Add STATUS_WORD to metadata for VIN_UV_FAULT

Change-Id: Iaa6001f7c5d0c558ad3bc01e209dc316236fea93
Signed-off-by: Brandon Wyman <bjwyman@gmail.com>

Update analyze to check VIN_UV_FAULT

The function is a pure virtual function in DeviceMonitor, add in the
implementation of that for PowerSupply. Read the file that represents
that bit from the STAT

Update analyze to check VIN_UV_FAULT

The function is a pure virtual function in DeviceMonitor, add in the
implementation of that for PowerSupply. Read the file that represents
that bit from the STATUS_WORD. If fault is on, report a fault.

Change-Id: I05a4bff997bb0c8b8b71db444e9db0e506765689
Signed-off-by: Brandon Wyman <bjwyman@gmail.com>

show more ...

Update PowerSupply to be derived from Device

The PowerSupply will pass a name and instance number down to the Device
class it is derived from, but will also have an inventory path and a
path to moni

Update PowerSupply to be derived from Device

The PowerSupply will pass a name and instance number down to the Device
class it is derived from, but will also have an inventory path and a
path to monitor for PMBus interfaces.

Change-Id: I29f875fda1f07d031b58ec7ffd381d655495f248
Signed-off-by: Brandon Wyman <bjwyman@gmail.com>

show more ...

Update build framework for PowerSupply fault app

Change-Id: I98a75efc88d92de0ab016a77d5dd4a1e9345df83
Signed-off-by: Brandon Wyman <bjwyman@gmail.com>

Shutdown system on GPU over-temps

Resolves openbmc/openbmc#1726

Change-Id: If3263678bc03df7714f31aa097f38ee6c09389f4
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

Find and call out faulted GPUs

Isolate down to the GPU that caused the GPU PGOOD or
overtemp summary fault bit to turn on. On Witherspoon
this involves reading GPIOs on a pca9552 device to find
the

Find and call out faulted GPUs

Isolate down to the GPU that caused the GPU PGOOD or
overtemp summary fault bit to turn on. On Witherspoon
this involves reading GPIOs on a pca9552 device to find
the GPU signaling the fault.

GPUs are not currently in the inventory, so the code
isn't doing the standard callout by adding a certain
metadata field. The GPU number that failed will just
be added to the error log metadata, and work will be done
with support to make sure that is documented. Also, the
other power fault callouts don't use the standard inventory
callouts either as they are more complicated than just a single
FRU, so this method is consistent with that.

Note that these faults do not cause the system to
power off automatically like other power faults, though
a future commit will power off the system on a GPU overtemp.

Change-Id: If4053f32a06a335a6612a04a8164d34306530b22
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add system data for doing GPU callouts

The 2 GPU fault types (pgood & overtemp) are wired from
each GPU to a summary bit on the UCD90160. If code
detects a summary bit on, it will need to read some

Add system data for doing GPU callouts

The 2 GPU fault types (pgood & overtemp) are wired from
each GPU to a summary bit on the UCD90160. If code
detects a summary bit on, it will need to read some GPIOs
on an IO expander (pca9552) to tell which actual GPU failed
in order to call it out.

This commit provides the data to know when to read the extra
GPIOs, and how those map to the specific faults and GPUs.

Change-Id: I688ddf2ef08b0313b73ed8737eeb01dec059bf40
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add GPU error logging functions

Add functions to log the GPU PGOOD and overtemp
errors.

Change-Id: I6f58d76883f8a78a3301481dbacd111c74b396d4
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

Refactor findGPIODevice

Update findGPIODevice() to work on the path passed
into it instead of just on the gpioDevice member
variable.

Now it can be called both for finding the path for
the UCD chip

Refactor findGPIODevice

Update findGPIODevice() to work on the path passed
into it instead of just on the gpioDevice member
variable.

Now it can be called both for finding the path for
the UCD chip as well as for the device that has the
GPIOs used for GPU isolation.

Change-Id: I01b93ece63cf28a11f0b438741689823280c7e2f
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add GPU errors

Add GPU power fault and overtemp errors.

