scaling issue
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these normal behaviors? Or do I miss something?
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Thanks,
Hank Liou
Quanta Computer Inc.
Quoted 5 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these normal behaviors? Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
Quoted 15 lines
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Thanks,
Hank Liou
Quanta Computer Inc.
Hi Patrick,
Quoted 20 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 5 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these normal behaviors? Or do I miss something?
Quoted 4 lines
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min != conf::inheritValueFromDbus (in dbus/dbuspassive.cpp)
After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
[3]
{
"name": "fan1",
"type": "fan",
"readPath": "/sys/class/hwmon/hwmon1/fan1_input",
"writePath": "/sys/class/hwmon/hwmon1/pwm1",
"min": 0,
"max": 255
},
{
"name": "temp1",
"type": "temp",
"readPath": "/xyz/openbmc_project/sensors/temperature/temp1",
"writePath": "",
"min": 0,
"max": 0
}[4] https://github.com/openbmc/phosphor-pid-control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassive.cpp#L158
Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
Hank
Quoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 4 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
So, iirc, the min/max in the json is only for sensors that write, and not read. Sorry, I'm ramping up on a new team and slower to catch up to emails.
Yeah, the min/max in the json are for basically converting a percentage value to a PWM values -- that's its initial goal. Temperature sensors typically don't have minimum and maximum values in the code's use-cases. I added James to this because they use the daemon for other cases -- where perhaps that'll make sense.
Quoted 125 lines
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 10 lines · Patrick Venture
Quoted 4 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
So, iirc, the min/max in the json is only for sensors that write, and not read. Sorry, I'm ramping up on a new team and slower to catch up to emails.
Quoted 5 lines
Yeah, the min/max in the json are for basically converting a percentage value to a PWM values -- that's its initial goal. Temperature sensors typically don't have minimum and maximum values in the code's use-cases. I added James to this because they use the daemon for other cases -- where perhaps that'll make sense.
I believe the history of min/max being removed for passive sensors was here: https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/23470
The only other use case we have for min / max is for fan RPM so that we can create a PID based on Tach %, so that the same PID can be reused on multiple platforms with different sized fans.
-James
Quoted 126 lines
Quoted 126 lines
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 22 lines · James Feist
Quoted 10 lines · Patrick Venture
Quoted 4 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
So, iirc, the min/max in the json is only for sensors that write, and not read. Sorry, I'm ramping up on a new team and slower to catch up to emails.
Quoted 5 lines
Yeah, the min/max in the json are for basically converting a percentage value to a PWM values -- that's its initial goal. Temperature sensors typically don't have minimum and maximum values in the code's use-cases. I added James to this because they use the daemon for other cases -- where perhaps that'll make sense.
I believe the history of min/max being removed for passive sensors was here:
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/23470
Quoted 132 lines
The only other use case we have for min / max is for fan RPM so that we can create a PID based on Tach %, so that the same PID can be reused on multiple platforms with different sized fans.
-James
Quoted 126 lines
Quoted 126 lines
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Hank, do you need to use the min/max on the temperature sensors? Try your build with those entries deleted from the json file (they're optional).
James, I think I should add a warning and ignore those fields in that case. What do you think? (I honestly thought they were already ignored for temperature sensors -- have to look at the code to verify).
Quoted 156 lines · Patrick Venture
Quoted 22 lines · James Feist
Quoted 10 lines · Patrick Venture
Quoted 4 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
So, iirc, the min/max in the json is only for sensors that write, and not read. Sorry, I'm ramping up on a new team and slower to catch up to emails.
Quoted 5 lines
Yeah, the min/max in the json are for basically converting a percentage value to a PWM values -- that's its initial goal. Temperature sensors typically don't have minimum and maximum values in the code's use-cases. I added James to this because they use the daemon for other cases -- where perhaps that'll make sense.
I believe the history of min/max being removed for passive sensors was here:
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/23470Quoted 132 lines
The only other use case we have for min / max is for fan RPM so that we can create a PID based on Tach %, so that the same PID can be reused on multiple platforms with different sized fans.
-James
Quoted 126 lines
Quoted 126 lines
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 8 lines
Hank, do you need to use the min/max on the temperature sensors? Try your build with those entries deleted from the json file (they're optional).
James, I think I should add a warning and ignore those fields in that case. What do you think? (I honestly thought they were already ignored for temperature sensors -- have to look at the code to verify).
I don't have much of an opinion here as I don't use the json configuration, but more warnings never hurt.
