Fan PWM settings via Redfish
Bruce Lee (李昀峻)
Bruce_Lee at quantatw.com
Wed Mar 17 21:17:03 AEDT 2021
> -----Original Message-----
> From: Ed Tanous <edtanous at google.com>
> Sent: Tuesday, March 16, 2021 11:18 PM
> To: Bruce Lee (李昀峻) <Bruce_Lee at quantatw.com>
> Cc: Nan Zhou <nanzhou at google.com>; rhanley at google.com;
> openbmc at lists.ozlabs.org
> Subject: Re: Fan PWM settings via Redfish
>
> On Tue, Mar 16, 2021 at 2:35 AM Bruce Lee (李昀峻)
> <Bruce_Lee at quantatw.com> wrote:
> >
> >
> >
> > -----Original Message-----
> > From: Ed Tanous <edtanous at google.com>
> > Sent: Saturday, March 13, 2021 1:40 AM
> > To: Bruce Lee (李昀峻) <Bruce_Lee at quantatw.com>
> > Cc: Nan Zhou <nanzhou at google.com>; rhanley at google.com;
> > openbmc at lists.ozlabs.org
> > Subject: Re: Fan PWM settings via Redfish
> >
> > On Thu, Mar 11, 2021 at 10:37 PM Bruce Lee (李昀峻)
> <Bruce_Lee at quantatw.com> wrote:
> > >
> > > Hi All,
> > >
> > >
> > >
> > > We are designing and implementing the Fan PWM settings via Redfish. The
> goal is that clients can set sensor value to bmc via Redfish.
> > >
> > >
> > >
> > > We divide the work into three phases.
> > >
> > >
> > >
> > > Phase 1 is to remove the definition
> “BMCWEB_INSECURE_UNRESTRICTED_SENSOR_OVERRIDE” and use new
> definition to “BMCWEB_SPECIAL_MODE_SENSOR_OVERRIDE”.
> > >
> > > The “BMCWEB_INSECURE_UNRESTRICTED_SENSOR_OVERRIDE” was added
> by
> > > Intel group, please refer to
> > > https://apc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fge
> > > rr
> > >
> it.openbmc-project.xyz%2Fc%2Fopenbmc%2Fbmcweb%2F%2B%2F30000&am
> p;data
> > > =0
> > >
> 4%7C01%7CBruce_Lee%40quantatw.com%7C64a1153cd45b46eeca4008d8e5
> 7df35c
> > > %7
> > >
> C179b032707fc4973ac738de7313561b2%7C1%7C0%7C63751167640422711
> 3%7CUnk
> > > no
> > >
> wn%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1ha
> Ww
> > > iL
> > >
> CJXVCI6Mn0%3D%7C1000&sdata=f604Piz1vDfItDZ3docZOPfryJesavkbOw
> hKy
> > > wJ
> > > oXlU%3D&reserved=0,
> > >
> > > The Intel solution has 4 conditions needs to match one of them and that can
> be work to override sensor but actually not all project needs those conditions, so
> we want to propose to remove the insecure definition and use new definition to
> include the intel solution and execute when compile. It would be no compile time
> with option for common project. And the insecure issue we will discuss in phase
> 2.
> > >
> > >
> > >
> > > Example below:
> > >
> > > --------------------------------------------------------------------
> > > --
> > > -------------------------------
> > >
> > > [Before modified]
> > >
> > > #ifdef BMCWEB_INSECURE_UNRESTRICTED_SENSOR_OVERRIDE
> > >
> > > // Proceed with sensor override
> > >
> > > setSensorsOverride(sensorAsyncResp, allCollections);
> > >
> > > return;
> > >
> > > #endif
> > >
> > > doIntelSpecialModeManager code …
> > >
> > > --------------------------------------------------------------------
> > > --
> > > -------------------------------
> > >
> > > [After modified]
> > >
> > > #ifdef BMCWEB_SPECIAL_MODE_SENSOR_OVERRIDE
> > >
> > > doIntelSpecialModeManager code …
> > >
> > > return;
> > >
> > > #endif
> > >
> > > //Proceed with sensor override
> > >
> > > setSensorsOverride(sensorAsyncResp, allCollections);
> > >
> > > --------------------------------------------------------------------
> > > --
> > > -------------------------------
> > >
> > >
> > >
> > >
> >
> > >I suspect this check and option needs to be moved into the individual sensors,
> so that we can differentiate between "should be settable in a test context" and
> "should be settable in a normal context".
> > 1. Does you mean don't change the Intel definition and keep the origin code
> when compile time?
>
> No, this means that the checking code needs to move from redfish into
> dbus-sensors.
>
> > 2. What do you mean this option needs to be moved into the individual sensors
> so that we can differentiate between "should be settable in a test context" and
> "should be settable in a normal context".
> > Please provide more details about your thinking.
>
> Individual sensors need to provide an appropriate dbus interface.
> Part of that is enforcing whether or not they're writable, and checking for the
> debug state of the system to do so.
>
> >
> >
> > >
> > > Phase 2 is to add a condition to check the sensor name’s Mutable value of
> EM if the value is true do the sensor override function else not do.
> >
> > >I suspect this patchset needs to be moved forward if you're hoping to use the
> mutable param:
> > >https://apc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fger
> >
> >rit.openbmc-project.xyz%2Fc%2Fopenbmc%2Fphosphor-dbus-interfaces%2F%
> 2
> >
> >B%2F36333&data=04%7C01%25>7CBruce_Lee%40quantatw.com%7C64
> a1153cd4
> >
> >5b46eeca4008d8e57df35c%7C179b032707fc4973ac738de7313561b2%7C1
> %7C0%7C6
> >
> >37511676404227113%7CUnknown%>7CTWFpbGZsb3d8eyJWIjoiMC4wLjAw
> MDAiLCJQIj
> >
> >oiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=tdExxB%
> 2BY7
> > >O1cKb%2FYMdvPGnw7YThW7J55jytDPh4YWYo%3D&reserved=0
>
> Not quite, but close. I wouldn't expect the configurability to be directly
> configurable. External sensor types should be mutable, all other types should
> not be mutable (except in a debug context). I don't think there's any reason to
> add a separate "IsMutable" parameter into the EM json, unless it really needs to
> be configurable per sensor, which I don't think is the case.
