|
From: Matthias H. <mat...@tu...> - 2016-04-09 17:50:51
|
Hi, I discovered that the thinkpad-acpi module does not recognize the second fan on my Thinkpad P50. I discovered some support for a second fan in the source code, so I hard-coded the 'second_fan' bit to 1 an added some functions to expose a second PWM interface. This way I managed to control both fans individually. As I mentioned, this code is just a hack and what's totally missing is some detection of the second fan. I would appreciate some help there. Here is the patch: http://pastebin.com/ymDwN4hj Regards, Matthias Hochsteger PS: The P50 also has HKEY version 0x200. I could do some testing there if needed. |
|
From: Henrique de M. H. <hm...@hm...> - 2016-04-09 18:59:13
|
On Sat, 09 Apr 2016, Matthias Hochsteger wrote: > This way I managed to control both fans individually. As I mentioned, > this code is just a hack and what's totally missing is some detection of > the second fan. I would appreciate some help there. Two-fan thinkpads are rare enough that we can use a quirk list for now. This thing needs to be whitelist-based anyway, as it is a direct-write-to-EC feature, so we can only enable it by default on tested models. >From your patch, I understand the select_fan() thing we already have for the X60/X61 actually changes which fan is being controlled on the P50? I wonder if it actually worked that way on the X60/X61... I don't recall if we tested that at the time. Is anyone with a X60 or X61 around that could test it? Otherwise, we can simply add a new quirk for second-fan-control-ok, and enable second-fan control only for the P50, and not for the X60/X61. > PS: The P50 also has HKEY version 0x200. I could do some testing there if > needed. Now we just need someone with the original adaptive keyboard (I understand the P50 doesn't have it) to help, but let's keep HKEY 0x200-related stuff on its own thread. -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh |
|
From: Matthias H. <mat...@tu...> - 2016-04-10 15:01:59
|
Here is the latest version of my patch. I added quirks to recognize the second fan and cleaned the code, any comments are welcome: http://pastebin.com/eVPcnvfK > From your patch, I understand the select_fan() thing we already have for the > X60/X61 actually changes which fan is being controlled on the P50? That is correct, the same mechanism/bit is used to select the active fan. But there is some strange behavior when reading the status register. Regardless of which fan is currently selected, the status register always reads the status of fan1. This means I can set the fan speed levels individually, but reading pwm2_enable or pwm2 always returns the corresponding value of fan1. > Otherwise, we can simply add a new quirk for second-fan-control-ok, and > enable second-fan control only for the P50, and not for the X60/X61. I added a quirk based on 'bios_model' (first two characters of BIOS version), since the EC version is not read correctly by the module. The Lenovo P70, which is very similar to the P50, shares these first two characters 'N1' according to Lenovo's release notes, so it should be automatically detected too. Some testing there is needed. https://download.lenovo.com/pccbbs/mobiles/n1duj12w.txt Regards, Matthias |
|
From: Henrique de M. H. <hm...@hm...> - 2016-04-10 19:07:23
|
On Sun, 10 Apr 2016, Matthias Hochsteger wrote: > Here is the latest version of my patch. I added quirks to recognize the > second fan and cleaned the code, any comments are welcome: > http://pastebin.com/eVPcnvfK > > > From your patch, I understand the select_fan() thing we already have for the > > X60/X61 actually changes which fan is being controlled on the P50? > > That is correct, the same mechanism/bit is used to select the active > fan. But there is some strange behavior when reading the status > register. Regardless of which fan is currently selected, the status > register always reads the status of fan1. This means I can set the fan > speed levels individually, but reading pwm2_enable or pwm2 always returns > the corresponding value of fan1. So, we'd have to implement them in software through a quirk, sort of like the T43 fan quirk, only worse. We have to assume the initial state (like the T43 quirk). Unlike the t43 quirk, we have to track the states we write, and assume nobody else will change it behind our backs. It is probably safe to assume fan2 defaults to "auto" mode like fan1 by the time the operating system is started. However, please check if there are any related BIOS toggles... > > Otherwise, we can simply add a new quirk for second-fan-control-ok, and > > enable second-fan control only for the P50, and not for the X60/X61. > > I added a quirk based on 'bios_model' (first two characters of BIOS > version), since the EC version is not read correctly by the module. The EC version is not present in dmidecode output anymore, is it? Because we CAN get it out of the EC register dump directly (the BIOS and the windows drivers certainly manage to do so somehow), but DMI data is *much* preferred. > The Lenovo P70, which is very similar to the P50, shares these first > two characters 'N1' according to Lenovo's release notes, so it should > be automatically detected too. Some testing there is needed. > https://download.lenovo.com/pccbbs/mobiles/n1duj12w.txt If it has the same EC firmware, it will not misbehave if we try to mess with fan2. But if it doesn't have a fan2, it would be nice to not expose it. -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh |
|
From: Matthias H. <mat...@tu...> - 2016-04-20 21:42:39
Attachments:
signature.asc
|
> On Sun, 10 Apr 2016, Matthias Hochsteger wrote: > > Here is the latest version of my patch. I added quirks to recognize the > > second fan and cleaned the code, any comments are welcome: > > http://pastebin.com/eVPcnvfK > > > > > From your patch, I understand the select_fan() thing we already have for the > > > X60/X61 actually changes which fan is being controlled on the P50? > > > > That is correct, the same mechanism/bit is used to select the active > > fan. But there is some strange behavior when reading the status > > register. Regardless of which fan is currently selected, the status > > register always reads the status of fan1. This means I can set the fan > > speed levels individually, but reading pwm2_enable or pwm2 always returns > > the corresponding value of fan1. > > So, we'd have to implement them in software through a quirk, sort of like > the T43 fan quirk, only worse. > > We have to assume the initial state (like the T43 quirk). Unlike the t43 > quirk, we have to track the states we write, and assume nobody else will > change it behind our backs. > > It is probably safe to assume fan2 defaults to "auto" mode like fan1 by the > time the operating system is started. However, please check if there are > any related BIOS toggles... > This sounds much nastier than I first thought. I will try to find a somewhat clean way to implement this. > > > Otherwise, we can simply add a new quirk for second-fan-control-ok, and > > > enable second-fan control only for the P50, and not for the X60/X61. > > > > I added a quirk based on 'bios_model' (first two characters of BIOS > > version), since the EC version is not read correctly by the module. > > The EC version is not present in dmidecode output anymore, is it? Actually, the EC version is in dmidecode, but in the OEM-specific section. My EC version is 'N1EHT32W'. Handle 0x003F, DMI type 140, 15 bytes OEM-specific Type Header and Data: 8C 0F 3F 00 4C 45 4E 4F 56 4F 0B 07 01 01 02 Strings: N1EHT32W 02/02/2016 > > The Lenovo P70, which is very similar to the P50, shares these first > > two characters 'N1' according to Lenovo's release notes, so it should > > be automatically detected too. Some testing there is needed. > > https://download.lenovo.com/pccbbs/mobiles/n1duj12w.txt > > If it has the same EC firmware, it will not misbehave if we try to mess with > fan2. But if it doesn't have a fan2, it would be nice to not expose it. The P70 has two fans like the P50, so I'm confident there is no need for any special treatment. |