Quoted 1 line
PTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
Quoted 129 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
[1] https://github.com/openbmc/phosphor-sel-logger/blob/3d300fca24b30864b3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59<https://g>
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issue
PTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
Quoted 129 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
As a result, fan will be at full speed while temp is low. Before the commit fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Hank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issue
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
[1] https://github.com/openbmc/phosphor-sel-logger/blob/3d300fca24b30864b3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59<https://g>
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issue
PTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
Quoted 129 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 10 lines · Hank Liou (劉晉翰)
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it? In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
If that's right, James do you have a cycle to look at this one-liner?
Quoted 171 lines
As a result, fan will be at full speed while temp is low. Before the commit fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Hank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueHi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
Quoted 129 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Sorry about the regression issue, Hank.
IIRC I proposed to have separate groups of min/max values in D-Bus and json files since they convey different semantics.
At the end the min/max value was merged in the current way, because the assumption that min/max values in json is only used for fan pwm and nothing else -- which obviously is not true now.
Should we revisit the approach? Having something like 'fanPwmScale' as a separate config param is a cleaner approach in my opinion.
Quoted 223 lines · Patrick Venture
Quoted 8 lines · Hank Liou (劉晉翰)
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as
input for temp sensor ->
Quoted 2 lines
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it? In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
If that's right, James do you have a cycle to look at this one-liner?
Quoted 3 lines
As a result, fan will be at full speed while temp is low. Before the
commit fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
Quoted 3 lines
[1]
https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Quoted 13 lines
Hank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 4 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 3 lines
[1]
Quoted 16 lines
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰) <[email protected]>
Quoted 5 lines · wrote:
Quoted 5 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Quoted 8 lines
Quoted 8 lines
Any suggestion?
Thanks,
HankQuoted 3 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of HankLiou
Quoted 20 lines
Quoted 20 lines
Quoted 20 lines
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan
input
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
to be 0 (since the input value of fan is from an integer array [2]).
Are these
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
normal behaviors?
Quoted 3 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach
your config.
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Since you're saying it was working and now isn't, I'm assuming
there's
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then
(before
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json
file (Json
Quoted 1 line
Quoted 1 line
Quoted 1 line
file [3]). However, after commit [1] the min and max values of config
would
Quoted 1 line
Quoted 1 line
Quoted 1 line
not be fed into the fan control process. The scaling issue is
originated from
Quoted 1 line
Quoted 1 line
Quoted 1 line
commit [4] with the aim to treat fan rpm as percentage. It seems that
commit
Quoted 1 line
Quoted 1 line
Quoted 1 line
[1] unexpectedly affects temp sensors in the sense that the temp is
the
Quoted 1 line
Quoted 1 line
Quoted 1 line
integer reading not percentage. Before commit [1], it would not
re-scale the
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
temp reading since my min and max are 0 [6].
There is another issue with commit [1]. Now the fan rpm would be
something
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the
fan input is 0
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 38 lines
Quoted 38 lines
Quoted 38 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Signature · 2 lines
Regards, Kun
Quoted 5 lines · Kun Yi
Sorry about the regression issue, Hank.
IIRC I proposed to have separate groups of min/max values in D-Bus and json files since they convey different semantics.
At the end the min/max value was merged in the current way, because the assumption that min/max values in json is only used for fan pwm and nothing else -- which obviously is not true now. Should we revisit the approach? Having something like 'fanPwmScale' as a separate config param is a cleaner approach in my opinion.
Kun, it was my understanding that the issue they're seeing is that they're reading min/max from dbus. But maybe I have if flipped.
I disagree that fanPwmScale is cleaner because it's now specific to fan pwm. In theory, using min/max could have some future purpose. Perhaps minWrite/maxWrite would be more ... separate. However, I think the issue seen is unrelated.
Quoted 201 lines · Patrick Venture
Quoted 13 lines · Hank Liou (劉晉翰)
Quoted 10 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
Quoted 180 lines
Ok, so you're getting a dbus-based min/max value and you want to ignore it? In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
If that's right, James do you have a cycle to look at this one-liner?
Quoted 171 lines
As a result, fan will be at full speed while temp is low. Before the commit fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Hank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueHi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
Quoted 129 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank--
Regards,
Kun
Quoted 8 lines · Patrick Venture
Quoted 7 lines · Hank Liou (劉晉翰)
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
Quoted 6 lines
Quoted 3 lines
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
What is providing the dbus-based min/max?
