commit | 099306430f5219b5f5805368f6052f920af3c260 | [log] [tgz] |
---|---|---|
author | Andrew Jeffery <andrew@aj.id.au> | Thu Jun 08 16:11:53 2023 +0930 |
committer | Andrew Jeffery <andrew@aj.id.au> | Mon Jun 12 21:53:07 2023 +0000 |
tree | 3d4d9a95900c3b905713ee849c4fe745ad3c9e71 | |
parent | 1dc30e8b30f95f6c5a4fbe91ec6e68857857c320 [diff] |
meta-ibm: first-boot-disable-ipmi-net: Improve performance The documentation for systemctl's `disable` verb[1] states: > This command implicitly reloads the system manager configuration after > completing the operation. Note that this command does not implicitly > stop the units that are being disabled. If this is desired, either > combine this command with the --now switch, or invoke the stop command > with appropriate arguments later. [1]: https://www.man7.org/linux/man-pages/man1/systemctl.1.html In simulation, reloads get expensive as they involve a bunch of IO. During boot, under qemu, such a reload can take on the order of 15 seconds. Further, Andrew Geissler states (no reference, private slack): > The service-config-manager is responsible for enabling and disabling > IPMI. On startup, it looks for the service and socket files via > systemd d-bus queries and caches in a file, /etc/srvcfg-mgr.json. If > you don’t run the services once on a fresh startup, > service-config-manager never sees them and therefore never provides a > way to enable/disable them. So on a fresh factory reset boot, you let > them start, service-config-manager sees them and caches them, and then > you disable them from there on out. As evidenced by the `After` statements in the unit, the expectation is that the units we're disabling have already been started. The documentation for systemctl's `mask` verb[1] provides a useful perspective: > Mask one or more units, as specified on the command line. This will > link these unit files to /dev/null, making it impossible to start > them. This is a stronger version of disable, since it prohibits all > kinds of activation of the unit, including enablement and manual > activation. The key insight here is the `disable` keyword tends to best be used to prevent the unit from being started as part of a given target, e.g. the default target. Given that we would only need to reload the systemd configuration to prevent the units from being started as part of a default target, and the fact that they must have already been started in accordance with `After` directive, there's no need to force an immediate reload of the systemd configuration upon disabling the units. Further, it's possible to combine the stop and the disable operation into one with `--now` as demonstrated in the `disable` documentation above. And finally, as the disable verb takes a unit PATTERN and not a singular unit, we can compress the operation down to a single invocation. Tested: 1. Booted up fresh p10bmc (factory reset), verified IPMI disabled by default via Redfish API 2. Enabled IPMI via Redfish API, verified bmcweb indicated IPMI enabled and services running 3. Rebooted BMC, verified IPMI was still enabled after BMC reboot 4. Disabled IPMI over Redfish, rebooted BMC, confirmed IPMI was still disabled Change-Id: I0926e9d16a56c2f022e415f4f40c35695dd155b8 Tested-by: Andrew Geissler <geissonator@yahoo.com> Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
OpenBMC is a Linux distribution for management controllers used in devices such as servers, top of rack switches or RAID appliances. It uses Yocto, OpenEmbedded, systemd, and D-Bus to allow easy customization for your platform.
See the Yocto documentation for the latest requirements
sudo apt install git python3-distutils gcc g++ make file wget \ gawk diffstat bzip2 cpio chrpath zstd lz4 bzip2
sudo dnf install git python3 gcc g++ gawk which bzip2 chrpath cpio \ hostname file diffutils diffstat lz4 wget zstd rpcgen patch
git clone https://github.com/openbmc/openbmc cd openbmc
Any build requires an environment set up according to your hardware target. There is a special script in the root of this repository that can be used to configure the environment as needed. The script is called setup
and takes the name of your hardware target as an argument.
The script needs to be sourced while in the top directory of the OpenBMC repository clone, and, if run without arguments, will display the list of supported hardware targets, see the following example:
$ . setup <machine> [build_dir] Target machine must be specified. Use one of: bletchley mori s8036 dl360poc mtjade swift e3c246d4i mtmitchell tatlin-archive-x86 ethanolx nicole tiogapass evb-ast2500 olympus-nuvoton transformers evb-ast2600 on5263m5 vegman-n110 evb-npcm750 p10bmc vegman-rx20 f0b palmetto vegman-sx20 fp5280g2 qcom-dc-scm-v1 witherspoon g220a quanta-q71l witherspoon-tacoma gbs romed8hm3 x11spi greatlakes romulus yosemitev2 gsj s2600wf zaius kudo s6q lannister s7106
Once you know the target (e.g. romulus), source the setup
script as follows:
. setup romulus
bitbake obmc-phosphor-image
Additional details can be found in the docs repository.
The OpenBMC community maintains a set of tutorials new users can go through to get up to speed on OpenBMC development out here
Commits submitted by members of the OpenBMC GitHub community are compiled and tested via our Jenkins server. Commits are run through two levels of testing. At the repository level the makefile make check
directive is run. At the system level, the commit is built into a firmware image and run with an arm-softmmu QEMU model against a barrage of CI tests.
Commits submitted by non-members do not automatically proceed through CI testing. After visual inspection of the commit, a CI run can be manually performed by the reviewer.
Automated testing against the QEMU model along with supported systems are performed. The OpenBMC project uses the Robot Framework for all automation. Our complete test repository can be found here.
Support of additional hardware and software packages is always welcome. Please follow the contributing guidelines when making a submission. It is expected that contributions contain test cases.
Issues are managed on GitHub. It is recommended you search through the issues before opening a new one.
First, please do a search on the internet. There's a good chance your question has already been asked.
For general questions, please use the openbmc tag on Stack Overflow. Please review the discussion on Stack Overflow licensing before posting any code.
For technical discussions, please see contact info below for Discord and mailing list information. Please don't file an issue to ask a question. You'll get faster results by using the mailing list or Discord.
This is a common question, particularly regarding boards from popular COTS (commercial off-the-shelf) vendors such as Supermicro and ASRock. You can see the list of supported boards by running . setup
(with no further arguments) in the root of the OpenBMC source tree. Most of the platforms supported by OpenBMC are specialized servers operated by companies running large datacenters, but some more generic COTS servers are supported to varying degrees.
If your motherboard is not listed in the output of . setup
it is not currently supported. Porting OpenBMC to a new platform is a non-trivial undertaking, ideally done with the assistance of schematics and other documentation from the manufacturer (it is not completely infeasible to take on a porting effort without documentation via reverse engineering, but it is considerably more difficult, and probably involves a greater risk of hardware damage).
However, even if your motherboard is among those listed in the output of . setup
, there are two significant caveats to bear in mind. First, not all ports are equally mature -- some platforms are better supported than others, and functionality on some "supported" boards may be fairly limited. Second, support for a motherboard is not the same as support for a complete system -- in particular, fan control is critically dependent on not just the motherboard but also the fans connected to it and the chassis that the board and fans are housed in, both of which can vary dramatically between systems using the same board model. So while you may be able to compile and install an OpenBMC build on your system and get some basic functionality, rough edges (such as your cooling fans running continuously at full throttle) are likely.
Dive deeper into OpenBMC by opening the docs repository.
The Technical Steering Committee (TSC) guides the project. Members are: