Skip to contentSkip to conversation

Call for Gardening Tasks

DISCUSSION Sui Chen 7 participants 9 messages Apr 20, 2020 – Apr 23, 2020

lore
Share & tools
lore

Sui Chen

author
#0 lore raw
Quoted 3 lines · Andrew Geissler

Systemd Visualization
Another complicated area of OpenBMC is our systemd targets and services. Building on the upstream tools to visualize our systemd targets and services would be useful to new people.


+1 to visualization, and I have a few thoughts on this ---
My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned – visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.
The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?
I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.

Thanks
Sui

#1 lore raw

about power control.回复: Call for Gardening Tasks

Hi

This is a very good idea. As the first time to do openbmc, I look forward to it. Currently, I am being troubled by the problems you describe.

The following questions, please also help you answer it, thank you very much.

1.I modified the dts file, how to use the command "bitbake -f linux-aspeed" to Generate fitimage and dub?
Now, I modify the dts file in the kernel source directory. When I compile, bitbake will re-fetch the source code, decompress, patch, configure, and compile, causing my changes to be overwritten. I currently use manual call scripts to compile dts files and make images for debugging.Is there a way to make bitbake skip the previous steps and compile directly?

2.The size of image-rwfs is 0, I want to use rwfs debugging, how to configure to generate a normal image-rwfs?
-rw-r--r--. 2 harleyzhou harleyzhou 9443 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.manifest
-rw-r--r--. 2 harleyzhou harleyzhou 18214912 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.squashfs-xz
-rw-rw-r--. 2 harleyzhou harleyzhou 33554432 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.static.mtd
-rw-rw-r--. 2 harleyzhou harleyzhou 33566720 4月 7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.all.tar
-rw-rw-r--. 2 harleyzhou harleyzhou 22384640 4月 7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.tar
-rw-r--r--. 2 harleyzhou harleyzhou 308218 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.testdata.json
-rw-r--r--. 2 harleyzhou harleyzhou 0 4月 7 17:27 obmc-phosphor-image-xxh.jffs2

3. I did not find the source code of "pgood_wait" and "power_control.exe" in obmc-op-control-power? where can I get it ?

4.OpenBMC uses systemd<https://www.freedesktop.org/wiki/Software/systemd/> to manage all processes, So i analyzed the process of systemd starting process and saw "default.target-> basic.target-> sysinit.target-> local-fs.target", there is no content behind.please help to provide clues of systemd starting process.

5.I understand the following execution process of power-on , help check if it is correct.
a).rest or busctl send power-on commands to systemd through d-bus messages.
b).systemd find the [email protected]
c).systemd find the [email protected]
d).systemd find the [email protected]
e).systemd start a thread and execute "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState I 1"

appendix:

ls [email protected]

[email protected] [email protected]

cat [email protected]

[Unit]

Description=Chassis%i (Power On)

Wants=multi-user.target

After=multi-user.target

Wants=mapper-wait@-xyz-openbmc_project-state-chassis%i.service

After=mapper-wait@-xyz-openbmc_project-state-chassis%i.service

Wants=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

After=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

cat [email protected]

[Unit]

Description=Start Power%i

Wants=obmc-power-start@%i.target

Before=obmc-power-start@%i.target

After=obmc-power-start-pre@%i.target

After=obmc-fan-control.target

Wants=mapper-wait@-org-openbmc-control-power%i.service

After=mapper-wait@-org-openbmc-control-power%i.service

[Service]

RemainAfterExit=yes

Type=oneshot

ExecStart=/bin/sh -c "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState i 1"

SyslogIdentifier=op-power-start

[Install]

WantedBy=obmc-host-start@%i.target

cat [email protected]

[Unit]

Description=Power%i On

After=obmc-power-start-pre@%i.target

Wants=multi-user.target

Conflicts=obmc-chassis-poweroff@%i.target

RefuseManualStart=yes

RefuseManualStop=yes

Thanks
Harley
________________________________
发件人: openbmc <[email protected]> 代表 Sui Chen <[email protected]> 发送时间: 2020年4月21日 1:05
收件人: [email protected] <[email protected]>
主题: Re: Call for Gardening Tasks

Quoted 5 lines · Andrew Geissler

Systemd Visualization

Another complicated area of OpenBMC is our systemd targets and services. Building on the upstream tools to visualize our systemd targets and services would be useful to new people.