Quoted 5 lines
In the json configuration, if you set the values to 0,
they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 174 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 171 lines
As a result, fan will be at full speed while temp is low. Before the commit fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
[1] https://github.com/openbmc/phosphor-pid-control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
Hank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueHi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
Quoted 129 lines · Hank Liou (劉晉翰)
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 119 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank Liou (劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 11 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Hank Liou (劉晉翰)
Quoted 8 lines
Quoted 5 lines
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 6 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your config. Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit
[1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the temp reading since my min and max are 0 [6].There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input", "writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath": "/xyz/openbmc_project/sensors/temperature/temp1", "writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontrolle r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-
control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspassi ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 8 lines
-----Original Message-----
From: James Feist [mailto:[email protected]]
Sent: Tuesday, September 10, 2019 1:27 AM
To: Patrick Venture <[email protected]>; Hank Liou (劉晉翰) <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issue
Quoted 17 lines
Quoted 1 line · Patrick Venture
On Thu, Sep 5, 2019 at 12:25 AM Hank Liou (劉晉翰)
Quoted 14 lines · <[email protected]> wrote:
Quoted 11 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
Yes, we want to ignore it.
Quoted 2 lines
What is providing the dbus-based min/max?
We use dbus min/max for phosphor-sel-logger utility. The dbus min/max is provided by config file of phosphor-hwmon. In our case, they are -128/127 respectively.
I have an observation:
I found that while fan readpath is of the form "/sys/...", the fan input would not be rescaled. The same is for other sensors since the input value would not enter rescaling function. However, in our case we don't have a sys path for some thermal sensors.
Quoted 7 lines
Quoted 6 lines
In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
I agree with the option to add a check since the min/max is only for fan now.
Quoted 204 lines
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 6 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 3 lines
As a result, fan will be at full speed while temp is low. Before the commit
fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
Quoted 4 lines
Quoted 4 lines
control/commit/fc2e803f5d9256
Quoted 14 lines
Quoted 14 lines
944e18c7c878a441606b1f121cHank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 4 lines
Quoted 4 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 4 lines
Quoted 4 lines
logger/blob/3d300fca24b30864b
Quoted 17 lines
Quoted 17 lines
3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰)
Quoted 39 lines · <[email protected]> wrote:
Quoted 39 lines
Quoted 39 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 29 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank bounces+Liou
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 16 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 6 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 3 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your
config.
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 3 lines
Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the
temp reading since my min and max are 0 [6].
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
lle
r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lle
r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspass
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
i
ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 22 lines · Hank Liou (劉晉翰)
Quoted 22 lines
-----Original Message-----
From: James Feist [mailto:[email protected]]
Sent: Tuesday, September 10, 2019 1:27 AM
To: Patrick Venture <[email protected]>; Hank Liou (劉晉翰) <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Patrick Venture
On Thu, Sep 5, 2019 at 12:25 AM Hank Liou (劉晉翰)
Quoted 10 lines · <[email protected]> wrote:
Quoted 10 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
Quoted 6 lines
Quoted 4 lines
Quoted 4 lines
Quoted 1 line
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
Yes, we want to ignore it.
It looks like the sensor value scaling should be ignored for non-fans.
Per https://github.com/openbmc/phosphor-pid-control/commit/75eb769d351434547899186f73ff70ae00d7934a,
this change was only meant to deal with a fan case where the goal
looks like it was to treat the fan values as percentages on reading
the sensors, and not only on writing them by leveraging the min/max
json sometimes?
I can change the dbuspassive constructor to set the values back to their input values (ignoring dbus) if the type is not "fan", but I'm not sure that makes sense because the real use-case isn't clear to me, and I know the scaling from dbus is gone:
https://github.com/openbmc/phosphor-pid-control/search?utf8=%E2%9C%93&q=inheritValueFromDbus&type=
I'll submit the patch real quick, but I'm not sure it's the right long-term fix. I'd like to clean this up if possible so that we don't leverage "unused" features of a configuration and instead explicitly use them (or not).
Quoted 225 lines
Quoted 3 lines
What is providing the dbus-based min/max?
We use dbus min/max for phosphor-sel-logger utility. The dbus min/max is provided by config file of phosphor-hwmon. In our case, they are -128/127 respectively.
I have an observation:
I found that while fan readpath is of the form "/sys/...", the fan input would not be rescaled. The same is for other sensors since the input value would not enter rescaling function. However, in our case we don't have a sys path for some thermal sensors.Quoted 7 lines
Quoted 6 lines
In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
I agree with the option to add a check since the min/max is only for fan now.