>
> >
> > >
> > > The Mutable value can be set in the sensor configuration of Entity-Manage,
> when using the patch command to override the sensor, it needs to check the
> EntityManager subtree’s sensor name and its interface
> “xyz.openbmc_project.Configuration.I2CFan.Connector” to check the
> corresponding property name’s mutable value to decide whether executing the
> override function.
> >
> > >See above. I suspect that the redfish code doesn't need to check the
> mutability of the sensor, the interface should just have the correct behavior.
> The only place I would expect to need to know the >mutability of a sensor is in
> the IPMI sdr, where we will need to set the modifiable bit appropriately.
> >
> > For now, the function to set sensor in redfish code is to set the d-bus value
> directly (internally writable), if we don't check the EM mutability in Redfish,
> follow the Add Mutable property to Sensor Value interface, we still need to check
> the sensor mutable property to know whether or not to write the d-bus value in
> redfish or we need other external services to know whether or not to grant write
> permission to their users like IPMI sensor.
>
> I'm not really following this. My point is that the only thing that really needs to
> "check" the mutability requirement is dbus-sensors.
> They should only allow setting when sensors are mutable, and reject when
> they're not.
>
IPMI has an external service to check the Mutability and allow setting when it is "Write" and reject when it's "not write".
Today if we add a mutable property in the d-bus sensor, we also need an external-service like IPMI sensor-handler to check the mutable value to grant write access or not.
You stated that "They should only allow setting when sensors are mutable and reject when they're not." are means an external-service as I mention above?
> >
> > >
> > > This achieves feature parity with the ipmi::sensor::Mutability
> > > parameter of the old hardcoded YAML configuration files
> >
> > >Not sure what you're referring to. That may have been something done in a
> fork.
> >
> > >
> > >
> > >
> > > Execute steps:
> > >
> > > 1. Patch command to override sensor.
> > >
> > > 2. Check the EM of sensor’s Mutable value
> > >
> > > 3. If Mutable value is true do sensor override action else not do.
> > >
> > >
> > >
> > >
> > >
> > > Phase 3 is to add a new get command to get the Zone_$id’s "Manual" value
> and patch command to change the fan mode from auto to manual mode
> ("Manual":true).
> > >
> > > Because the fan control is use package phosphor-pid-control, when we need
> to set fan pwm, it needs to set the fan mode from auto mode to manual mode,
> for now, the phosphor-pid-control has already provided ipmi-oem command to
> achieve this feature, so we need to implement this fan mode change via redfish
> command.
> >
> > >Doesn't this already work today? I thought we had all that sorted a long
> time ago. For some reason I thought we intentionally didn't expose the
> manual/automatic param, because that only applied to >the PID loops, and
> PWM sensor didn't expose that interface. I need to go look at the code at some
> point.
> >
> > Yes, ipmi-oem is work today. I agree it is not properly to show on redfish to let
> users can easily change the fan mode, the reason to change fan mode to the
> manual is for debugging. Maybe let users use ipmi-oem to replace show on
> Redfish URLs.
> >
> > >
> > >
> > >
> > > Example URLs |Method |Example
> Payload
> > >
> > > --------------------------------------- |-------------- |--
> > >
> > > /redfish/v1/Managers/bmc |GET |"Oem": {
> > >
> > > |
> | Fan": {
> > >
> > > |
> | "FanZones": {
> > >
> > > |
> | "@odata.id":
> "/redfish/v1/Managers/bmc#/Oem/OpenBmc/Fan/FanZones",
> > >
> > > |
> | "@odata.type": "#OemManager.FanZones",
> > >
> > > |
> | "Zone_0": {
> > >
> > > |
> | "@odata.id":
> "/redfish/v1/Managers/bmc#/Oem/OpenBmc/Fan/FanZones/Zone_0",
> > >
> > > |
> | "@odata.type":
> "#OemManager.FanZone",
> > >
> > > |
> | "Chassis": {
> > >
> > > |
> | "@odata.id":
> "/redfish/v1/Chassis/GSZ_EVT"
> > >
> > > |
> | },
> > >
> > > |
> | "FailSafePercent": 100.0,
> > >
> > > |
> | "MinThermalOutput": 0.0,
> > >
> > > |
> | "ZoneIndex": 0.0,
> > >
> > > |
> | "Manual":false
> > >
> > > |
> | },
> > >
> > > |
> | },
> > >
> > > |
> | },
> > >
> > > |
> |}
> > >
> > > --------------------------------------- |-------------- |----
> > >
> > > /redfish/v1/Managers/bmc | PATCH |"Oem": { "Fan":
> { "FanZones": { "Zone_0": { "Manual":true } } }
> > >
> > >
> >
> > >It should be noted, this schema needs some serious cleanup to make it proper
> resources, paths, and collections, and should version the schema files
> appropriately. If you're planning on extending it, I ?>would expect _some_
> effort to be put into cleanup. There's several github bugs that have more details,
> and I will leave it up to you to decide how much you'd like to do as part of this
> work, but please >plan on some.
> >
> > >
> > > If any thoughts on this topic, feel free to give your comments. Thanks!
> > >
> > >
> > >
> > > Sincerely,
> > >
> > > Bruce
More information about the openbmc
mailing list