+1 to visualization, and I have a few thoughts on this ---

My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned – visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.

The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?

I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.

Thanks

Sui

#6 lore raw
In reply to Sui Chen

Thank you everyone for your suggestions.

Scanning through this it's clear that documentation is one of the biggest most generally agreed upon area to work on.

Build tools and package management seems like another area with some agreed on improvements. I like the idea of moving to meson. Getting tidy files and revising our style guides also sound like good ideas. I also really agree that we could stand to do some consolidation of processes/repos.

There is also some use for DBus tooling for both improved code, and for visualizations.

So here are my next steps:
1) Since we have some new people getting started, I've asked them to add
documentation at the points they get stuck during on-boarding.
2) I'm going to take the ideas in this thread and put together a .md doc
with any relevant information I can find. I'll be adding it to openbmc/docs.
3) I'll follow up in another thread about some of the ideas around DBus
tooling and package consolidation. That seems to be an area where most people agree some improvements can be made, but I'm not sure there is an agreement on how we should do it.

-Richard

Quoted 59 lines · Sui Chen
Quoted 1 line

On Apr 11, 2020, at 8:15 AM, Andrew Geissler <geissonator at gmail.com>

Quoted 3 lines · wrote:

Systemd Visualization

Another complicated area of OpenBMC is our systemd targets and services.

Building on the upstream tools to visualize our systemd targets and services would be useful to new people.

+1 to visualization, and I have a few thoughts on this ---

My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned – visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.

The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?

I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.

Thanks

Sui

#2 lore raw

Re: about power control.回复: Call for Gardening Tasks

In reply to [email protected]

Regarding the power control related, sugest to check on the below:

https://github.com/openbmc/x86-power-control/blob/master/power-control-x86/src/power_control.cpp

It uses litter systems related services/targets, and monitor these gpio signals, to trigger different callback actions.

For the pgood_wait, it is in:

https://github.com/openbmc/skeleton/tree/master/op-pwrctl/pgood_wait

https://github.com/openbmc/skeleton/tree/master/op-pwrctl

Thanks,

Yong

Quoted 138 lines · 在 4/21/2020 10:50 PM, [email protected] 写道:

Hi

    This is a very good idea. As the first time to do openbmc, I look forward to it. Currently, I am being troubled by the problems you describe.

    The following questions, please also help you answer it, thank you very much.

    1.I modified the dts file, how to use the command "bitbake -f linux-aspeed" to Generate fitimage and dub?
     Now, I modify the dts file in the kernel source directory. When I compile, bitbake will re-fetch the source code, decompress, patch, configure, and compile, causing my changes to be overwritten. I currently use manual call scripts to compile dts files and make images for debugging.Is there a way to make bitbake skip the previous steps and compile directly?

     2.The size of image-rwfs is 0, I want to use rwfs debugging, how to configure to generate a normal image-rwfs? -rw-r--r--. 2 harleyzhou harleyzhou     9443     4月        7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.manifest
     -rw-r--r--. 2 harleyzhou harleyzhou 18214912 4月    7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.squashfs-xz -rw-rw-r--. 2 harleyzhou harleyzhou 33554432        4月   7 17:27 obmc-phosphor-image-xxh-20200407092558.static.mtd -rw-rw-r--. 2 harleyzhou harleyzhou 33566720        4月   7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.all.tar -rw-rw-r--. 2 harleyzhou harleyzhou 22384640        4月   7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.tar -rw-r--r--. 2 harleyzhou harleyzhou   308218  4月      7 17:27 obmc-phosphor-image-xxh-20200407092558.testdata.json -rw-r--r--. 2 harleyzhou harleyzhou           0  4月   7 17:27 obmc-phosphor-image-xxh.jffs2

    3. I did not find the source code of "pgood_wait" and "power_control.exe" in obmc-op-control-power? where can I get it ?

    4.OpenBMC usessystemd <https://www.freedesktop.org/wiki/Software/systemd/>to manage all processes, So i analyzed the process of systemd starting process and saw "default.target-> basic.target-> sysinit.target-> local-fs.target", there is no content behind.please help to provide clues of systemd starting process.

