| Message ID | 20211123232704.25394-1-linux@weissschuh.net (mailing list archive) |
|---|---|
| Headers | show |
| Series | power: supply: add charge_behaviour property (force-discharge, inhibit-charge) | expand |
On Wed, 2021-11-24 at 00:27 +0100, Thomas Weißschuh wrote: > this series adds support for the charge_behaviour property to the power > subsystem and thinkpad_acpi driver. Wonderful! Thanks for working on this. I can confirm inhibit-charge and force-discharge states work with patch v2 on v5.16-rc2 on a T430 (2342-CTO) with BIOS G1ETC2WW (2.82 ), EC G1HT36WW and a single battery. Most behavior is as expected: - With force-discharge, status becomes "Discharging" and energy_now drops over time while AC remains connected. - With inhibit-charge, status becomes "Unknown" and energy_now is stable over time, even when not fully charged. - With auto, status becomes "Charging" and energy_now rises over time. - charge_behaviour takes precedence over charge_control_{start,end}_threshold: status remains Discharging/Unknown when below the start threshold, either due to discharge or threshold change. - charge_behaviour is preserved over soft reboot. - inhibit-charge/auto are preserved across battery removal and reinsertion. - inhibit-charge/auto are preserved across s2ram (S3). - With force-discharge, if the battery is removed, the machine immediately powers off. Some behavior is a little surprising: - charge_behaviour can not be set to force-discharge if AC is disconnected (EIO). If charge_behaviour is force-discharge when AC is disconnected, it changes to auto, unlike inhibit-charge. - charge_behavior force-discharge is not preserved across s2ram (S3), unlike inhibit-charge. - charge_behaviour is not preserved across hard reset (unlike charge thresholds). Interestingly, it appears that inhibit-charge is preserved across power-off (no charging is observed while powered off) but not power-on, even though it is preserved across soft reboot, as noted above. I assume the behavior is under the control of the EC, so these aren't criticisms of the patch. Just some observations. Tested-by: Kevin Locke <kevin@kevinlocke.name> Thanks again, Kevin
Hi, On Wed, Nov 24, 2021 at 12:27:00AM +0100, Thomas Weißschuh wrote: > This series adds support for the charge_behaviour property to the power > subsystem and thinkpad_acpi driver. > > As thinkpad_acpi has to use the 'struct power_supply' created by the generic > ACPI driver it has to rely on custom sysfs attributes instead of proper > power_supply properties to implement this property. > > Patch 1: Adds the power_supply documentation and basic public API > Patch 2: Adds helpers to power_supply core to help drivers implement the > charge_behaviour attribute > Patch 3: Adds support for force-discharge to thinkpad_acpi. > Patch 4: Adds support for inhibit-discharge to thinkpad_acpi. > > Patch 3 and 4 are largely taken from other patches and adapted to the new API. > (Links are in the patch trailer) > > Ognjen Galic: > > Your S-o-b is on the original inhibit_charge and force_discharge patches. > I would like to add you as Co-developed-by but to do that it will also require > your S-o-b. Could you give your sign-offs for the new patches, so you can be > properly attributed? > > Sebastian Reichel: > > Currently the series does not actually support the property as a proper > powersupply property handled fully by power_supply_sysfs.c because there would > be no user for this property. I'm not too happy how the acpi-battery hooks work, but that's not your fault and this patchset does not really make the situation worse. So: Acked-by: Sebastian Reichel <sebastian.reichel@collabora.com> -- Sebastian > Previous discussions about the API: > > https://lore.kernel.org/platform-driver-x86/20211108192852.357473-1-linux@weissschuh.net/ > https://lore.kernel.org/platform-driver-x86/21569a89-8303-8573-05fb-c2fec29983d1@gmail.com/ > > v1: https://lore.kernel.org/lkml/20211113104225.141333-1-linux@weissschuh.net/ > v1 -> v2: > > * Use sysfs_emit-APIs instead of plain sprintf > * More cecks for actual feature availability > * Validation of the written values > * Read inhibit-charge via BICG instead of PSSG (peak shift state) > * Don't mangle error numbers in charge_behaviour_store() > > Open points: > > Thomas Koch has observed that on a T450s with two batteries > inhibit-charge on BAT0 will affect both batteries and for BAT1 it is ignored > entirely, this seems to be a bug in the EC. > On my T460s with two batteries it works correctly. > > Thomas Weißschuh (4): > power: supply: add charge_behaviour attributes > power: supply: add helpers for charge_behaviour sysfs > platform/x86: thinkpad_acpi: support force-discharge > platform/x86: thinkpad_acpi: support inhibit-charge > > Documentation/ABI/testing/sysfs-class-power | 14 ++ > drivers/platform/x86/thinkpad_acpi.c | 191 +++++++++++++++++++- > drivers/power/supply/power_supply_sysfs.c | 51 ++++++ > include/linux/power_supply.h | 16 ++ > 4 files changed, 268 insertions(+), 4 deletions(-) > > > base-commit: 66f4beaa6c1d28161f534471484b2daa2de1dce0 > -- > 2.34.0 >
Hi, On 12/3/21 22:33, Sebastian Reichel wrote: > Hi, > > On Wed, Nov 24, 2021 at 12:27:00AM +0100, Thomas Weißschuh wrote: >> This series adds support for the charge_behaviour property to the power >> subsystem and thinkpad_acpi driver. >> >> As thinkpad_acpi has to use the 'struct power_supply' created by the generic >> ACPI driver it has to rely on custom sysfs attributes instead of proper >> power_supply properties to implement this property. >> >> Patch 1: Adds the power_supply documentation and basic public API >> Patch 2: Adds helpers to power_supply core to help drivers implement the >> charge_behaviour attribute >> Patch 3: Adds support for force-discharge to thinkpad_acpi. >> Patch 4: Adds support for inhibit-discharge to thinkpad_acpi. >> >> Patch 3 and 4 are largely taken from other patches and adapted to the new API. >> (Links are in the patch trailer) >> >> Ognjen Galic: >> >> Your S-o-b is on the original inhibit_charge and force_discharge patches. >> I would like to add you as Co-developed-by but to do that it will also require >> your S-o-b. Could you give your sign-offs for the new patches, so you can be >> properly attributed? >> >> Sebastian Reichel: >> >> Currently the series does not actually support the property as a proper >> powersupply property handled fully by power_supply_sysfs.c because there would >> be no user for this property. > > I'm not too happy how the acpi-battery hooks work, but that's not > your fault and this patchset does not really make the situation > worse. So: > > Acked-by: Sebastian Reichel <sebastian.reichel@collabora.com> I haven't looked at the thinkpad_apci.c bits closely yet (for this new version), but assuming those are ready for merging too, we need to discuss about how to merge this. The thinkpad_acpi code has already seen quite a lot of changes in -next, so I would like the thinkpad_acpi changes to go upstream through the platform-drivers-x86.git tree to avoid conflicts. As such I think it is best if you (Sebastian) can prepare an immutable branch with patch 1 + 2 for me to merge. Then even if patch 3 + 4 need more work, Thomas can just respin those on top of the immutable branch. Alternatively I can take the entire series upstream through the platform-drivers-x86.git tree if that is ok with you (Sebastian). Either way please let me know how you want to proceed with this. Regards, Hans >> Previous discussions about the API: >> >> https://lore.kernel.org/platform-driver-x86/20211108192852.357473-1-linux@weissschuh.net/ >> https://lore.kernel.org/platform-driver-x86/21569a89-8303-8573-05fb-c2fec29983d1@gmail.com/ >> >> v1: https://lore.kernel.org/lkml/20211113104225.141333-1-linux@weissschuh.net/ >> v1 -> v2: >> >> * Use sysfs_emit-APIs instead of plain sprintf >> * More cecks for actual feature availability >> * Validation of the written values >> * Read inhibit-charge via BICG instead of PSSG (peak shift state) >> * Don't mangle error numbers in charge_behaviour_store() >> >> Open points: >> >> Thomas Koch has observed that on a T450s with two batteries >> inhibit-charge on BAT0 will affect both batteries and for BAT1 it is ignored >> entirely, this seems to be a bug in the EC. >> On my T460s with two batteries it works correctly. >> >> Thomas Weißschuh (4): >> power: supply: add charge_behaviour attributes >> power: supply: add helpers for charge_behaviour sysfs >> platform/x86: thinkpad_acpi: support force-discharge >> platform/x86: thinkpad_acpi: support inhibit-charge >> >> Documentation/ABI/testing/sysfs-class-power | 14 ++ >> drivers/platform/x86/thinkpad_acpi.c | 191 +++++++++++++++++++- >> drivers/power/supply/power_supply_sysfs.c | 51 ++++++ >> include/linux/power_supply.h | 16 ++ >> 4 files changed, 268 insertions(+), 4 deletions(-) >> >> >> base-commit: 66f4beaa6c1d28161f534471484b2daa2de1dce0 >> -- >> 2.34.0 >>
Hi Thomas, On 24.11.21 00:27, Thomas Weißschuh wrote: > Hi, > > this series adds support for the charge_behaviour property to the power > subsystem and thinkpad_acpi driver. > > As thinkpad_acpi has to use the 'struct power_supply' created by the generic > ACPI driver it has to rely on custom sysfs attributes instead of proper > power_supply properties to implement this property. > > Patch 1: Adds the power_supply documentation and basic public API > Patch 2: Adds helpers to power_supply core to help drivers implement the > charge_behaviour attribute > Patch 3: Adds support for force-discharge to thinkpad_acpi. > Patch 4: Adds support for inhibit-discharge to thinkpad_acpi. > > Patch 3 and 4 are largely taken from other patches and adapted to the new API. > (Links are in the patch trailer) > > Ognjen Galic: > > Your S-o-b is on the original inhibit_charge and force_discharge patches. > I would like to add you as Co-developed-by but to do that it will also require > your S-o-b. Could you give your sign-offs for the new patches, so you can be > properly attributed? > > Sebastian Reichel: > > Currently the series does not actually support the property as a proper > powersupply property handled fully by power_supply_sysfs.c because there would > be no user for this property. > > Previous discussions about the API: > > https://lore.kernel.org/platform-driver-x86/20211108192852.357473-1-linux@weissschuh.net/ > https://lore.kernel.org/platform-driver-x86/21569a89-8303-8573-05fb-c2fec29983d1@gmail.com/ > > v1: https://lore.kernel.org/lkml/20211113104225.141333-1-linux@weissschuh.net/ > v1 -> v2: > > * Use sysfs_emit-APIs instead of plain sprintf > * More cecks for actual feature availability > * Validation of the written values > * Read inhibit-charge via BICG instead of PSSG (peak shift state) > * Don't mangle error numbers in charge_behaviour_store() > > Open points: > > Thomas Koch has observed that on a T450s with two batteries > inhibit-charge on BAT0 will affect both batteries and for BAT1 it is ignored > entirely, this seems to be a bug in the EC. > On my T460s with two batteries it works correctly. > > Thomas Weißschuh (4): > power: supply: add charge_behaviour attributes > power: supply: add helpers for charge_behaviour sysfs > platform/x86: thinkpad_acpi: support force-discharge > platform/x86: thinkpad_acpi: support inhibit-charge > > Documentation/ABI/testing/sysfs-class-power | 14 ++ > drivers/platform/x86/thinkpad_acpi.c | 191 +++++++++++++++++++- > drivers/power/supply/power_supply_sysfs.c | 51 ++++++ > include/linux/power_supply.h | 16 ++ > 4 files changed, 268 insertions(+), 4 deletions(-) > > > base-commit: 66f4beaa6c1d28161f534471484b2daa2de1dce0 > Reviewed-by : Thomas Koch <linrunner@gmx.net> Tested-by: Thomas Koch <linrunner@gmx.net> Works well on ThinkPad X220, T450s, X1C6 with the exception mentioned above. The new API is included in TLP already [1]. [1] https://github.com/linrunner/TLP/commit/f0bf18f847470ae495a68f9f0e30130b96348936 -- Freundliche Grüße / Kind regards, Thomas Koch Mail : linrunner@gmx.net Web : https://linrunner.de/tlp
