diff options
Diffstat (limited to 'meta-openbmc-mods/meta-common/recipes-phosphor/watchdog')
6 files changed, 397 insertions, 0 deletions
diff --git a/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog/0001-Customize-phosphor-watchdog-for-Intel-platforms.patch b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog/0001-Customize-phosphor-watchdog-for-Intel-platforms.patch new file mode 100644 index 000000000..48bc4a086 --- /dev/null +++ b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog/0001-Customize-phosphor-watchdog-for-Intel-platforms.patch @@ -0,0 +1,348 @@ +From a7d1d3770a650df8fe61d594885fcb388ac2ca42 Mon Sep 17 00:00:00 2001 +From: James Feist <james.feist@linux.intel.com> +Date: Mon, 17 Jun 2019 12:00:58 -0700 +Subject: [PATCH] Customize phosphor-watchdog for Intel platforms + +This patch adds various changes to phosphor-watchdog that are +required for compatibility with Intel platforms. + + 1. Add Redfish messages for watchdog timeout and pre-interrupt + 2. Use dbus properties for power control insted of service files + 3. Use host status to enable/disable watchdog + 4. Set preTimeoutInterruptOccurFlag + 5. Assign watchdog cause for correct reset cause reporting + 6. Add NMI Pre-Interrupt support for IPMI watchdog timer. + +Signed-off-by: James Feist <james.feist@linux.intel.com> +Signed-off-by: Ren Yu <yux.ren@intel.com> +Signed-off-by: Yong Li <yong.b.li@linux.intel.com> +Signed-off-by: Jason M. Bills <jason.m.bills@linux.intel.com> +Signed-off-by: Johnathan Mantey <johnathanx.mantey@intel.com> +Signed-off-by: Sunita Kumari <sunitax.kumari@intel.com> + +%% original patch: 0001-Customize-phosphor-watchdog-for-Intel-platforms.patch +--- + src/watchdog.cpp | 229 ++++++++++++++++++++++++++++++++++++++++++++--- + src/watchdog.hpp | 22 ++++- + 2 files changed, 240 insertions(+), 11 deletions(-) + +diff --git a/src/watchdog.cpp b/src/watchdog.cpp +index 7de98ae3e70f..f96faebc7368 100644 +--- a/src/watchdog.cpp ++++ b/src/watchdog.cpp +@@ -5,8 +5,10 @@ + #include <phosphor-logging/elog.hpp> + #include <phosphor-logging/log.hpp> + #include <sdbusplus/exception.hpp> ++#include <systemd/sd-journal.h> + #include <string_view> + #include <xyz/openbmc_project/Common/error.hpp> ++#include <xyz/openbmc_project/State/Host/server.hpp> + + namespace phosphor + { +@@ -19,10 +21,86 @@ using namespace phosphor::logging; + using sdbusplus::exception::SdBusError; + using sdbusplus::xyz::openbmc_project::Common::Error::InternalFailure; + +-// systemd service to kick start a target. +-constexpr auto SYSTEMD_SERVICE = "org.freedesktop.systemd1"; +-constexpr auto SYSTEMD_ROOT = "/org/freedesktop/systemd1"; +-constexpr auto SYSTEMD_INTERFACE = "org.freedesktop.systemd1.Manager"; ++const static constexpr char* currentHostState = "CurrentHostState"; ++const static constexpr char* hostStatusOff = ++ "xyz.openbmc_project.State.Host.HostState.Off"; ++ ++const static constexpr char* actionDescription = " due to Watchdog timeout"; ++const static constexpr char* hardResetDescription = "Hard Reset - System reset"; ++const static constexpr char* powerOffDescription = ++ "Power Down - System power down"; ++const static constexpr char* powerCycleDescription = ++ "Power Cycle - System power cycle"; ++const static constexpr char* timerExpiredDescription = "Timer expired"; ++ ++const static constexpr char* preInterruptActionNone = ++ "xyz.openbmc_project.State.Watchdog.PreTimeoutInterruptAction.None"; ++ ++const static constexpr char* preInterruptDescriptionSMI = "SMI"; ++const static constexpr char* preInterruptDescriptionNMI = "NMI"; ++const static constexpr char* preInterruptDescriptionMI = "Messaging Interrupt"; ++ ++const static constexpr char* reservedDescription = "Reserved"; ++ ++const static constexpr char* timerUseDescriptionBIOSFRB2 = "BIOS FRB2"; ++const static constexpr char* timerUseDescriptionBIOSPOST = "BIOS/POST"; ++const static constexpr char* timerUseDescriptionOSLoad = "OSLoad"; ++const static constexpr char* timerUseDescriptionSMSOS = "SMS/OS"; ++const static constexpr