    5.I understand the following execution process of power-on , help check if it is correct.
       a).rest or busctl send power-on commands to systemd through d-bus messages.
       b).systemd find the [email protected]
  c).systemd find the [email protected]
d).systemd find the [email protected]
e).systemd start a thread and execute "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState I 1"

appendix:

ls [email protected]

[email protected]@0.service

cat [email protected]

[Unit]

Description=Chassis%i (Power On)

Wants=multi-user.target

After=multi-user.target

Wants=mapper-wait@-xyz-openbmc_project-state-chassis%i.service

After=mapper-wait@-xyz-openbmc_project-state-chassis%i.service

Wants=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

After=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

cat [email protected]

[Unit]

Description=Start Power%i

Wants=obmc-power-start@%i.target

Before=obmc-power-start@%i.target

After=obmc-power-start-pre@%i.target

After=obmc-fan-control.target

Wants=mapper-wait@-org-openbmc-control-power%i.service

After=mapper-wait@-org-openbmc-control-power%i.service

[Service]

RemainAfterExit=yes

Type=oneshot

ExecStart=/bin/sh -c "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState i 1"

SyslogIdentifier=op-power-start

[Install]

WantedBy=obmc-host-start@%i.target

cat [email protected]

[Unit]

Description=Power%i On

After=obmc-power-start-pre@%i.target

Wants=multi-user.target

Conflicts=obmc-chassis-poweroff@%i.target

RefuseManualStart=yes

RefuseManualStop=yes

Thanks
Harley
------------------------------------------------------------------------
*发件人:* openbmc <[email protected]> 代表 Sui Chen <[email protected]> *发送时间:* 2020年4月21日 1:05
*收件人:* [email protected] <[email protected]>
*主题:* Re: Call for Gardening Tasks

Quoted 5 lines · Andrew Geissler

Systemd Visualization

Another complicated area of OpenBMC is our systemd targets and services. Building on the upstream

tools to visualize our systemd targets and services would be useful to new people.

+1 to visualization, and I have a few thoughts on this ---

My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned – visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.

The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?

I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.

Thanks

Sui

#3 lore raw

Re: about power control.回复: Call for Gardening Tasks

In reply to [email protected]

Hello! I'll try to answer your question about rwfs image.I had the same issue plus I wanted to be able to create some data in rwfs segment.Thus I've created obmc-phosphor-image.bbappend where I do the following:======================================================================RWFS_SIZE ?= "4194304"OVERLAY_MKFS_OPTS = " --pad=${RWFS_SIZE}" # Override JFFS image creationdo_generate_rwfs_static() {rwdir=$(pwd)rwdir=${rwdir}/jffs2image=rwfs.jffs2 rm -rf $rwdir $image > /dev/null 2>&1mkdir -p ${rwdir}/cowrwdir=${rwdir}/cow bbplain "DEBUG: Creating RWFS image " mv ${IMAGE_ROOTFS}/rwfs/* ${rwdir}/rm -rf ${IMAGE_ROOTFS}/rwfs # Create necessary files, directories, etc in ${rwdir}${JFFS2_RWFS_CMD} ${OVERLAY_MKFS_OPTS} ${OVERLAY_MKFS_OPTS} --squash-uids}====================================================================== At the end I have the normal JFFS2 image instead of empty. Hope this can help. 21.04.2020, 18:24, "[email protected]" <[email protected]>:Hi      This is a very good idea. As the first time to do openbmc, I look forward to it. Currently, I am being troubled by the problems you describe.     The following questions, please also help you answer it, thank you very much.     1.I modified the dts file, how to use the command "bitbake -f linux-aspeed" to Generate fitimage and dub?     Now, I modify the dts file in the kernel source directory. When I compile, bitbake will re-fetch the source code, decompress, patch, configure, and compile, causing my changes to be overwritten. I currently use manual call scripts to compile dts files and make images for debugging.Is there a way to make bitbake skip the previous steps and compile directly?       2.The size of image-rwfs is 0, I want to use rwfs debugging, how to configure to generate a normal image-rwfs?      -rw-r--r--. 2 harleyzhou harleyzhou     9443             4月        7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.manifest     -rw-r--r--. 2 harleyzhou harleyzhou 18214912         4月    7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.squashfs-xz     -rw-rw-r--. 2 harleyzhou harleyzhou 33554432        4月   7 17:27 obmc-phosphor-image-xxh-20200407092558.static.mtd     -rw-rw-r--. 2 harleyzhou harleyzhou 33566720        4月   7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.all.tar     -rw-rw-r--. 2 harleyzhou harleyzhou 22384640        4月   7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.tar     -rw-r--r--. 2 harleyzhou harleyzhou   308218            4月      7 17:27 obmc-phosphor-image-xxh-20200407092558.testdata.json     -rw-r--r--. 2 harleyzhou harleyzhou           0               4月   7 17:27 obmc-phosphor-image-xxh.jffs2      3. I did not find the source code of "pgood_wait" and "power_control.exe" in obmc-op-control-power? where can I get it ?      4.OpenBMC uses systemd to manage all processes, So i analyzed the process of systemd starting process and saw "default.target-> basic.target-> sysinit.target-> local-fs.target", there is no content behind.please help to provide clues of systemd starting process.     5.I understand the following execution process of power-on , help check if it is correct.       a).rest or busctl send power-on commands to systemd through d-bus messages.       b).systemd find the [email protected]        c).systemd find the [email protected]        d).systemd find the [email protected]         e).systemd start a thread and execute "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState I 1" appendix:ls [email protected] [email protected]  [email protected]