Quoted 204 lines
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 6 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 3 lines
As a result, fan will be at full speed while temp is low. Before the commit
fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
Quoted 4 lines
Quoted 4 lines
control/commit/fc2e803f5d9256
Quoted 14 lines
Quoted 14 lines
944e18c7c878a441606b1f121cHank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 4 lines
Quoted 4 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 4 lines
Quoted 4 lines
logger/blob/3d300fca24b30864b
Quoted 17 lines
Quoted 17 lines
3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰)
Quoted 39 lines · <[email protected]> wrote:
Quoted 39 lines
Quoted 39 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 29 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank bounces+Liou
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 16 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 6 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 3 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your
config.
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 3 lines
Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the
temp reading since my min and max are 0 [6].
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
lle
r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lle
r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspass
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
i
ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 44 lines · Patrick Venture
Quoted 22 lines · Hank Liou (劉晉翰)
Quoted 22 lines
-----Original Message-----
From: James Feist [mailto:[email protected]]
Sent: Tuesday, September 10, 2019 1:27 AM
To: Patrick Venture <[email protected]>; Hank Liou (劉晉翰) <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Patrick Venture
On Thu, Sep 5, 2019 at 12:25 AM Hank Liou (劉晉翰)
Quoted 10 lines · <[email protected]> wrote:
Quoted 10 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
Quoted 6 lines
Quoted 4 lines
Quoted 4 lines
Quoted 1 line
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
Yes, we want to ignore it.
It looks like the sensor value scaling should be ignored for non-fans.
Per https://github.com/openbmc/phosphor-pid-control/commit/75eb769d351434547899186f73ff70ae00d7934a,
this change was only meant to deal with a fan case where the goal
looks like it was to treat the fan values as percentages on reading
the sensors, and not only on writing them by leveraging the min/max
json sometimes?I can change the dbuspassive constructor to set the values back to their input values (ignoring dbus) if the type is not "fan", but I'm not sure that makes sense because the real use-case isn't clear to me, and I know the scaling from dbus is gone:
https://github.com/openbmc/phosphor-pid-control/search?utf8=%E2%9C%93&q=inheritValueFromDbus&type=
Quoted 4 lines
I'll submit the patch real quick, but I'm not sure it's the right long-term fix. I'd like to clean this up if possible so that we don't leverage "unused" features of a configuration and instead explicitly use them (or not).
This patch should address what you're seeing. Overall, I think there should be a larger change addressing the goal of the original change.
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/25070
Quoted 226 lines
Quoted 226 lines
Quoted 4 lines
What is providing the dbus-based min/max?
We use dbus min/max for phosphor-sel-logger utility. The dbus min/max is provided by config file of phosphor-hwmon. In our case, they are -128/127 respectively.
I have an observation:
I found that while fan readpath is of the form "/sys/...", the fan input would not be rescaled. The same is for other sensors since the input value would not enter rescaling function. However, in our case we don't have a sys path for some thermal sensors.Quoted 7 lines
Quoted 6 lines
In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
I agree with the option to add a check since the min/max is only for fan now.
Quoted 204 lines
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 6 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 3 lines
As a result, fan will be at full speed while temp is low. Before the commit
fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
Quoted 4 lines
Quoted 4 lines
control/commit/fc2e803f5d9256
Quoted 14 lines
Quoted 14 lines
944e18c7c878a441606b1f121cHank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 4 lines
Quoted 4 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 4 lines
Quoted 4 lines
logger/blob/3d300fca24b30864b
Quoted 17 lines
Quoted 17 lines
3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰)
Quoted 39 lines · <[email protected]> wrote:
Quoted 39 lines
Quoted 39 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 29 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank bounces+Liou
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 16 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 6 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 3 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your
config.
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 3 lines
Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the
temp reading since my min and max are 0 [6].
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
lle
r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lle
r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspass
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
i
ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 51 lines · Patrick Venture
Quoted 44 lines · Patrick Venture
Quoted 22 lines · Hank Liou (劉晉翰)
Quoted 22 lines
-----Original Message-----
From: James Feist [mailto:[email protected]]
Sent: Tuesday, September 10, 2019 1:27 AM
To: Patrick Venture <[email protected]>; Hank Liou (劉晉翰) <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 1 line · Patrick Venture
On Thu, Sep 5, 2019 at 12:25 AM Hank Liou (劉晉翰)
Quoted 10 lines · <[email protected]> wrote:
Quoted 10 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following issue:
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
Quoted 6 lines
Quoted 4 lines
Quoted 4 lines
Quoted 1 line
stepwise fan table output 100 duty -> full speed fan
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
Yes, we want to ignore it.
