else {
pid = fork();
}
if (pid == 0) {
umask(077);
if (namespace_flags_ & CLONE_NEWPID) {
// This will fork again to run an init process inside the PID
// namespace.
SetUpPidNamespace(name_);
}
for (const auto& ei : envvars_) {
add_environment(ei.name.c_str(), ei.value.c_str());
}
std::for_each(descriptors_.begin(), descriptors_.end(),
std::bind(&DescriptorInfo::CreateAndPublish, std::placeholders::_1, scon));
// See if there were "writepid" instructions to write to files under /dev/cpuset/.
auto cpuset_predicate = [](const std::string& path) {
return android::base::StartsWith(path, "/dev/cpuset/");
};
auto iter = std::find_if(writepid_files_.begin(), writepid_files_.end(), cpuset_predicate);
if (iter == writepid_files_.end()) {
// There were no "writepid" instructions for cpusets, check if the system default
// cpuset is specified to be used for the process.
std::string default_cpuset = android::base::GetProperty("ro.cpuset.default", "");
if (!default_cpuset.empty()) {
// Make sure the cpuset name starts and ends with '/'.
// A single '/' means the 'root' cpuset.
if (default_cpuset.front() != '/') {
default_cpuset.insert(0, 1, '/');
}
if (default_cpuset.back() != '/') {
default_cpuset.push_back('/');
}
writepid_files_.push_back(
StringPrintf("/dev/cpuset%stasks", default_cpuset.c_str()));
}
}
std::string pid_str = StringPrintf("%d", getpid());
for (const auto& file : writepid_files_) {
if (!WriteStringToFile(pid_str, file)) {
PLOG(ERROR) << "couldn't write " << pid_str << " to " << file;
}
}
if (ioprio_class_ != IoSchedClass_NONE) {
if (android_set_ioprio(getpid(), ioprio_class_, ioprio_pri_)) {
PLOG(ERROR) << "failed to set pid " << getpid()
<< " ioprio=" << ioprio_class_ << "," << ioprio_pri_;
}
}
if (needs_console) {
setsid();
OpenConsole();
} else {
ZapStdio();
}
// As requested, set our gid, supplemental gids, uid, context, and
// priority. Aborts on failure.
SetProcessAttributes();
std::vector<char*> strs;
ExpandArgs(args_, &strs);
if (execve(strs[0], (char**) &strs[0], (char**) ENV) < 0) {
PLOG(ERROR) << "cannot execve('" << strs[0] << "')";
}
_exit(127);
}
if (pid < 0) {
PLOG(ERROR) << "failed to fork for '" << name_ << "'";
pid_ = 0;
return false;
}
if (oom_score_adjust_ != -1000) {
std::string oom_str = StringPrintf("%d", oom_score_adjust_);
std::string oom_file = StringPrintf("/proc/%d/oom_score_adj", pid);
if (!WriteStringToFile(oom_str, oom_file)) {
PLOG(ERROR) << "couldn't write oom_score_adj: " << strerror(errno);
}
}
time_started_ = boot_clock::now();
pid_ = pid;
flags_ |= SVC_RUNNING;
errno = -createProcessGroup(uid_, pid_);
if (errno != 0) {
PLOG(ERROR) << "createProcessGroup(" << uid_ << ", " << pid_ << ") failed for service '"
<< name_ << "'";
}
if ((flags_ & SVC_EXEC) != 0) {
LOG(INFO) << android::base::StringPrintf(
"SVC_EXEC pid %d (uid %d gid %d+%zu context %s) started; waiting...", pid_, uid_, gid_,
supp_gids_.size(), !seclabel_.empty() ? seclabel_.c_str() : "default");
}
NotifyStateChange("running");
return true;
}
bool Service::StartIfNotDisabled() {
if (!(flags_ & SVC_DISABLED)) {
return Start();
} else {
flags_ |= SVC_DISABLED_START;
}
return true;
}
bool Service::Enable() {
flags_ &= ~(SVC_DISABLED | SVC_RC_DISABLED);
if (flags_ & SVC_DISABLED_START) {
return Start();
}
return true;
}
void Service::Reset() {
StopOrReset(SVC_RESET);
}
void Service::Stop() {
StopOrReset(SVC_DISABLED);
}
void Service::Terminate() {
flags_ &= ~(SVC_RESTARTING | SVC_DISABLED_START);
flags_ |= SVC_DISABLED;
if (pid_) {
KillProcessGroup(SIGTERM);
NotifyStateChange("stopping");
}
}
void Service::Restart() {
if (flags_ & SVC_RUNNING) {
/* Stop, wait, then start the service. */
StopOrReset(SVC_RESTART);
} else if (!(flags_ & SVC_RESTARTING)) {
/* Just start the service since it's not running. */
Start();
} /* else: Service is restarting anyways. */
}
void Service::RestartIfNeeded(time_t* process_needs_restart_at) {
boot_clock::time_point now = boot_clock::now();
boot_clock::time_point next_start = time_started_ + 5s;
if (now > next_start) {
flags_ &= (~SVC_RESTARTING);
Start();
return;
}
time_t next_start_time_t = time(nullptr) +
time_t(std::chrono::duration_cast<std::chrono::seconds>(next_start - now).count());
if (next_start_time_t < *process_needs_restart_at || *process_needs_restart_at == 0) {
*process_needs_restart_at = next_start_time_t;
}
}
// The how field should be either SVC_DISABLED, SVC_RESET, or SVC_RESTART.