char* timerUseDescriptionOEM = "OEM"; ++ ++namespace restart ++{ ++static constexpr const char* busName = ++ "xyz.openbmc_project.Control.Host.RestartCause"; ++static constexpr const char* path = ++ "/xyz/openbmc_project/control/host0/restart_cause"; ++static constexpr const char* interface = ++ "xyz.openbmc_project.Control.Host.RestartCause"; ++static constexpr const char* property = "RequestedRestartCause"; ++} // namespace restart ++ ++// chassis state manager service ++namespace chassis ++{ ++static constexpr const char* busName = "xyz.openbmc_project.State.Chassis"; ++static constexpr const char* path = "/xyz/openbmc_project/state/chassis0"; ++static constexpr const char* interface = "xyz.openbmc_project.State.Chassis"; ++static constexpr const char* request = "RequestedPowerTransition"; ++} // namespace chassis ++ ++namespace host ++{ ++static constexpr const char* busName = "xyz.openbmc_project.State.Host"; ++static constexpr const char* path = "/xyz/openbmc_project/state/host0"; ++static constexpr const char* interface = "xyz.openbmc_project.State.Host"; ++static constexpr const char* request = "RequestedHostTransition"; ++} // namespace host ++ ++namespace nmi ++{ ++static constexpr const char* busName = "xyz.openbmc_project.Control.Host.NMI"; ++static constexpr const char* path = "/xyz/openbmc_project/control/host0/nmi"; ++static constexpr const char* interface = "xyz.openbmc_project.Control.Host.NMI"; ++static constexpr const char* request = "NMI"; ++ ++} // namespace nmi ++ ++void Watchdog::powerStateChangedHandler( ++ const std::map<std::string, std::variant<std::string>>& props) ++{ ++ const auto iter = props.find(currentHostState); ++ if (iter != props.end()) ++ { ++ const std::string* powerState = std::get_if<std::string>(&iter->second); ++ if (powerState && (*powerState == hostStatusOff)) ++ { ++ if (timerEnabled()) ++ { ++ enabled(false); ++ } ++ } ++ } ++} + + void Watchdog::resetTimeRemaining(bool enableWatchdog) + { +@@ -108,13 +186,111 @@ uint64_t Watchdog::interval(uint64_t value) + // Optional callback function on timer expiration + void Watchdog::timeOutHandler() + { ++ PreTimeoutInterruptAction preTimeoutInterruptAction = preTimeoutInterrupt(); ++ std::string preInterruptActionMessageArgs{}; ++ + Action action = expireAction(); ++ std::string actionMessageArgs{}; ++ ++ expiredTimerUse(currentTimerUse()); ++ ++ TimerUse timeUser = expiredTimerUse(); ++ std::string timeUserMessage{}; ++ + if (!this->enabled()) + { + action = fallback->action; + } + +- expiredTimerUse(currentTimerUse()); ++ switch (timeUser) ++ { ++ case Watchdog::TimerUse::BIOSFRB2: ++ timeUserMessage = timerUseDescriptionBIOSFRB2; ++ break; ++ case Watchdog::TimerUse::BIOSPOST: ++ timeUserMessage = timerUseDescriptionBIOSPOST; ++ break; ++ case Watchdog::TimerUse::OSLoad: ++ timeUserMessage = timerUseDescriptionOSLoad; ++ break; ++ case Watchdog::TimerUse::SMSOS: ++ timeUserMessage = timerUseDescriptionSMSOS; ++ break; ++ case Watchdog::TimerUse::OEM: ++ timeUserMessage = timerUseDescriptionOEM; ++ break; ++ default: ++ timeUserMessage = reservedDescription; ++ break; ++ } ++ ++ switch (action) ++ { ++ case Watchdog::Action::HardReset: ++ actionMessageArgs = std::string(hardResetDescription) + ++ std::string(actionDescription); ++ break; ++ case Watchdog::Action::PowerOff: ++ actionMessageArgs = std::string(powerOffDescription) + ++ std::string(actionDescription); ++ break; ++ case Watchdog::Action::PowerCycle: ++ actionMessageArgs = std::string(powerCycleDescription) + ++ std::string(actionDescription); ++ break; ++ case Watchdog::Action::None: ++ actionMessageArgs = timerExpiredDescription; ++ break; ++ default: ++ actionMessageArgs = reservedDescription; ++ break; ++ } ++ ++ // Log into redfish event log ++ sd_journal_send("MESSAGE=IPMIWatchdog: Timed out ACTION=%s", ++ convertForMessage(action).c_str(), "PRIORITY=%i", LOG_INFO, ++ "REDFISH_MESSAGE_ID=%s", "OpenBMC.0.1.IPMIWatchdog", ++ "REDFISH_MESSAGE_ARGS=%s. timer use: %s", ++ actionMessageArgs.c_str(), timeUserMessage.c_str(), NULL); ++ ++ switch (preTimeoutInterruptAction) ++ { ++ case Watchdog::PreTimeoutInterruptAction::SMI: ++ preInterruptActionMessageArgs = preInterruptDescriptionSMI; ++ break; ++ case Watchdog::PreTimeoutInterruptAction::NMI: ++ preInterruptActionMessageArgs = preInterruptDescriptionNMI; ++ break; ++ case Watchdog::PreTimeoutInterruptAction::MI: ++ preInterruptActionMessageArgs = preInterruptDescriptionMI; ++ break; ++ default: ++ preInterruptActionMessageArgs = reservedDescription; ++ break; ++ } ++ ++ if (preInterruptActionNone != convertForMessage(preTimeoutInterruptAction)) ++ { ++ preTimeoutInterruptOccurFlag(true); ++ ++ sd_journal_send("MESSAGE=IPMIWatchdog: Pre Timed out Interrupt=%s", ++ convertForMessage(preTimeoutInterruptAction).c_str(), ++ "PRIORITY=%i", LOG_INFO, "REDFISH_MESSAGE_ID=%s", ++ "OpenBMC.0.1.IPMIWatchdog", ++ "REDFISH_MESSAGE_ARGS=Timer interrupt - %s due to " ++ "Watchdog timeout. timer use: %s", ++ preInterruptActionMessageArgs.c_str(), ++ timeUserMessage.c_str(), NULL); ++ ++ if (preTimeoutInterruptAction == ++ Watchdog::PreTimeoutInterruptAction::NMI) ++ { ++ sdbusplus::message::message preTimeoutInterruptHandler; ++ preTimeoutInterruptHandler = bus.new_method_call( ++ nmi::busName, nmi::path, nmi::interface, nmi::request); ++ bus.call_noreply(preTimeoutInterruptHandler); ++ } ++ } + + auto target = actionTargetMap.find(action); + if (target == actionTargetMap.end()) +@@ -147,12 +323,45 @@ void Watchdog::timeOutHandler() + + try + { +- auto method = bus.new_method_call(SYSTEMD_SERVICE, SYSTEMD_ROOT, +- SYSTEMD_INTERFACE, "StartUnit"); +- method.append(target->second); +- method.append("replace"); ++ sdbusplus::message::message method; ++ if (action == Watchdog::Action::HardReset) ++ { ++ auto method = bus.new_method_call( ++ restart::busName, restart::path, ++ "org.freedesktop.DBus.Properties", "Set"); ++ method.append( ++ restart::interface, restart::property, ++ std::variant<std::string>("xyz.openbmc_project.State.Host." ++ "RestartCause.WatchdogTimer")); ++ bus.call_noreply(method); + +- bus.call_noreply(method); ++ method = bus.new_method_call(host::busName, host::path, ++ "org.freedesktop.DBus.Properties", ++ "Set"); ++ method.append(host::interface, host::request, ++ std::variant<std::string>(target->second)); ++ bus.call_noreply(method); ++ } ++ else ++ { ++ if (action == Watchdog::Action::PowerCycle) ++ { ++ auto method = bus.new_method_call( ++ restart::busName, restart::path, ++ "org.freedesktop.DBus.Properties", "Set"); ++ method.append(restart::interface, restart::property, ++ std::variant<std::string>( ++ "xyz.openbmc_project.State.Host." ++ "RestartCause.WatchdogTimer")); ++ bus.call_noreply(method); ++ } ++ method = bus.new_method_call(chassis::busName, chassis::path, ++ "org.freedesktop.DBus.Properties", ++ "Set"); ++ method.append(chassis::interface, chassis::request, ++ std::variant<std::string>(target->second)); ++ bus.call_noreply(method); ++ } + } + catch (const SdBusError& e) + { +diff --git a/src/watchdog.hpp b/src/watchdog.hpp +index 736e71f68fcc..79158f192f40 100644 +--- a/src/watchdog.hpp ++++ b/src/watchdog.hpp +@@ -73,7 +73,17 @@ class Watchdog : public WatchdogInherits + bus(bus), actionTargetMap(std::move(actionTargetMap)), + fallback(std::move(fallback)), minInterval(minInterval), + timer(event, std::bind(&Watchdog::timeOutHandler, this)), +- objPath(objPath) ++ objPath(objPath), powerStateChangedSignal( ++ bus, ++ sdbusplus::bus::match::rules::propertiesChanged( ++ "/xyz/openbmc_project/state/host0", ++ "xyz.openbmc_project.State.Host"), ++ [this](sdbusplus::message::message& msg) { ++ std::string objectName; ++ std::map<std::string, std::variant<std::string>> props; ++ msg.read(objectName, props); ++ powerStateChangedHandler(props); ++ }) + { + // Use default if passed in otherwise just use default that comes + // with object +@@ -90,6 +100,12 @@ class Watchdog : public WatchdogInherits + tryFallbackOrDisable(); + } + ++ /** @brief Disable watchdog when power status change meet ++ * the specific requirement ++ */ ++ void powerStateChangedHandler( ++ const std::map<std::string, std::variant<std::string>>& props); ++ + /** @brief Resets the TimeRemaining to the configured Interval + * Optionally enables the watchdog. + * +@@ -178,6 +194,10 @@ class Watchdog : public WatchdogInherits + /** @brief Contained timer object */ + sdeventplus::utility::Timer<sdeventplus::ClockId::Monotonic> timer; + ++ /** @brief Optional Callback handler when power status change meet ++ * the specific requirement */ ++ sdbusplus::bus::match_t powerStateChangedSignal; ++ + /** @brief Optional Callback handler on timer expirartion */ + void timeOutHandler(); + +-- +2.17.1 + diff --git a/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog/phosphor-watchdog.service b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog/phosphor-watchdog.service new file mode 100644 index 000000000..35d3e8e0f --- /dev/null +++ b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog/phosphor-watchdog.service @@ -0,0 +1,15 @@ +[Unit] +Description=Phosphor Watchdog + +[Service] +ExecStart=/usr/bin/phosphor-watchdog --continue --service=xyz.openbmc_project.Watchdog \ + --path=/xyz/openbmc_project/watchdog/host0 \ + --action_target=xyz.openbmc_project.State.Watchdog.Action.HardReset=xyz.openbmc_project.State.Host.Transition.ForceWarmReboot \ + --action_target=xyz.openbmc_project.State.Watchdog.Action.PowerOff=xyz.openbmc_project.State.Chassis.Transition.Off \ + --action_target=xyz.openbmc_project.State.Watchdog.Action.PowerCycle=xyz.openbmc_project.State.Chassis.Transition.PowerCycle + +BusName=xyz.openbmc_project.Watchdog +Type=dbus + +[Install] +WantedBy=basic.target diff --git a/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog_%.bbappend b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog_%.bbappend new file mode 100644 index 000000000..cc439b2c5 --- /dev/null +++ b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/phosphor-watchdog_%.bbappend @@ -0,0 +1,8 @@ +FILESEXTRAPATHS:append := ":${THISDIR}/${PN}" + +SRC_URI += "file://0001-Customize-phosphor-watchdog-for-Intel-platforms.patch \ + " + +# Remove the override to keep service running after DC cycle +SYSTEMD_OVERRIDE:${PN}:remove = "poweron.conf:phosphor-watchdog@poweron.service.d/poweron.conf" +SYSTEMD_SERVICE:${PN} = "phosphor-watchdog.service" diff --git a/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog.bb b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog.bb new file mode 100644 index 000000000..ac1302d9c --- /dev/null +++ b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog.bb @@ -0,0 +1,12 @@ +SUMMARY = "System watchdog" +DESCRIPTION = "BMC hardware watchdog service that is used to reset BMC \ + when unrecoverable events occurs" + +inherit allarch +inherit obmc-phosphor-systemd + +LICENSE = "Apache-2.0" +LIC_FILES_CHKSUM = "file://${INTELBASE}/COPYING.apache-2.0;md5=34400b68072d710fecd0a2940a0d1658" + +SYSTEMD_SERVICE:${PN} += "system-watchdog.service" +SYSTEMD_ENVIRONMENT_FILE:${PN} += "obmc/system-watchdog/system-watchdog.conf" diff --git a/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog/obmc/system-watchdog/system-watchdog.conf b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog/obmc/system-watchdog/system-watchdog.conf new file mode 100644 index 000000000..defe830a1 --- /dev/null +++ b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog/obmc/system-watchdog/system-watchdog.conf @@ -0,0 +1,3 @@ +TIMEOUT=60 +INTERVAL=10 +DEVICE=/dev/watchdog1 diff --git a/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog/system-watchdog.service b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog/system-watchdog.service new file mode 100644 index 000000000..1564fda20 --- /dev/null +++ b/meta-openbmc-mods/meta-common/recipes-phosphor/watchdog/system-watchdog/system-watchdog.service @@ -0,0 +1,11 @@ +[Unit] +Description=BMC Hardware Watchdog Daemon + +[Service] +EnvironmentFile=/etc/default/obmc/system-watchdog/system-watchdog.conf +ExecStart=/sbin/watchdog -T ${{TIMEOUT}} -t ${{INTERVAL}} -F ${{DEVICE}} +KillSignal=SIGKILL + +[Install] +WantedBy=basic.target + |