It looks like the sensor value scaling should be ignored for non-fans.
Per https://github.com/openbmc/phosphor-pid-control/commit/75eb769d351434547899186f73ff70ae00d7934a,
this change was only meant to deal with a fan case where the goal
looks like it was to treat the fan values as percentages on reading
the sensors, and not only on writing them by leveraging the min/max
json sometimes?I can change the dbuspassive constructor to set the values back to their input values (ignoring dbus) if the type is not "fan", but I'm not sure that makes sense because the real use-case isn't clear to me, and I know the scaling from dbus is gone:
https://github.com/openbmc/phosphor-pid-control/search?utf8=%E2%9C%93&q=inheritValueFromDbus&type=Quoted 4 lines
I'll submit the patch real quick, but I'm not sure it's the right long-term fix. I'd like to clean this up if possible so that we don't leverage "unused" features of a configuration and instead explicitly use them (or not).
Quoted 4 lines
This patch should address what you're seeing. Overall, I think there should be a larger change addressing the goal of the original change.
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/25070
I decided to take on the larger change because chances are we might hit it some other way:
https://gerrit.openbmc-project.xyz/c/25072/
This lets you specify a json configuration option that will ignore the MinValue/MaxValue properties on dbus.
Quoted 227 lines
Quoted 227 lines
Quoted 227 lines
Quoted 5 lines
What is providing the dbus-based min/max?
We use dbus min/max for phosphor-sel-logger utility. The dbus min/max is provided by config file of phosphor-hwmon. In our case, they are -128/127 respectively.
I have an observation:
I found that while fan readpath is of the form "/sys/...", the fan input would not be rescaled. The same is for other sensors since the input value would not enter rescaling function. However, in our case we don't have a sys path for some thermal sensors.Quoted 7 lines
Quoted 6 lines
In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
I agree with the option to add a check since the min/max is only for fan now.
Quoted 204 lines
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 6 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 3 lines
As a result, fan will be at full speed while temp is low. Before the commit
fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger scaling function.
Quoted 4 lines
Quoted 4 lines
control/commit/fc2e803f5d9256
Quoted 14 lines
Quoted 14 lines
944e18c7c878a441606b1f121cHank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 4 lines
Quoted 4 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 4 lines
Quoted 4 lines
logger/blob/3d300fca24b30864b
Quoted 17 lines
Quoted 17 lines
3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issuePTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into openbmc/openbmc in a day or two.
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰)
Quoted 39 lines · <[email protected]> wrote:
Quoted 39 lines
Quoted 39 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 29 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf Of Hank bounces+Liou
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueHi Patrick,
Quoted 16 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 6 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
normal behaviors?
Quoted 3 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your
config.
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 3 lines
Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
info->min!=conf::inheritValueFromDbus(in dbus/dbuspassive.cpp)Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan control process. The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the
temp reading since my min and max are 0 [6].
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see [5]). Hence the fan input is 0 not percentage. Is there something wrong?
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
lle
r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancontro
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lle
r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspass
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
i
ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Hi Patrick,
Thanks for your help. This change does solve the issue.
Hank
Quoted 328 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, September 11, 2019 1:52 AM
To: Hank Liou (劉晉翰) <[email protected]>
Cc: James Feist <[email protected]>; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 2 lines · Patrick Venture
On Tue, Sep 10, 2019 at 7:39 AM Patrick Venture <[email protected]>
Quoted 2 lines · wrote:
Quoted 2 lines
On Tue, Sep 10, 2019 at 1:02 AM Hank Liou (劉晉翰)
Quoted 9 lines · <[email protected]> wrote:
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
-----Original Message-----
From: James Feist [mailto:[email protected]]
Sent: Tuesday, September 10, 2019 1:27 AM
To: Patrick Venture <[email protected]>; Hank Liou (劉晉翰) <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines · Patrick Venture
Quoted 1 line
On Thu, Sep 5, 2019 at 12:25 AM Hank Liou (劉晉翰)
Quoted 5 lines · <[email protected]> wrote:
Quoted 5 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following
issue:
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
Yes, we want to ignore it.
