Update GE2/1 demonstrator layout tree

See merge request !300

Pieter Everaerts authored Jan 19, 2023 and Laurent Petre committed Jan 19, 2023

Implement the module, OptoHybrid and VFAT hardware changes into the
layout tree. Note the M2 module naming convention update, given it is a
production module now.

Monitor the GBTx/lpGBT downlink FEC counters

See merge request !299

Implement lpGBT v1 & ASIAGO v2 compatibility

See merge request !295

Laurent Petre authored Nov 02, 2022 and Laurent Petre committed Dec 14, 2022

So far, the eLink numbers for GEx/1 (GBTx) were given as "logical"
numbers, that is indexing from 0 only the *used* eLinks. In order to
match the ME0 (lpGBT) codeflow, this commits ensures that the eLink
hardware numbering is used throughout the codebase. This change also
simplifies any FPGA, SCA, or S-bit phase scans.

Laurent Petre authored Oct 25, 2022 and Laurent Petre committed Dec 14, 2022

The lpGBT topology on ME0 is much more complicated than the GBTx one on
GEx/1. Each ASIAGO is composed by 2 lpGBT, one master that has both
downlink and uplink connectivity and one slave that has only uplink
connectivity. Clock and slow-controls are provided by the master to the
slave.

Two sets of changes are implemented:

1. Disable the slave lpGBT if the master lpGBT is not ready

2. Perform the appropriate lpGBT configuration sequence:
   * Always reset the chips before pushing a new configuration
   * Configure first the master, then the slave
   * Ensure that the VTRx+ lasers are enabled before starting safe IC
     transactions. Note that this beahvior enforces a few blind
     transactions to reset the lpGBT master and enable the lasers

Laurent Petre authored Sep 30, 2022 and Laurent Petre committed Dec 14, 2022

Despite the firmware abstraction between the GBTx, lpGBT v0, and lpGBT
v1 chips, some particularities cannot be hidden to the user.

This commit implement a software level abstraction of the GBT IC
communication channel. The GBTx and lpGBT v1 version are fully
supported; the lpGBT v0 is not supported.

The following abstractions are implemented:

* Select the right IC frame format and I2C address
* Allow the usage of different register ranges
* Implement the lpGBT default configuration file format

Based on the initial work from:
 Antonello Pellecchia <antonello.pellecchia@cern.ch>

Fix dumb typo in local run number generation

See merge request !298

Before this commit and since their introduction, the generated local run
numbers were not following the expectations. While not strictly a
problem for their uniqueness, the (approximate) run start time could not
be computed.

Expose monitorables for the front-end power monitoring tools

Closes #248

See merge request !297

Add unpacker for the v301b and v302a data formats

Closes #200

See merge request !261

The DWARF symbols compression step of the find-debuginfo.sh (using dwc)
is broken for some of the .so files (e.g. unknown symbol types, failed
asserts,...) Since the debuginfo RPM packages are not heavy and rarely
installed, just skip this this compression step.

Raul Rabadan authored Jul 21, 2022 and Laurent Petre committed Nov 03, 2022

Co-authored-by: Laurent Pétré <laurent.petre@cern.ch>

Raul Rabadan authored May 03, 2022 and Laurent Petre committed Nov 03, 2022

Co-authored-by: Antonello Pellecchia <antonello.pellecchia@cern.ch>
Co-authored-by: Laurent Pétré <laurent.petre@cern.ch>

Implement automatic DAQ power on

See merge request !296

It was shown during recent experimentations that simultaneous DIM RPC
calls to the DCS command services do not trigger a segmentation fault or
any other kind of error. It is unclear was happened in the first
place...

Give another try to simultaneous commands and fix the issue would it
occur again.

Reverts 1cc1d6e7.

Before this commit, an existing installation of PostgreSQL 9 from the
standard repositories in the standard paths was always privileged over
an installation of PostgreSQL 10 from the SCL repositories in the SCL
paths.

The previous heuristic was trying to find a suitable PostgreSQL version
via find_package. Such a method was setting the PostgreSQL_* cache
variables to the standard paths regardless of the presence of a suitable
version. It was de facto defining the PostgreSQL paths for any
subsequent package lookup.

Since calling find_package without cache is between very unpractical and
impossible, the PostgreSQL_ROOT variable is now set first whenever the
PostgreSQL 10 SCL is detected. The subsequent find_package ensure the
right version could be found (and sets the cache variables for the
future lookups).

Miscellenaous bug-fixes from p5 deployment

See merge request !294

The DCS exporter was looking for monitorables into the JSON object root
instead than in each of the monitorable groups. Look in all groups until
there is a match instead.

While trying to deploy in production the layout tree from repository,
multiples inconsistencies arose. Fix them.

* Use in CMS the same convention between GE1/1 and GE2/1 for the display
  and wire names.

* Fix the GE2/1 YAML filenames. They should be based on the UID.

* Use the same name between the "parts" and the layout tree root
  directory. This is meant to simplify the deployment.

Recent updates of the DAQ at p5 show some glitches to reset the MGT,
usually just after the updates. Make the MGT reset more rugged and add
debugging information.

In theory, the run number is sent during the Start transition and
well-defined only during the Running state.

However, the current workflow imposes the GEMSupervisor to generate a
new local run number for every configuration during configuration. This
is required to provide the the calibration scans a pristine run number.

As a consequence, the GEM function manager must set the isLocalRunNumber
parameter of the GEMSupervisor prior to triggering the Configure state
transition.

Although the AnyState to Error transition was defined, the SetError
input and Error state were not included in the FSM description, leading to
a malformed state machine. Include all required states, inputs, and
state transtions.

During the JSON object key renaming of 22f39fb7, the FEDMonitor
JavaScript update script was forgotten. Consequently, the state value
was never updated in the FEDMonitor Expert Page tab, despite the state
being well defined. Fix the issue and get the current state from the
'fsm-state' key (instead of 'value').