Change-Id: Iac5782a0db0c5cda2fe478ef8c4e1d7cd3ff4560
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

Add testcases for NamesValues

Change-Id: I61c52f2a196a32dbd4d03a2ccafa2b94414f7555
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

Remove clearFaults calls

The community wasn't thrilled with the device driver
providing the clear_logged_faults command. As it isn't
absolutely necessary for this code to do now, it is being
remove

Remove clearFaults calls

The community wasn't thrilled with the device driver
providing the clear_logged_faults command. As it isn't
absolutely necessary for this code to do now, it is being
removed.

Note: Currently Device::clearFaults is a pure virtual function
so it still needs to be defined in the UCD90160 class.

A future commit may also remove these.

Change-Id: I0b3a33d56987dd97ab7253eb6b5d3b5afd835d67
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add helper class to handle PMBus metadata

Add a NamesValues class to help when adding multiple
PMBus registers to a single metadata entry.

It allows one to add multiple name/value pairs to
the obje

Add helper class to handle PMBus metadata

Add a NamesValues class to help when adding multiple
PMBus registers to a single metadata entry.

It allows one to add multiple name/value pairs to
the object, and provides a get() method to return them
all concatenated.

It currently supports numeric values, up to a uint64_t,
but more types could be added if ever necessary.

The resulting string will look like:
"name1=value1|name2=value2|etc"

Change-Id: I4def8ac313882a0b9efc7cb33488f19ad33f1727
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add method to turn off UCD90160 hardware accesses

As the fault monitoring functionality is going in toward
the end of a release, a flag is being provided to quickly
turn off the hardware accesses wh

Add method to turn off UCD90160 hardware accesses

As the fault monitoring functionality is going in toward
the end of a release, a flag is being provided to quickly
turn off the hardware accesses while still leaving the
ability to create general errors on PGOOD fails, as
well as issuing a shutdown on a runtime PGOOD fail.

This is meant to used if it turns out there are problems
with the hardware that end up taking a lot of time to debug.

The flag is --enable-turn-off-ucd90160-access.

Change-Id: I03f0ab5dc4010bf20ef2871f2e737ce310b4398f
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Monitor UCD90160 for faults at runtime

Add the RuntimeMonitor class that will monitor the
UCD90160 faults in 2 ways:

1) Watch for the PowerLost signal, meaning system
PGOOD was lost. When it oc

Monitor UCD90160 for faults at runtime

Add the RuntimeMonitor class that will monitor the
UCD90160 faults in 2 ways:

1) Watch for the PowerLost signal, meaning system
PGOOD was lost. When it occurs, analyze the
chip for errors and then issue a proper shutdown
so a faulted device doesn't keep getting power.

2) Poll on an interval for nonfatal errors that need
to be logged but don't cause a PGOOD loss.

The main executable can now launch either the PGOODMonitor
or the RuntimeMonitor based on commandline arguments.

Change-Id: If2856f173d5d6288d8333538334b4b4cb4a60097
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Derive PGOODMonitor from DeviceMonitor

Adding this base class so that PGOODMonitor can
fulfill its true purpose, which is checking the
UCD90160 for errors on PGOOD failures.

Change-Id: Ie0637676ae5

Derive PGOODMonitor from DeviceMonitor

Adding this base class so that PGOODMonitor can
fulfill its true purpose, which is checking the
UCD90160 for errors on PGOOD failures.

Change-Id: Ie0637676ae5239c677d60f14d738ff9709d2b7b0
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Fill in the createPowerFaultLog function

Only PGOOD and voltage faults are expected to occur
during normal operation, but when the system PGOOD is lost
and one of those 2 errors wasn't found, we wil

Fill in the createPowerFaultLog function

Only PGOOD and voltage faults are expected to occur
during normal operation, but when the system PGOOD is lost
and one of those 2 errors wasn't found, we will log a generic
power fault log that captures the status_word and mfr_status
registers.