It looks like the sensor value scaling should be ignored for non-fans. Per
https://github.com/openbmc/phosphor-pid-control/commit/75eb769d35143
Quoted 10 lines
Quoted 10 lines
4547899186f73ff70ae00d7934a, this change was only meant to deal with a fan case where the goal looks like it was to treat the fan values as percentages on reading the sensors, and not only on writing them by leveraging the min/max json sometimes?I can change the dbuspassive constructor to set the values back to their input values (ignoring dbus) if the type is not "fan", but I'm not sure that makes sense because the real use-case isn't clear to me, and I know the scaling from dbus is gone:
https://github.com/openbmc/phosphor-pid-control/search?utf8=%E2%9C%9
Quoted 7 lines
Quoted 6 lines
3&q=inheritValueFromDbus&type=
I'll submit the patch real quick, but I'm not sure it's the right long-term fix. I'd like to clean this up if possible so that we don't leverage "unused" features of a configuration and instead explicitly use them (or not).
Quoted 5 lines
This patch should address what you're seeing. Overall, I think there should be a larger change addressing the goal of the original change.
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/25
070I decided to take on the larger change because chances are we might hit it some other way:
https://gerrit.openbmc-project.xyz/c/25072/This lets you specify a json configuration option that will ignore the MinValue/MaxValue properties on dbus.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 5 lines
What is providing the dbus-based min/max?
We use dbus min/max for phosphor-sel-logger utility. The dbus min/max
is provided by config file of phosphor-hwmon.
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
In our case, they are -128/127 respectively.
I have an observation:
I found that while fan readpath is of the form "/sys/...", the fan inputwould not be rescaled. The same is for other sensors since the input value would not enter rescaling function.
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
However, in our case we don't have a sys path for some thermal sensors.
Quoted 7 lines
Quoted 6 lines
In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
I agree with the option to add a check since the min/max is only for fan
now.
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 5 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 2 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines
As a result, fan will be at full speed while temp is low. Before the commit
fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger
scaling function.
Quoted 47 lines
Quoted 47 lines
Quoted 47 lines
Quoted 47 lines
Quoted 4 lines
Quoted 4 lines
control/commit/fc2e803f5d9256
Quoted 14 lines
Quoted 14 lines
944e18c7c878a441606b1f121cHank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 5 lines
Quoted 5 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 4 lines
Quoted 4 lines
logger/blob/3d300fca24b30864b
Quoted 13 lines
Quoted 13 lines
3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
PTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into
openbmc/openbmc in a day or two.
Quoted 24 lines
Quoted 24 lines
Quoted 24 lines
Quoted 24 lines
Quoted 2 lines
Quoted 2 lines
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰)
Quoted 21 lines · <[email protected]> wrote:
Quoted 21 lines
Quoted 21 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 11 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf bounces+Of Hank Liou
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueQuoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Hi Patrick,
Quoted 4 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; JamesFeist
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
<[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 6 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
normal behaviors?
Quoted 4 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your
config.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
info->min!=conf::inheritValueFromDbus(in
info->dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan
control process.
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the
temp reading since my min and max are 0 [6].
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see
[5]).
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Hence the fan input is 0 not percentage. Is there something
wrong?
Quoted 70 lines
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancont
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
ro
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
lle
r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancont
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
ro
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lle
r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspa
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
ss
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
i
ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank
Quoted 4 lines · Hank Liou (劉晉翰)
Hi Patrick,
Thanks for your help. This change does solve the issue.
No problem! I enjoy it when someone hits an edge case or a situation not yet considered. We don't have nearly enough variety of configurations to really test quirks and unit-testing only covers so much.
I saw your review requesting the configuration change, and +1'd it. Presumably you're well on your way.
Thanks for your patience in this back-and-forth to resolve the issue.
Quoted 331 lines
Hank
Quoted 328 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, September 11, 2019 1:52 AM
To: Hank Liou (劉晉翰) <[email protected]>
Cc: James Feist <[email protected]>; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 2 lines · Patrick Venture
On Tue, Sep 10, 2019 at 7:39 AM Patrick Venture <[email protected]>
Quoted 2 lines · wrote:
Quoted 2 lines
On Tue, Sep 10, 2019 at 1:02 AM Hank Liou (劉晉翰)
Quoted 9 lines · <[email protected]> wrote:
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
-----Original Message-----
From: James Feist [mailto:[email protected]]
Sent: Tuesday, September 10, 2019 1:27 AM
To: Patrick Venture <[email protected]>; Hank Liou (劉晉翰) <[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines · Patrick Venture
Quoted 1 line
On Thu, Sep 5, 2019 at 12:25 AM Hank Liou (劉晉翰)
Quoted 5 lines · <[email protected]> wrote:
Quoted 5 lines
Hi Patrick,
Sorry for not clearly stating our problem. We have the following
issue:
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
temp sensor gets 31(C) -> 31 / 0.255 = 121.57 -> pid uses 127.57 as input for temp sensor ->
stepwise fan table output 100 duty -> full speed fan
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
Ok, so you're getting a dbus-based min/max value and you want to ignore it?
