chassis power status seems to not be changed when type command to shutdown OS
Hi Team,
When I executed command "poweroff" or "shutdown -h" in Host OS, the chassis power status seems to not be changed. Is there any package could support update power status for this situation?
Thanks,
Andrew
Quoted 4 lines · On Thu, Aug 22, 2019 at 5:01 AM Andrew MS1 Peng <[email protected]> wrote:
Hi Team,
When I executed command "poweroff" or "shutdown -h" in Host OS, the chassis power status seems to not be changed. Is there any package could support update power status for this situation?
Sounds like your host is not sending the inband IPMI message to the BMC to turn off the chassis once it completes the shutdown?
Quoted 5 lines
Thanks,
Andrew
There's also quite a bit of infrastructure on the OpenBMC side that needs to be running to be able for the BMC to actually power on / off the host.
Can you power-on / off the host using "ipmitool power on" / "off" from the BMC? If that works, then the host should be able to do the same.
Oskar.
Quoted 18 lines · Andrew Geissler
Quoted 4 lines · On Thu, Aug 22, 2019 at 5:01 AM Andrew MS1 Peng <[email protected]> wrote:
Hi Team,
When I executed command "poweroff" or "shutdown -h" in Host OS, the
chassis power status seems to not be changed. Is there any package could support update power status for this situation?
Sounds like your host is not sending the inband IPMI message to the BMC to turn off the chassis once it completes the shutdown?
Quoted 5 lines
Thanks,
Andrew
How does the "ipmitool power on/off" do to power on/off the host? I assume it need to trigger the power button connected to BMC gpio, right? But I am not quite sure about the path which the message goes from ipmitool to gpio.
What I have got:
1. The gpio defined in machine layer, gpio_defs.json, which is handled by obmc-libobmc-intf.
2. obmc-libobmc-intf is part of "skeleton“.
The decription of skeleton on github says "Area for initial design & reference code. Will be replaced with proper implementation and design docs. " What does it mean?
Am I on track? Is power on/off gpio handled by skeleton? Thanks.
Can any big boy give some hint? Thanks.
Zheng
________________________________
From: openbmc <[email protected]> on behalf of Oskar Senft <[email protected]>
Sent: Wednesday, August 28, 2019 9:29 PM
To: Andrew Geissler <[email protected]>
Cc: Andrew MS1 Peng <[email protected]>; [email protected] <[email protected]>
Subject: Re: chassis power status seems to not be changed when type command to shutdown OS
There's also quite a bit of infrastructure on the OpenBMC side that needs to be running to be able for the BMC to actually power on / off the host.
Can you power-on / off the host using "ipmitool power on" / "off" from the BMC? If that works, then the host should be able to do the same.
Oskar.
On Thu, Aug 22, 2019 at 11:47 AM Andrew Geissler <[email protected]<mailto:[email protected]>> wrote:
Quoted 4 lines · On Thu, Aug 22, 2019 at 5:01 AM Andrew MS1 Peng…
Hi Team,
When I executed command "poweroff" or "shutdown -h" in Host OS, the chassis power status seems to not be changed. Is there any package could support update power status for this situation?
Sounds like your host is not sending the inband IPMI message to the BMC to turn off the chassis once it completes the shutdown?
Quoted 5 lines
Thanks,
Andrew
Quoted 11 lines · Zheng Bao
How does the "ipmitool power on/off" do to power on/off the host? I assume it need to trigger the power button connected to BMC gpio, right? But I am not quite sure about the path which the message goes from ipmitool to gpio.
What I have got:
1. The gpio defined in machine layer, gpio_defs.json, which is handled by obmc-libobmc-intf.
2. obmc-libobmc-intf is part of "skeleton“.
The decription of skeleton on github says "Area for initial design & reference code. Will be replaced with proper implementation and design docs. " What does it mean?Am I on track? Is power on/off gpio handled by skeleton? Thanks.
The out of band and inband path goes through this code:
https://github.com/openbmc/phosphor-host-ipmid/blob/master/chassishandler.cpp#L1120
You can follow it to see how the power off occurs. But basically the code uses phosphor-state-manager interfaces to request the power off which then runs the appropriate systemd targets.
I think the power button code goes through a similar path of using phosphor-state-manager systemd targets but is not required for IPMI interfaces to work.
Quoted 34 lines
Can any big boy give some hint? Thanks.
Zheng
________________________________
From: openbmc <[email protected]> on behalf of Oskar Senft <[email protected]>
Sent: Wednesday, August 28, 2019 9:29 PM
To: Andrew Geissler <[email protected]>
Cc: Andrew MS1 Peng <[email protected]>; [email protected] <[email protected]>
Subject: Re: chassis power status seems to not be changed when type command to shutdown OSThere's also quite a bit of infrastructure on the OpenBMC side that needs to be running to be able for the BMC to actually power on / off the host.
Can you power-on / off the host using "ipmitool power on" / "off" from the BMC? If that works, then the host should be able to do the same.
Oskar.
On Thu, Aug 22, 2019 at 11:47 AM Andrew Geissler <[email protected]> wrote:
Quoted 4 lines · On Thu, Aug 22, 2019 at 5:01 AM Andrew MS1 Peng <[email protected]> wrote:
Hi Team,
When I executed command "poweroff" or "shutdown -h" in Host OS, the chassis power status seems to not be changed. Is there any package could support update power status for this situation?
Sounds like your host is not sending the inband IPMI message to the BMC to turn off the chassis once it completes the shutdown?
Quoted 5 lines
Thanks,
Andrew