Change-Id: I583c5f6c825adce00ac1b458444fcb8e05900c91
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Fill in PGOOD fault checking code

Fill in the code to check for a PGOOD fault. This is where
the power sequencer device detects that one of its child devices
lost PGOOD. A separate error log will

Fill in PGOOD fault checking code

Fill in the code to check for a PGOOD fault. This is where
the power sequencer device detects that one of its child devices
lost PGOOD. A separate error log will be created for each input
that has a fault. Each input will only have an error logged against
it once for the lifetime of the object.

Errors are detected by reading the real time status of the PGOOD input,
which is exposed as a GPIO by the device driver.

Ideally we would be able to use a summary bit in the status_word
register to see if there is an error before doing any GPIO reads,
but as the device drivers sends a clear faults every time we read
that register, and the GPI fault bits are edge triggered in the
mfr_status register that feeds status_word, we would never detect
any failures. If this was ever fixed in the core PMBus device
driver code, we could add this functionality in and save some
CPU cycles.

Change-Id: I5000f2ebc2b22dcca946154afd2405b29734ccaf
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Find the path for the GPIO device

This path is required to access a GPIO later.

Change-Id: I4ec64adbf939c5f0eaa12b7e18345d0fa2247a7d
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

Add GPIO class

This class will be used for accessing GPIOs off of the
UCD90160 to check for PGOOD faults, and also later for
isolating GPU overtemps and PGOOD faults down to the specific
GPU that fa

Add GPIO class

This class will be used for accessing GPIOs off of the
UCD90160 to check for PGOOD faults, and also later for
isolating GPU overtemps and PGOOD faults down to the specific
GPU that failed.

The FileDescriptor class is used by the GPIO class to manage
the lifetime of the GPIO file handles.

The class only supports reading a GPIO value. At this point
there is no requirement for doing writes.

This class was copied from phosphor-gpio-monitor/gpio-util.

Change-Id: Iee276aed67e1cba549c3070c08238ab5f621c320
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Fill in voltage fault checking code

Fill in the code to check for a voltage fault. This is where
the power sequencer device detects that one of its child
devices has a bad voltage. A separate erro

Fill in voltage fault checking code

Fill in the code to check for a voltage fault. This is where
the power sequencer device detects that one of its child
devices has a bad voltage. A separate error log will be created
for each voltage rail that has a fault. Each rail will only have
an error logged against it once for the lifetime of the object.

There will be support documentation that maps the failing rail
name to the hardware that the rail corresponds to.

Change-Id: I13380b9898613bf8e76d66a72e1fbe005f816dad
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add power sequencer errors

Add the errors that can be created by the UCD90160 device.

Change-Id: Ic12c8cbdaf01602583226c93feeb6b22cdbed283
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

Add clear faults support

Add support to send the clear-logged-faults command to
the UCD90160 chip. This is a manufacturer specific command.
The PMBus standard clear-faults command is done automatic

Add clear faults support

Add support to send the clear-logged-faults command to
the UCD90160 chip. This is a manufacturer specific command.
The PMBus standard clear-faults command is done automatically
by the device driver every time a PMBus sysfs file is read.

Change-Id: I6a9f670502ce1e1b4ffbaab63db335daa0865f46
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add UCD90160 class

This class represents the UCD90160 power sequencer
chip, and provides the ability to check that chip for
voltage and PGOOD faults.

This commit just adds function stubs.

Change-I

Add UCD90160 class

This class represents the UCD90160 power sequencer
chip, and provides the ability to check that chip for
voltage and PGOOD faults.

This commit just adds function stubs.

Change-Id: Iec6e83e9bcddbd476bdd86a887db08f5875f11cd
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...

Add device debug path to PMBus class

Some devices may also have files in
/sys/kernel/debug/<device driver name>.<instance>,
so adding a DeviceDebug path type to support that.

The device driver name

Add device debug path to PMBus class

Some devices may also have files in
/sys/kernel/debug/<device driver name>.<instance>,
so adding a DeviceDebug path type to support that.

The device driver name and chip instance number are then
required to be passed in to the constructor.

Change-Id: I301d730a29ac7c2c39198e4eb7125aff70d727dc
Signed-off-by: Matt Spinler <spinler@us.ibm.com>

show more ...