void Service::StopOrReset(int how) {
// The service is still SVC_RUNNING until its process exits, but if it has
// already exited it shoudn't attempt a restart yet.
flags_ &= ~(SVC_RESTARTING | SVC_DISABLED_START);
if ((how != SVC_DISABLED) && (how != SVC_RESET) && (how != SVC_RESTART)) {
// An illegal flag: default to SVC_DISABLED.
how = SVC_DISABLED;
}
// If the service has not yet started, prevent it from auto-starting with its class.
if (how == SVC_RESET) {
flags_ |= (flags_ & SVC_RC_DISABLED) ? SVC_DISABLED : SVC_RESET;
} else {
flags_ |= how;
}
if (pid_) {
KillProcessGroup(SIGKILL);
NotifyStateChange("stopping");
} else {
NotifyStateChange("stopped");
}
}
void Service::ZapStdio() const {
int fd;
fd = open("/dev/null", O_RDWR);
dup2(fd, 0);
dup2(fd, 1);
dup2(fd, 2);
close(fd);
}
void Service::OpenConsole() const {
int fd = open(console_.c_str(), O_RDWR);
if (fd == -1) fd = open("/dev/null", O_RDWR);
ioctl(fd, TIOCSCTTY, 0);
dup2(fd, 0);
dup2(fd, 1);
dup2(fd, 2);
close(fd);
}
int ServiceManager::exec_count_ = 0;
ServiceManager::ServiceManager() {
}
ServiceManager& ServiceManager::GetInstance() {
static ServiceManager instance;
return instance;
}
void ServiceManager::AddService(std::unique_ptr<Service> service) {
Service* old_service = FindServiceByName(service->name());
if (old_service) {
LOG(ERROR) << "ignored duplicate definition of service '" << service->name() << "'";
return;
}
services_.emplace_back(std::move(service));
}
bool ServiceManager::Exec(const std::vector<std::string>& args) {
Service* svc = MakeExecOneshotService(args);
if (!svc) {
LOG(ERROR) << "Could not create exec service";
return false;
}
if (!svc->ExecStart(&exec_waiter_)) {
LOG(ERROR) << "Could not start exec service";
ServiceManager::GetInstance().RemoveService(*svc);
return false;
}
return true;
}
bool ServiceManager::ExecStart(const std::string& name) {
Service* svc = FindServiceByName(name);
if (!svc) {
LOG(ERROR) << "ExecStart(" << name << "): Service not found";
return false;
}
if (!svc->ExecStart(&exec_waiter_)) {
LOG(ERROR) << "ExecStart(" << name << "): Could not start Service";
return false;
}
return true;
}
bool ServiceManager::IsWaitingForExec() const { return exec_waiter_ != nullptr; }
Service* ServiceManager::MakeExecOneshotService(const std::vector<std::string>& args) {
// Parse the arguments: exec [SECLABEL [UID [GID]*] --] COMMAND ARGS...