Hi, On 12/3/21 22:33, Sebastian Reichel wrote: > Hi, > > On Wed, Nov 24, 2021 at 12:27:00AM +0100, Thomas Weißschuh wrote: >> This series adds support for the charge_behaviour property to the power >> subsystem and thinkpad_acpi driver. >> >> As thinkpad_acpi has to use the 'struct power_supply' created by the generic >> ACPI driver it has to rely on custom sysfs attributes instead of proper >> power_supply properties to implement this property. >> >> Patch 1: Adds the power_supply documentation and basic public API >> Patch 2: Adds helpers to power_supply core to help drivers implement the >> charge_behaviour attribute >> Patch 3: Adds support for force-discharge to thinkpad_acpi. >> Patch 4: Adds support for inhibit-discharge to thinkpad_acpi. >> >> Patch 3 and 4 are largely taken from other patches and adapted to the new API. >> (Links are in the patch trailer) >> >> Ognjen Galic: >> >> Your S-o-b is on the original inhibit_charge and force_discharge patches. >> I would like to add you as Co-developed-by but to do that it will also require >> your S-o-b. Could you give your sign-offs for the new patches, so you can be >> properly attributed? >> >> Sebastian Reichel: >> >> Currently the series does not actually support the property as a proper >> powersupply property handled fully by power_supply_sysfs.c because there would >> be no user for this property. > > I'm not too happy how the acpi-battery hooks work, but that's not > your fault and this patchset does not really make the situation > worse. So: > > Acked-by: Sebastian Reichel <sebastian.reichel@collabora.com> Sebastian, what is the plan for taking this upstream ? Does your ack mean that you are ok with me taking the entire series upstream through the pdx86 tree? Or do you plan to apply patches 1-2 through linux-power-supply.git; and in that case can you provide an inmmutable branch with those patches for me to merge into pdx86/for-next so that I can then apply patches 3 + 4 there ? Note merging everything through the linux-power-supply.git tree is non ideal in this case because the thinkpad_acpi.c code already has a lot of changes in pdx86/for-next. Regards, Hans >> Previous discussions about the API: >> >> https://lore.kernel.org/platform-driver-x86/20211108192852.357473-1-linux@weissschuh.net/ >> https://lore.kernel.org/platform-driver-x86/21569a89-8303-8573-05fb-c2fec29983d1@gmail.com/ >> >> v1: https://lore.kernel.org/lkml/20211113104225.141333-1-linux@weissschuh.net/ >> v1 -> v2: >> >> * Use sysfs_emit-APIs instead of plain sprintf >> * More cecks for actual feature availability >> * Validation of the written values >> * Read inhibit-charge via BICG instead of PSSG (peak shift state) >> * Don't mangle error numbers in charge_behaviour_store() >> >> Open points: >> >> Thomas Koch has observed that on a T450s with two batteries >> inhibit-charge on BAT0 will affect both batteries and for BAT1 it is ignored >> entirely, this seems to be a bug in the EC. >> On my T460s with two batteries it works correctly. >> >> Thomas Weißschuh (4): >> power: supply: add charge_behaviour attributes >> power: supply: add helpers for charge_behaviour sysfs >> platform/x86: thinkpad_acpi: support force-discharge >> platform/x86: thinkpad_acpi: support inhibit-charge >> >> Documentation/ABI/testing/sysfs-class-power | 14 ++ >> drivers/platform/x86/thinkpad_acpi.c | 191 +++++++++++++++++++- >> drivers/power/supply/power_supply_sysfs.c | 51 ++++++ >> include/linux/power_supply.h | 16 ++ >> 4 files changed, 268 insertions(+), 4 deletions(-) >> >> >> base-commit: 66f4beaa6c1d28161f534471484b2daa2de1dce0 >> -- >> 2.34.0 >>