cat [email protected]
[Unit]
Description=Chassis%i (Power On)
Wants=multi-user.target
After=multi-user.target
Wants=mapper-wait@-xyz-openbmc_project-state-chassis%i.service
After=mapper-wait@-xyz-openbmc_project-state-chassis%i.service
Wants=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service
After=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

cat [email protected]
[Unit]
Description=Start Power%i
Wants=obmc-power-start@%i.target
Before=obmc-power-start@%i.target
After=obmc-power-start-pre@%i.target
After=obmc-fan-control.target
Wants=mapper-wait@-org-openbmc-control-power%i.service
After=mapper-wait@-org-openbmc-control-power%i.service
[Service]
RemainAfterExit=yes
Type=oneshot
ExecStart=/bin/sh -c "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState i 1"
SyslogIdentifier=op-power-start
[Install]
WantedBy=obmc-host-start@%i.target

cat [email protected]
[Unit]
Description=Power%i On
After=obmc-power-start-pre@%i.target
Wants=multi-user.target
Conflicts=obmc-chassis-poweroff@%i.target
RefuseManualStart=yes
RefuseManualStop=yes
 ThanksHarley发件人:openbmc <[email protected]> 代表 Sui Chen <[email protected]>
发送时间:2020年4月21日 1:05
收件人: [email protected]<[email protected]>

Quoted 2 lines · Andrew Geissler

Systemd Visualization
Another complicated area of OpenBMC is our systemd targets and services. Building on the upstream tools to visualize our systemd targets and services would be useful to new people.


+1 to visualization, and I have a few thoughts on this ---
My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned C visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.
The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?
I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.

Thanks
Sui
  -- Best regards,Konstantin Klubnichkin,lead firmware engineer,server hardware R&D group,Yandex Moscow office.tel: +7-903-510-33-33

#4 lore raw

回复: about power control.回复: Call for Gardening Tasks

Thank you for the help, I will test it.

________________________________
发件人: Konstantin Klubnichkin <[email protected]>
发送时间: 2020年4月22日 15:25
收件人: [email protected] <[email protected]>; Sui Chen <[email protected]>; [email protected] <[email protected]> 抄送: [email protected] <[email protected]>; [email protected] <[email protected]>
主题: Re: about power control.回复: Call for Gardening Tasks

Hello!