// SECLABEL can be a - to denote default
std::size_t command_arg = 1;
for (std::size_t i = 1; i < args.size(); ++i) {
if (args[i] == "--") {
command_arg = i + 1;
break;
}
}
if (command_arg > 4 + NR_SVC_SUPP_GIDS) {
LOG(ERROR) << "exec called with too many supplementary group ids";
return nullptr;
}
if (command_arg >= args.size()) {
LOG(ERROR) << "exec called without command";
return nullptr;
}
std::vector<std::string> str_args(args.begin() + command_arg, args.end());
exec_count_++;
std::string name = StringPrintf("exec %d (%s)", exec_count_, str_args[0].c_str());
unsigned flags = SVC_EXEC | SVC_ONESHOT | SVC_TEMPORARY;
init: Add support for ambient capabilities. Ambient capabilities are inherited in a straightforward way across execve(2): " If you are nonroot but you have a capability, you can add it to pA. If you do so, your children get that capability in pA, pP, and pE. For example, you can set pA = CAP_NET_BIND_SERVICE, and your children can automatically bind low-numbered ports. " This will allow us to get rid of the special meaning for AID_NET_ADMIN and AID_NET_RAW, and if desired, to reduce the use of file capabilities (which grant capabilities to any process that can execute the file). An additional benefit of the latter is that a single .rc file can specify all properties for a service, without having to rely on a separate file for file capabilities. Ambient capabilities are supported starting with kernel 4.3 and have been backported to all Android common kernels back to 3.10. I chose to not use Minijail here (though I'm still using libcap) for two reasons: 1-The Minijail code is designed to work in situations where the process is holding any set of capabilities, so it's more complex. The situation when forking from init allows for simpler code. 2-The way Minijail is structured right now, we would not be able to make the required SELinux calls between UID/GID dropping and other priv dropping code. In the future, it will make sense to add some sort of "hook" to Minijail so that it can be used in situations where we want to do other operations between some of the privilege-dropping operations carried out by Minijail. Bug: 32438163 Test: Use sample service. Change-Id: I3226cc95769d1beacbae619cb6c6e6a5425890fb
2016-10-27 16:33:03 +02:00
CapSet no_capabilities;
unsigned namespace_flags = 0;
std::string seclabel = "";
if (command_arg > 2 && args[1] != "-") {
seclabel = args[1];
}
uid_t uid = 0;
if (command_arg > 3) {
uid = decode_uid(args[2].c_str());
}
gid_t gid = 0;
std::vector<gid_t> supp_gids;
if (command_arg > 4) {
gid = decode_uid(args[3].c_str());
std::size_t nr_supp_gids = command_arg - 1 /* -- */ - 4 /* exec SECLABEL UID GID */;
for (size_t i = 0; i < nr_supp_gids; ++i) {
supp_gids.push_back(decode_uid(args[4 + i].c_str()));
}
}
auto svc_p = std::make_unique<Service>(name, flags, uid, gid, supp_gids, no_capabilities,
namespace_flags, seclabel, str_args);
Service* svc = svc_p.get();
services_.emplace_back(std::move(svc_p));
return svc;
}
Service* ServiceManager::FindServiceByName(const std::string& name) const {
auto svc = std::find_if(services_.begin(), services_.end(),
[&name] (const std::unique_ptr<Service>& s) {
return name == s->name();
});
if (svc != services_.end()) {
return svc->get();
}
return nullptr;
}
Service* ServiceManager::FindServiceByPid(pid_t pid) const {
auto svc = std::find_if(services_.begin(), services_.end(),
[&pid] (const std::unique_ptr<Service>& s) {
return s->pid() == pid;
});
if (svc != services_.end()) {
return svc->get();
}
return nullptr;
}
Service* ServiceManager::FindServiceByKeychord(int keychord_id) const {
auto svc = std::find_if(services_.begin(), services_.end(),
[&keychord_id] (const std::unique_ptr<Service>& s) {
return s->keychord_id() == keychord_id;
});
if (svc != services_.end()) {
return svc->get();
}
return nullptr;
}
void ServiceManager::ForEachService(const std::function<void(Service*)>& callback) const {
for (const auto& s : services_) {
callback(s.get());
}
}
void ServiceManager::ForEachServiceInClass(const std::string& classname,
void (*func)(Service* svc)) const {
for (const auto& s : services_) {
if (s->classnames().find(classname) != s->classnames().end()) {
func(s.get());
}
}
}
void ServiceManager::ForEachServiceWithFlags(unsigned matchflags,
void (*func)(Service* svc)) const {
for (const auto& s : services_) {
if (s->flags() & matchflags) {
func(s.get());
}
}
}
void ServiceManager::RemoveService(const Service& svc) {
auto svc_it = std::find_if(services_.begin(), services_.end(),
[&svc] (const std::unique_ptr<Service>& s) {
return svc.name() == s->name();
});
if (svc_it == services_.end()) {
return;
}
services_.erase(svc_it);
}
void ServiceManager::DumpState() const {
for (const auto& s : services_) {
s->DumpState();
}
}
bool ServiceManager::ReapOneProcess() {
int status;
pid_t pid = TEMP_FAILURE_RETRY(waitpid(-1, &status, WNOHANG));
if (pid == 0) {
return false;
} else if (pid == -1) {
PLOG(ERROR) << "waitpid failed";
return false;
}
Service* svc = FindServiceByPid(pid);
std::string name;
if (svc) {
name = android::base::StringPrintf("Service '%s' (pid %d)",
svc->name().c_str(), pid);
} else {
name = android::base::StringPrintf("Untracked pid %d", pid);
}
if (WIFEXITED(status)) {
LOG(INFO) << name << " exited with status " << WEXITSTATUS(status);
