Add fpga-rsu role to install fpga packages

[ratailor]

This change adds fpga-rsu ansible-role to install
OPAE FPGA packages on hosts where fpga hardware
is already installed.

[matosatti]

Looks good in general, a few comments:

1. Depending on the currently installed firmware version, the upgrade procedure is different.
   see https://www.intel.com/content/www/us/en/programmable/documentation/xgz1560360700260.html#hpo1573151952874

2. There are a number of checks that should be performed before firmware upgrade. From the same document above:

Note: These upgrades erase the Static Region (SR) root entry hash and any CSK cancellation IDs previously programmed in the flash of the Intel® FPGA PAC N3000.
Remember:
Stop any service or daemon accessing the FPGA or XL710 before updating the Intel® FPGA PAC N3000 such as fpgad.
PLDM requests may return stale data. Avoid Host PLDM requests.
Ensure cooling requirements are met. The server can reboot if the FPGA Core temperature exceeds 95°C. For more information, refer to Cooling Requirements.

Tip: Before you proceed with upgrade, ensure that the FPGA Die Temperature is below 80°C using the following command:

sudo fpgainfo bmc

If it is higher than the threshold value, increase the fan speed to improve thermal condition.

[matosatti]

Also, perhaps its good to only attempt to perform RPM installation if we know the packages work on RHEL-8.

The current packages work on RHEL-7 only.

[ratailor]

Thank you Marcelo for review.

We are only trying to install rpm in this and I will update the README to improve doc and update the pull-request.

[matosatti]

On Tue, Jun 23, 2020 at 10:19:22PM -0700, ratailor wrote: @ratailor commented on this pull request. > + when: opae_packages_installed.rc != 0 + + - name: Verify opae and intel packages installed + shell: |- + set -o pipefail + rpm -qa | grep 'opae' + when: opae_packages_installed.rc == 0 + + - name: Verify opae driver installation + shell: |- + set -o pipefail + lsmod | grep fpga + lsmod | grep pac_n3000_net + register: verify_opae_driver + +- name: Verify module is loaded in kernel and update flash Thank you Marcelo for review. We are only trying to install rpm in this and I will update the README to improve doc and update the pull-request.

Right, but: 1) Installing the RPM triggers a firmware update. 2) To perform the firmware update a number of steps must be done (they are not difficult to do, i can help you find out the exact details if needed... drop me an email). So i assume this automation code has to perform those steps (stop any daemons using the card, check temperature).

…

-- You are receiving this because you commented. Reply to this email directly or view it on GitHub: #3 (comment)

[ratailor]

On Tue, Jun 23, 2020 at 10:19:22PM -0700, ratailor wrote: @ratailor commented on this pull request. > + when: opae_packages_installed.rc != 0 + + - name: Verify opae and intel packages installed + shell: |- + set -o pipefail + rpm -qa | grep 'opae' + when: opae_packages_installed.rc == 0 + + - name: Verify opae driver installation + shell: |- + set -o pipefail + lsmod | grep fpga + lsmod | grep pac_n3000_net + register: verify_opae_driver + +- name: Verify module is loaded in kernel and update flash Thank you Marcelo for review. We are only trying to install rpm in this and I will update the README to improve doc and update the pull-request.
Right, but: 1) Installing the RPM triggers a firmware update. 2) To perform the firmware update a number of steps must be done (they are not difficult to do, i can help you find out the exact details if needed... drop me an email). So i assume this automation code has to perform those steps (stop any daemons using the card, check temperature).

Yes, it would be good, if you could provide the steps to perform after RPM installation is complete.

…
-- You are receiving this because you commented. Reply to this email directly or view it on GitHub: #3 (comment)