I'll try to answer your question about rwfs image.
I had the same issue plus I wanted to be able to create some data in rwfs segment. Thus I've created obmc-phosphor-image.bbappend where I do the following:
======================================================================
RWFS_SIZE ?= "4194304"
OVERLAY_MKFS_OPTS = " --pad=${RWFS_SIZE}"

# Override JFFS image creation
do_generate_rwfs_static() {
rwdir=$(pwd)
rwdir=${rwdir}/jffs2
image=rwfs.jffs2

rm -rf $rwdir $image > /dev/null 2>&1
mkdir -p ${rwdir}/cow
rwdir=${rwdir}/cow

bbplain "DEBUG: Creating RWFS image "

mv ${IMAGE_ROOTFS}/rwfs/* ${rwdir}/
rm -rf ${IMAGE_ROOTFS}/rwfs

# Create necessary files, directories, etc in ${rwdir} ${JFFS2_RWFS_CMD} ${OVERLAY_MKFS_OPTS} ${OVERLAY_MKFS_OPTS} --squash-uids }
======================================================================

At the end I have the normal JFFS2 image instead of empty.

Hope this can help.

21.04.2020, 18:24, "[email protected]" <[email protected]>:
Hi

This is a very good idea. As the first time to do openbmc, I look forward to it. Currently, I am being troubled by the problems you describe.

The following questions, please also help you answer it, thank you very much.

1.I modified the dts file, how to use the command "bitbake -f linux-aspeed" to Generate fitimage and dub?
Now, I modify the dts file in the kernel source directory. When I compile, bitbake will re-fetch the source code, decompress, patch, configure, and compile, causing my changes to be overwritten. I currently use manual call scripts to compile dts files and make images for debugging.Is there a way to make bitbake skip the previous steps and compile directly?

2.The size of image-rwfs is 0, I want to use rwfs debugging, how to configure to generate a normal image-rwfs?
-rw-r--r--. 2 harleyzhou harleyzhou 9443 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.manifest
-rw-r--r--. 2 harleyzhou harleyzhou 18214912 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.squashfs-xz
-rw-rw-r--. 2 harleyzhou harleyzhou 33554432 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.static.mtd
-rw-rw-r--. 2 harleyzhou harleyzhou 33566720 4月 7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.all.tar
-rw-rw-r--. 2 harleyzhou harleyzhou 22384640 4月 7 17:28 obmc-phosphor-image-xxh-20200407092558.static.mtd.tar
-rw-r--r--. 2 harleyzhou harleyzhou 308218 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.testdata.json
-rw-r--r--. 2 harleyzhou harleyzhou 0 4月 7 17:27 obmc-phosphor-image-xxh.jffs2

3. I did not find the source code of "pgood_wait" and "power_control.exe" in obmc-op-control-power? where can I get it ?

4.OpenBMC uses systemd<https://www.freedesktop.org/wiki/Software/systemd/> to manage all processes, So i analyzed the process of systemd starting process and saw "default.target-> basic.target-> sysinit.target-> local-fs.target", there is no content behind.please help to provide clues of systemd starting process.

5.I understand the following execution process of power-on , help check if it is correct.
a).rest or busctl send power-on commands to systemd through d-bus messages.
b).systemd find the [email protected]
c).systemd find the [email protected]
d).systemd find the [email protected]<mailto:[email protected]>
e).systemd start a thread and execute "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState I 1"

appendix:

ls [email protected]<mailto:[email protected]>

[email protected]<mailto:[email protected]>e [email protected]<mailto:[email protected]>

cat [email protected]

[Unit]

Description=Chassis%i (Power On)

Wants=multi-user.target

After=multi-user.target

Wants=mapper-wait@-xyz-openbmc_project-state-chassis%i.service<mailto:Wants=mapper-wait@-xyz-openbmc_project-state-chassis%i.service>

After=mapper-wait@-xyz-openbmc_project-state-chassis%i.service<mailto:After=mapper-wait@-xyz-openbmc_project-state-chassis%i.service>

Wants=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

After=mapper-subtree-remove@-xyz-openbmc\x5fproject-software\x3Axyz.openbmc_project.Software.ActivationBlocksTransition.service

cat [email protected]<mailto:[email protected]>e

[Unit]

Description=Start Power%i

Wants=obmc-power-start@%i.target<mailto:Wants=obmc-power-start@%i.target>

Before=obmc-power-start@%i.target<mailto:Before=obmc-power-start@%i.target>

After=obmc-power-start-pre@%i.target<mailto:After=obmc-power-start-pre@%i.target>

After=obmc-fan-control.target

Wants=mapper-wait@-org-openbmc-control-power%i.service<mailto:Wants=mapper-wait@-org-openbmc-control-power%i.service>