} else if (WIFSIGNALED(status)) {
LOG(INFO) << name << " killed by signal " << WTERMSIG(status);
} else if (WIFSTOPPED(status)) {
LOG(INFO) << name << " stopped by signal " << WSTOPSIG(status);
} else {
LOG(INFO) << name << " state changed";
}
if (!svc) {
return true;
}
svc->Reap();
if (svc->flags() & SVC_EXEC) {
LOG(INFO) << "Wait for exec took " << *exec_waiter_;
exec_waiter_.reset();
}
if (svc->flags() & SVC_TEMPORARY) {
RemoveService(*svc);
}
return true;
}
void ServiceManager::ReapAnyOutstandingChildren() {
while (ReapOneProcess()) {
}
}
init: Stop combining actions In the past, I had thought it didn't make sense to have multiple Action classes with identical triggers within ActionManager::actions_, and opted to instead combine these into a single action. In theory, it should reduce memory overhead as only one copy of the triggers needs to be stored. In practice, this ends up not being a good idea. Most importantly, given a file with the below three sections in this same order: on boot setprop a b on boot && property:true=true setprop c d on boot setprop e f Assuming that property 'true' == 'true', when the `boot` event happens, the order of the setprop commands will actually be: setprop a b setprop e f setprop c d instead of the more intuitive order of: setprop a b setprop c d setprop e f This is a mistake and this CL fixes it. It also documents this order. Secondly, with a given 'Action' now spanning multiple files, in order to keep track of which file a command is run from, the 'Command' itself needs to store this. Ironically to the original intention, this increases total ram usage. This change now only stores the file name in each 'Action' instead of each 'Command'. All in all this is a negligible trade off of ram usage. Thirdly, this requires a bunch of extra code and assumptions that don't help anything else. In particular it forces to keep property triggers sorted for easy comparison, which I'm using an std::map for currently, but that is not the best data structure to contain them. Lastly, I added the filename and line number to the 'processing action' LOG(INFO) message. Test: Boot bullhead, observe above changes Test: Boot sailfish, observe no change in boot time Change-Id: I3fbcac4ee677351314e33012c758145be82346e9
2017-04-18 22:21:54 +02:00
bool ServiceParser::ParseSection(const std::vector<std::string>& args, const std::string& filename,
int line, std::string* err) {
if (args.size() < 3) {
*err = "services must have a name and a program";
return false;
}
const std::string& name = args[1];
if (!IsValidName(name)) {
*err = StringPrintf("invalid service name '%s'", name.c_str());
return false;
}
std::vector<std::string> str_args(args.begin() + 2, args.end());
service_ = std::make_unique<Service>(name, str_args);
return true;
}
init: Stop combining actions In the past, I had thought it didn't make sense to have multiple Action classes with identical triggers within ActionManager::actions_, and opted to instead combine these into a single action. In theory, it should reduce memory overhead as only one copy of the triggers needs to be stored. In practice, this ends up not being a good idea. Most importantly, given a file with the below three sections in this same order: on boot setprop a b on boot && property:true=true setprop c d on boot setprop e f Assuming that property 'true' == 'true', when the `boot` event happens, the order of the setprop commands will actually be: setprop a b setprop e f setprop c d instead of the more intuitive order of: setprop a b setprop c d setprop e f This is a mistake and this CL fixes it. It also documents this order. Secondly, with a given 'Action' now spanning multiple files, in order to keep track of which file a command is run from, the 'Command' itself needs to store this. Ironically to the original intention, this increases total ram usage. This change now only stores the file name in each 'Action' instead of each 'Command'. All in all this is a negligible trade off of ram usage. Thirdly, this requires a bunch of extra code and assumptions that don't help anything else. In particular it forces to keep property triggers sorted for easy comparison, which I'm using an std::map for currently, but that is not the best data structure to contain them. Lastly, I added the filename and line number to the 'processing action' LOG(INFO) message. Test: Boot bullhead, observe above changes Test: Boot sailfish, observe no change in boot time Change-Id: I3fbcac4ee677351314e33012c758145be82346e9
2017-04-18 22:21:54 +02:00
bool ServiceParser::ParseLineSection(const std::vector<std::string>& args, int line,
std::string* err) {
return service_ ? service_->ParseLine(args, err) : false;
}
void ServiceParser::EndSection() {
if (service_) {
ServiceManager::GetInstance().AddService(std::move(service_));
}
}
bool ServiceParser::IsValidName(const std::string& name) const {
// Property names can be any length, but may only contain certain characters.
// Property values can contain any characters, but may only be a certain length.
// (The latter restriction is needed because `start` and `stop` work by writing
// the service name to the "ctl.start" and "ctl.stop" properties.)
return is_legal_property_name("init.svc." + name) && name.size() <= PROP_VALUE_MAX;
}