Hi, On 12/3/21 22:33, Sebastian Reichel wrote: > Hi, > > On Wed, Nov 24, 2021 at 12:27:00AM +0100, Thomas Weißschuh wrote: >> This series adds support for the charge_behaviour property to the power >> subsystem and thinkpad_acpi driver. >> >> As thinkpad_acpi has to use the 'struct power_supply' created by the generic >> ACPI driver it has to rely on custom sysfs attributes instead of proper >> power_supply properties to implement this property. >> >> Patch 1: Adds the power_supply documentation and basic public API >> Patch 2: Adds helpers to power_supply core to help drivers implement the >> charge_behaviour attribute >> Patch 3: Adds support for force-discharge to thinkpad_acpi. >> Patch 4: Adds support for inhibit-discharge to thinkpad_acpi. >> >> Patch 3 and 4 are largely taken from other patches and adapted to the new API. >> (Links are in the patch trailer) >> >> Ognjen Galic: >> >> Your S-o-b is on the original inhibit_charge and force_discharge patches. >> I would like to add you as Co-developed-by but to do that it will also require >> your S-o-b. Could you give your sign-offs for the new patches, so you can be >> properly attributed? >> >> Sebastian Reichel: >> >> Currently the series does not actually support the property as a proper >> powersupply property handled fully by power_supply_sysfs.c because there would >> be no user for this property. > > I'm not too happy how the acpi-battery hooks work, but that's not > your fault and this patchset does not really make the situation > worse. So: > > Acked-by: Sebastian Reichel <sebastian.reichel@collabora.com> Sebastian, I have taken the liberty to assume that this means that you are ok with merging the entire series through the pdx86 tree (I've done a test-merge with linux-power-supply/for-next and there are no conflicts). Thomas, Thank you for your patch-series, I've applied the series to my review-hans branch: https://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86.git/log/?h=review-hans Once I've run some tests on this branch the patches there will be added to the platform-drivers-x86/for-next branch and eventually will be included in the pdx86 pull-request to Linus for the next merge-window. Regards, Hans > > -- Sebastian > >> Previous discussions about the API: >> >> https://lore.kernel.org/platform-driver-x86/20211108192852.357473-1-linux@weissschuh.net/ >> https://lore.kernel.org/platform-driver-x86/21569a89-8303-8573-05fb-c2fec29983d1@gmail.com/ >> >> v1: https://lore.kernel.org/lkml/20211113104225.141333-1-linux@weissschuh.net/ >> v1 -> v2: >> >> * Use sysfs_emit-APIs instead of plain sprintf >> * More cecks for actual feature availability >> * Validation of the written values >> * Read inhibit-charge via BICG instead of PSSG (peak shift state) >> * Don't mangle error numbers in charge_behaviour_store() >> >> Open points: >> >> Thomas Koch has observed that on a T450s with two batteries >> inhibit-charge on BAT0 will affect both batteries and for BAT1 it is ignored >> entirely, this seems to be a bug in the EC. >> On my T460s with two batteries it works correctly. >> >> Thomas Weißschuh (4): >> power: supply: add charge_behaviour attributes >> power: supply: add helpers for charge_behaviour sysfs >> platform/x86: thinkpad_acpi: support force-discharge >> platform/x86: thinkpad_acpi: support inhibit-charge >> >> Documentation/ABI/testing/sysfs-class-power | 14 ++ >> drivers/platform/x86/thinkpad_acpi.c | 191 +++++++++++++++++++- >> drivers/power/supply/power_supply_sysfs.c | 51 ++++++ >> include/linux/power_supply.h | 16 ++ >> 4 files changed, 268 insertions(+), 4 deletions(-) >> >> >> base-commit: 66f4beaa6c1d28161f534471484b2daa2de1dce0 >> -- >> 2.34.0 >>