After=mapper-wait@-org-openbmc-control-power%i.service<mailto:After=mapper-wait@-org-openbmc-control-power%i.service>

[Service]

RemainAfterExit=yes

Type=oneshot

ExecStart=/bin/sh -c "busctl call mapper get-service /org/openbmc/control/power%i /org/openbmc/control/power%i org.openbmc.control.Power setPowerState i 1"

SyslogIdentifier=op-power-start

[Install]

WantedBy=obmc-host-start@%i.target<mailto:WantedBy=obmc-host-start@%i.target>

cat [email protected]

[Unit]

Description=Power%i On

After=obmc-power-start-pre@%i.target<mailto:After=obmc-power-start-pre@%i.target>

Wants=multi-user.target

Conflicts=obmc-chassis-poweroff@%i.target<mailto:Conflicts=obmc-chassis-poweroff@%i.target>

RefuseManualStart=yes

RefuseManualStop=yes

Thanks
Harley
________________________________
发件人: openbmc <[email protected]<mailto:[email protected]>> 代表 Sui Chen <[email protected]<mailto:[email protected]>> 发送时间: 2020年4月21日 1:05
收件人: [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>>
主题: Re: Call for Gardening Tasks

Quoted 5 lines · Andrew Geissler

Systemd Visualization

Another complicated area of OpenBMC is our systemd targets and services. Building on the upstream tools to visualize our systemd targets and services would be useful to new people.

+1 to visualization, and I have a few thoughts on this ---

My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned C visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.

The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?

I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.

Thanks

Sui

--
Best regards,
Konstantin Klubnichkin,
lead firmware engineer,
server hardware R&D group,
Yandex Moscow office.
tel: +7-903-510-33-33

#7 lore raw
In reply to Richard Hanley
Quoted 6 lines · Richard Hanley

Thank you everyone for your suggestions.

Scanning through this it's clear that documentation is one of the biggest most generally agreed upon area to work on.

Build tools and package management seems like another area with some agreed on improvements. I like the idea of moving to meson. Getting tidy files and revising our style guides also sound like good ideas. I also really agree that we could stand to do some consolidation of processes/repos.

Quoted 7 lines

There is also some use for DBus tooling for both improved code, and for visualizations.

So here are my next steps:
  1) Since we have some new people getting started, I've asked them to
add documentation at the points they get stuck during on-boarding.
  2) I'm going to take the ideas in this thread and put together a .md
doc with any relevant information I can find. I'll be adding it to openbmc/docs.

That would work, but I would also suggest a wiki page. Docs have a tendency to get stale quickly, where a wiki page has a lower pain point for a small changes. I can get that started if it would help.

Kurt Taylor (krtaylor)

Quoted 68 lines

  3) I'll follow up in another thread about some of the ideas around
DBus tooling and package consolidation. That seems to be an area where most people agree some improvements can be made, but I'm not sure there is an agreement on how we should do it.

-Richard

Quoted 5 lines · On Mon, Apr 20, 2020 at 10:19 AM Sui Chen <[email protected]…

On Apr 11, 2020, at 8:15 AM, Andrew Geissler <geissonator at gmail.com <http://gmail.com>> wrote:

Systemd Visualization

Another complicated area of OpenBMC is our systemd targets and services. Building on the upstream tools to visualize our systemd targets and services would be useful to new people.

__ __

+1 to visualization, and I have a few thoughts on this ---____

My lacking a mental model of how an OpenBMC system works had been my pain point in the first few months working with OpenBMC (I’m a bit new to this area), so after learning the minimal set of survival skills I did something similar to what you mentioned – visualizing the messages passed between different dbus peers (and conveniently, IPMI traffic, as IPMI-related dbus messages exposing all IPMI payload comprise most of the dbus traffic on that particular system I was working on.)____

__ __

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.____

__ __

The IPMI stack is interesting because it’s one noticeable workload on the particular BMC system I had been working on; in my current limited understanding, having lots of I/O capability to connect to hundreds of sensors is one of the many features that set apart a BMC chip and a similarly powerful smartphone chip, and the broad use of dbus is what sets apart OpenBMC and the desktop Linux distro I had been using. I heard optimization is best done workload by workload, perhaps this rationale applies to visualization too?____