Yes, we want to ignore it.
It looks like the sensor value scaling should be ignored for non-fans. Per
https://github.com/openbmc/phosphor-pid-control/commit/75eb769d35143
Quoted 10 lines
Quoted 10 lines
4547899186f73ff70ae00d7934a, this change was only meant to deal with a fan case where the goal looks like it was to treat the fan values as percentages on reading the sensors, and not only on writing them by leveraging the min/max json sometimes?I can change the dbuspassive constructor to set the values back to their input values (ignoring dbus) if the type is not "fan", but I'm not sure that makes sense because the real use-case isn't clear to me, and I know the scaling from dbus is gone:
https://github.com/openbmc/phosphor-pid-control/search?utf8=%E2%9C%9
Quoted 7 lines
Quoted 6 lines
3&q=inheritValueFromDbus&type=
I'll submit the patch real quick, but I'm not sure it's the right long-term fix. I'd like to clean this up if possible so that we don't leverage "unused" features of a configuration and instead explicitly use them (or not).
Quoted 5 lines
This patch should address what you're seeing. Overall, I think there should be a larger change addressing the goal of the original change.
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-pid-control/+/25
070I decided to take on the larger change because chances are we might hit it some other way:
https://gerrit.openbmc-project.xyz/c/25072/This lets you specify a json configuration option that will ignore the MinValue/MaxValue properties on dbus.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 5 lines
What is providing the dbus-based min/max?
We use dbus min/max for phosphor-sel-logger utility. The dbus min/max
is provided by config file of phosphor-hwmon.
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
In our case, they are -128/127 respectively.
I have an observation:
I found that while fan readpath is of the form "/sys/...", the fan inputwould not be rescaled. The same is for other sensors since the input value would not enter rescaling function.
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
However, in our case we don't have a sys path for some thermal sensors.
Quoted 7 lines
Quoted 6 lines
In the json configuration, if you set the values to 0, they are set to the default (0), so there'd be no clean way to know to ignore dbus in this case, without adding a small check to only consider min/max from dbus when sensor is not temp/margin. Basically, only care about min/max on dbus if type is "fan"
I agree with the option to add a check since the min/max is only for fan
now.
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Why is the dbus-based min/max even there if you don't plan on using it?
Quoted 5 lines
If that's right, James do you have a cycle to look at this one-liner?
Quoted 2 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines
As a result, fan will be at full speed while temp is low. Before the commit
fc2e803 [1], this won't happen. The root cause is that, before fc2e803, pid will use config min/max, which is 0 in our case. This would not trigger scaling function, namely scaleSensorReading, in util.cpp. However, after such commit, min/max would be non-zero (-128/127 from DBus). This will trigger
scaling function.
Quoted 47 lines
Quoted 47 lines
Quoted 47 lines
Quoted 47 lines
Quoted 4 lines
Quoted 4 lines
control/commit/fc2e803f5d9256
Quoted 14 lines
Quoted 14 lines
944e18c7c878a441606b1f121cHank
________________________________
From: Hank Liou (劉晉翰)
Sent: Monday, September 2, 2019 4:52 PM
To: Patrick Venture
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 5 lines
Quoted 5 lines
Hi Patrick,
Since we use phosphor-sel-logger [1] at the same time, we do assign min
and max of temp sensors to Dbus (max: 127, min: -128). So in the present case, our temp value will be divided by 0.255 (also due to exponent is -3 here). This will cause re-scaling problem. Therefore there should be an statement to distinguish sensor types. If it is "temp", then one assigns 0 to the min and max from Dbus.
Quoted 4 lines
Quoted 4 lines
logger/blob/3d300fca24b30864b
Quoted 13 lines
Quoted 13 lines
3e9a4f5768cfe5e769458ae/include/sensorutils.hpp#L59
Hank
________________________________
From: Patrick Venture <[email protected]>
Sent: Friday, August 30, 2019 1:47 AM
To: Hank Liou (劉晉翰)
Cc: James Feist; [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
PTAL - https://gerrit.openbmc-project.xyz/24827 - this is merged, and the srcrev bump should propagate into
openbmc/openbmc in a day or two.
Quoted 24 lines
Quoted 24 lines
Quoted 24 lines
Quoted 24 lines
Quoted 2 lines
Quoted 2 lines
On Wed, Aug 28, 2019 at 11:00 PM Hank Liou (劉晉翰)
Quoted 21 lines · <[email protected]> wrote:
Quoted 21 lines
Quoted 21 lines
Hi Patrick,
I think it's OK to parse the config min&max for temp sensors.
Any suggestion?
Thanks,
HankQuoted 11 lines
-----Original Message-----
From: openbmc [mailto:openbmc- [email protected]] On Behalf bounces+Of Hank Liou
(劉晉翰)
Sent: Friday, August 23, 2019 4:31 PM
To: Patrick Venture <[email protected]>; James Feist <[email protected]>
Cc: [email protected]
Subject: RE: [phosphor-pid-control] scaling issueQuoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Hi Patrick,
Quoted 4 lines
-----Original Message-----
From: Patrick Venture [mailto:[email protected]]
Sent: Wednesday, August 21, 2019 10:32 PM
To: Hank Liou (劉晉翰) <[email protected]>; JamesFeist
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
<[email protected]>
Cc: [email protected]
Subject: Re: [phosphor-pid-control] scaling issueQuoted 6 lines · Hank Liou (劉晉翰)
Hi All,
After commit [1], I found my temp sensor reading would be re-scaled by
multiplying 1 over 255, making temperature into unfamiliar unit. Also the fan rpm reading would lie in [0,1] interval, letting the fan input to be 0 (since the input value of fan is from an integer array [2]). Are these
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
normal behaviors?
Quoted 4 lines
Or do I miss something?
Are you using dbus configuration or json? If json, can you attach your
config.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Since you're saying it was working and now isn't, I'm assuming there's something about the config being treated differently with the code changes in an unexpected way.
I found pid control will first read min and max from dbus and then (before commit [1]) revise them if
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
info->min!=conf::inheritValueFromDbus(in
info->dbus/dbuspassive.cpp)After value initialization, the min and max would be the ones in json file (Json file [3]). However, after commit [1] the min and max values of config would not be fed into the fan
control process.
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
The scaling issue is originated from commit [4] with the aim to treat fan rpm as percentage. It seems that commit [1] unexpectedly affects temp sensors in the sense that the temp is the integer reading not percentage. Before commit [1], it would not re-scale the
temp reading since my min and max are 0 [6].
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
There is another issue with commit [1]. Now the fan rpm would be something like
1500 / 20000 = 0.075
where rpm max 20000 is from dbus. However the fan input function would transfer double into int, which is 0 for 0.075 (see
[5]).
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Hence the fan input is 0 not percentage. Is there something
wrong?
Quoted 70 lines
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
[1] https://github.com/openbmc/phosphor-pid-
control/commit/fc2e803f5d9256944e18c7c878a441606b1f121c
[2] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancont
Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
ro
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
lle
r.cpp#L86
[3]{ "name": "fan1", "type": "fan", "readPath": "/sys/class/hwmon/hwmon1/fan1_input",Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "/sys/class/hwmon/hwmon1/pwm1", "min": 0, "max": 255 }, { "name": "temp1", "type": "temp", "readPath":"/xyz/openbmc_project/sensors/temperature/temp1",
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
"writePath": "", "min": 0, "max": 0 }[4] https://github.com/openbmc/phosphor-pid-
control/commit/75eb769d351434547899186f73ff70ae00d7934a
[5] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/pid/fancont
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
ro
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lle
r.cpp#L64
[6] https://github.com/openbmc/phosphor-pid-control/blob/a7ec8350d17b70153cebe666d3fbe88bddd02a1a/dbus/dbuspa
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
ss
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
Quoted 28 lines
i
ve.cpp#L158Quoted 22 lines
Quoted 5 lines
control/commit/fc2e803f5d92569
Quoted 4 lines
44e18c7c878a441606b1f121ccontrol/blob/a7ec8350d17b70153
Quoted 11 lines
cebe666d3fbe88bddd02a1a/pid/fancontroller.cpp#L86
Thanks,
Hank Liou
Quanta Computer Inc.
Sincerely,
Hank