Merge "Use properties for various userspace reboot timeouts"

This commit is contained in:
Nikita Ioffe 2020-03-25 20:02:05 +00:00 committed by Gerrit Code Review
commit fc51af215b
2 changed files with 26 additions and 10 deletions

View file

@ -96,6 +96,7 @@
using android::base::Basename;
using android::base::GetBoolProperty;
using android::base::GetUintProperty;
using android::base::Readlink;
using android::base::Realpath;
using android::base::SetProperty;
@ -1545,11 +1546,16 @@ int fs_mgr_umount_all(android::fs_mgr::Fstab* fstab) {
return ret;
}
static std::chrono::milliseconds GetMillisProperty(const std::string& name,
std::chrono::milliseconds default_value) {
auto value = GetUintProperty(name, static_cast<uint64_t>(default_value.count()));
return std::chrono::milliseconds(std::move(value));
}
static bool fs_mgr_unmount_all_data_mounts(const std::string& block_device) {
LINFO << __FUNCTION__ << "(): about to umount everything on top of " << block_device;
Timer t;
// TODO(b/135984674): should be configured via a read-only property.
std::chrono::milliseconds timeout = 5s;
auto timeout = GetMillisProperty("init.userspace_reboot.userdata_remount.timeoutmillis", 5s);
while (true) {
bool umount_done = true;
Fstab proc_mounts;

View file

@ -72,6 +72,7 @@ using namespace std::literals;
using android::base::boot_clock;
using android::base::GetBoolProperty;
using android::base::GetUintProperty;
using android::base::SetProperty;
using android::base::Split;
using android::base::Timer;
@ -732,6 +733,12 @@ static Result<void> UnmountAllApexes() {
return Error() << "'/system/bin/apexd --unmount-all' failed : " << status;
}
static std::chrono::milliseconds GetMillisProperty(const std::string& name,
std::chrono::milliseconds default_value) {
auto value = GetUintProperty(name, static_cast<uint64_t>(default_value.count()));
return std::chrono::milliseconds(std::move(value));
}
static Result<void> DoUserspaceReboot() {
LOG(INFO) << "Userspace reboot initiated";
auto guard = android::base::make_scope_guard([] {
@ -769,10 +776,13 @@ static Result<void> DoUserspaceReboot() {
sync();
LOG(INFO) << "sync() took " << sync_timer;
}
// TODO(b/135984674): do we need shutdown animation for userspace reboot?
// TODO(b/135984674): control userspace timeout via read-only property?
StopServicesAndLogViolations(stop_first, 10s, true /* SIGTERM */);
if (int r = StopServicesAndLogViolations(stop_first, 20s, false /* SIGKILL */); r > 0) {
auto sigterm_timeout = GetMillisProperty("init.userspace_reboot.sigterm.timeoutmillis", 5s);
auto sigkill_timeout = GetMillisProperty("init.userspace_reboot.sigkill.timeoutmillis", 10s);
LOG(INFO) << "Timeout to terminate services : " << sigterm_timeout.count() << "ms"
<< "Timeout to kill services: " << sigkill_timeout.count() << "ms";
StopServicesAndLogViolations(stop_first, sigterm_timeout, true /* SIGTERM */);
if (int r = StopServicesAndLogViolations(stop_first, sigkill_timeout, false /* SIGKILL */);
r > 0) {
// TODO(b/135984674): store information about offending services for debugging.
return Error() << r << " post-data services are still running";
}
@ -782,8 +792,8 @@ static Result<void> DoUserspaceReboot() {
if (auto result = CallVdc("volume", "reset"); !result.ok()) {
return result;
}
if (int r = StopServicesAndLogViolations(GetDebuggingServices(true /* only_post_data */), 5s,
false /* SIGKILL */);
if (int r = StopServicesAndLogViolations(GetDebuggingServices(true /* only_post_data */),
sigkill_timeout, false /* SIGKILL */);
r > 0) {
// TODO(b/135984674): store information about offending services for debugging.
return Error() << r << " debugging services are still running";
@ -827,8 +837,8 @@ static void UserspaceRebootWatchdogThread() {
return;
}
LOG(INFO) << "Starting userspace reboot watchdog";
// TODO(b/135984674): this should be configured via a read-only sysprop.
std::chrono::milliseconds timeout = 60s;
auto timeout = GetMillisProperty("init.userspace_reboot.watchdog.timeoutmillis", 5min);
LOG(INFO) << "UserspaceRebootWatchdog timeout: " << timeout.count() << "ms";
if (!WaitForProperty("sys.boot_completed", "1", timeout)) {
LOG(ERROR) << "Failed to boot in " << timeout.count() << "ms. Switching to full reboot";
// In this case device is in a boot loop. Only way to recover is to do dirty reboot.