__ __

I realize I was mostly talking about visualizing the run-time state of the system rather than build-time, but we could visualize the run-time aspect of systemd units too since I have seen many times a dbus message eventually triggering a systemd target to acutate the system, so it would be good to consider both dbus and systemd (and maybe other parts of the system?) to have a more holistic view of the BMC’s operations.____

__ __

Thanks____

Sui____

#5 lore raw

Re: about power control.回复: Call for Gardening Tasks

In reply to [email protected]
Quoted 8 lines · On Wed, 22 Apr 2020, at 00:20, [email protected] wrote:

Hi

This is a very good idea. As the first time to do openbmc, I look
forward to it. Currently, I am being troubled by the problems you describe.

The following questions, please also help you answer it, thank you very much.

Quoted 8 lines

1.I modified the dts file, how to use the command "bitbake -f
linux-aspeed" to Generate fitimage and dub?
Now, I modify the dts file in the kernel source directory. When I
compile, bitbake will re-fetch the source code, decompress, patch, configure, and compile, causing my changes to be overwritten. I currently use manual call scripts to compile dts files and make images for debugging.Is there a way to make bitbake skip the previous steps and compile directly?

Two approaches here:

1. Use devtool[1], (devtool modify linux-aspeed), and then hack on the source that gets extracted

2. Use your own fork of the kernel hack the linux-aspeed recipe to point to your local repo.

[1] https://www.yoctoproject.org/docs/2.0/dev-manual/dev-manual.html#dev-modifying-source-code

Diverging further, you can just build your own kernel out of your kernel source tree and netboot it on the BMC (either hardware or qemu). It depends on what your needs are, but netbooting is great for doing kernel development and testing (as opposed to building images to be deployed).

Quoted 18 lines

2.The size of image-rwfs is 0, I want to use rwfs debugging, how to
configure to generate a normal image-rwfs?
-rw-r--r--. 2 harleyzhou harleyzhou 9443 4月 7 17:27 obmc-phosphor-image-xxh-20200407092558.rootfs.manifest
-rw-r--r--. 2 harleyzhou harleyzhou 18214912 4月 7 17:27
obmc-phosphor-image-xxh-20200407092558.rootfs.squashfs-xz
-rw-rw-r--. 2 harleyzhou harleyzhou 33554432 4月 7 17:27
obmc-phosphor-image-xxh-20200407092558.static.mtd
-rw-rw-r--. 2 harleyzhou harleyzhou 33566720 4月 7 17:28
obmc-phosphor-image-xxh-20200407092558.static.mtd.all.tar
-rw-rw-r--. 2 harleyzhou harleyzhou 22384640 4月 7 17:28
obmc-phosphor-image-xxh-20200407092558.static.mtd.tar
-rw-r--r--. 2 harleyzhou harleyzhou 308218 4月 7 17:27
obmc-phosphor-image-xxh-20200407092558.testdata.json
-rw-r--r--. 2 harleyzhou harleyzhou 0 4月 7 17:27
obmc-phosphor-image-xxh.jffs2

As long as the flash partition is erased jffs2 will be happy to mount it, so depending on your circumstances you may not need any data in image-rwfs.

Hope that helps,

Andrew

#8 lore raw
In reply to Sui Chen
Quoted 5 lines · Sui Chen
Quoted 1 line · Andrew Geissler
Quoted 10 lines

I think packet analysis tools such as Wireshark and graphics frame analysis tools such as RenderDoc, or system event-based Windows Performance tool like GPUView provide great examples of what people might expect to achieve with a visualization tool: capture, inspect and (sometimes) playback, across multiple layers in the software/hardware stack. Many similar existing tools process sequences of events, and in this case of BMCs, the events could be dbus messages. I found a prototype visualizer made at work greatly helpful in explaining to new team members some basic concepts and the IPMI stack on the BMC.

Not really a visualisation tool, but I'm going to plug my dbus pcap parser here - it's helped me analyse a number of silly problems of dbus interactions between a number of applications:

https://github.com/openbmc/openbmc-tools/blob/master/amboar/obmc-scripts/dbus-pcap/dbus-pcap

It desperately needs a README, but you can slice and dice busctl capture output with DBus filter specifications, and if you like dump the output as json for further processing.

Andrew