| Message ID | 20210919200735.142862-2-henrik@grimler.se (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | power: supply: max17042_battery: use VFSOC for capacity when no rsns | expand |
On 19/09/2021 22:07, Henrik Grimler wrote: > On Galaxy S3 (i9300/i9305), which has the max17047 fuel gauge and no > current sense resistor (rsns), the RepSOC register does not provide an > accurate state of charge value. The reported value is wrong, and does > not change over time. VFSOC however, which uses the voltage fuel gauge > to determine the state of charge, always shows an accurate value. > > At least one max170xx driver, found in Asus' Z00D kernel [1], chooses > how to get the capacity based on if current sense is available or not. > Lets change the mainline driver to match the Asus Z00D driver's > behaviour and thereby fix so that correct state of charge values are > obtained on Galaxy S3. > > [1] https://github.com/LineageOS/android_kernel_asus_Z00D/blob/c7ab0e3ec5b5/drivers/power/max17042_battery.c#L1103-L1105 > > Suggested-by: Wolfgang Wiedmeyer <wolfgit@wiedmeyer.de> > Signed-off-by: Henrik Grimler <henrik@grimler.se> > --- > drivers/power/supply/max17042_battery.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/drivers/power/supply/max17042_battery.c b/drivers/power/supply/max17042_battery.c > index 622bdae6182c..7233670978d0 100644 > --- a/drivers/power/supply/max17042_battery.c > +++ b/drivers/power/supply/max17042_battery.c > @@ -317,7 +317,10 @@ static int max17042_get_property(struct power_supply *psy, > val->intval = data * 625 / 8; > break; > case POWER_SUPPLY_PROP_CAPACITY: > - ret = regmap_read(map, MAX17042_RepSOC, &data); > + if (chip->pdata->enable_current_sense) > + ret = regmap_read(map, MAX17042_RepSOC, &data); > + else > + ret = regmap_read(map, MAX17042_VFSOC, &data); Thanks for the patch. I found also my comments to Wolfgang's patch in 2016, which you resolve here but I have more. :) I think my previous message about current sense are not correct. What is important is whether ModelGauge is being used/configured. For example none of DT platforms support it but ACPI might. There is incoming patch around it: https://lore.kernel.org/lkml/5702731.UytLkSCjyO@pliszka/ If you switch to VSSoc, I think you need to modify the SOC Alert Config in MiscCFG register (bits 0:1 to 0x1). Otherwise the alerts will be generated on different value. Different topic: When touching Exynos-based boards (like Galaxy S3), please Cc me as well, even if I don't pop up in the maintainers. For max17042 we need to Cc broader group of users, for example using it in ACPI platforms. The best is to pick the contributors. Best regards, Krzysztof
On Fri, Sep 24, 2021 at 01:45:29PM +0200, Krzysztof Kozlowski wrote: > Thanks for the patch. I found also my comments to Wolfgang's patch in > 2016, which you resolve here but I have more. :) > > I think my previous message about current sense are not correct. What is > important is whether ModelGauge is being used/configured. For example > none of DT platforms support it but ACPI might. > There is incoming patch around it: > https://lore.kernel.org/lkml/5702731.UytLkSCjyO@pliszka/ > > If you switch to VSSoc, I think you need to modify the SOC Alert Config > in MiscCFG register (bits 0:1 to 0x1). Otherwise the alerts will be > generated on different value. So, 0x1 should correspond to AvSOC (i.e. non-filtered RepSOC), while right now we write 0x3 (VFSOC) to MiscCFG for devices without current sense [1]. Could you elaborate on why AvSOC should be used for alert if we use VFSOC to get PROP_CAPACITY? On this particular device it does not seem to make a difference what I use for the SOC alert, the alert triggers all the time in any case since RepSOC does not give an accurate value. Supposedly this happens because ModelGauge configuration is incomplete, as you said. Looking at the registers used by the ModelGauge it seems that only the "characterization table" at 0x80 - 0xAF is missing. The rest (FullCap, DesignCap, ICHGTerm, ..) are set to the same values as with vendor kernel. > Different topic: > When touching Exynos-based boards (like Galaxy S3), please Cc me as > well, even if I don't pop up in the maintainers. > > For max17042 we need to Cc broader group of users, for example using it > in ACPI platforms. The best is to pick the contributors. Thanks for taking the time to explain this. > Best regards, > Krzysztof [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/power/supply/max17042_battery.c#n1092 Best regards, Henrik Grimler
On 27/09/2021 21:05, Henrik Grimler wrote: > On Fri, Sep 24, 2021 at 01:45:29PM +0200, Krzysztof Kozlowski wrote: >> Thanks for the patch. I found also my comments to Wolfgang's patch in >> 2016, which you resolve here but I have more. :) >> >> I think my previous message about current sense are not correct. What is >> important is whether ModelGauge is being used/configured. For example >> none of DT platforms support it but ACPI might. >> There is incoming patch around it: >> https://lore.kernel.org/lkml/5702731.UytLkSCjyO@pliszka/ >> >> If you switch to VSSoc, I think you need to modify the SOC Alert Config >> in MiscCFG register (bits 0:1 to 0x1). Otherwise the alerts will be >> generated on different value. > > So, 0x1 should correspond to AvSOC (i.e. non-filtered RepSOC), while > right now we write 0x3 (VFSOC) to MiscCFG for devices without current > sense [1]. Could you elaborate on why AvSOC should be used for alert > if we use VFSOC to get PROP_CAPACITY? I meant that same measurement should be used for both: for PROP_CAPACITY and for alerts. I double checked the driver and your change is actually aligned with it. If !enable_current_sense, the driver will set MiscCFG to 0x3 to use VFSOC for alerts. I think you can ignore that part of my comment before. However still remaining issue is that switching to VFSoc should happen not only if !enable_current_sense but also if ModelGauge m3 is not configured. > On this particular device it does not seem to make a difference what I > use for the SOC alert, the alert triggers all the time in any case > since RepSOC does not give an accurate value. Supposedly this happens > because ModelGauge configuration is incomplete, as you said. Looking > at the registers used by the ModelGauge it seems that only the > "characterization table" at 0x80 - 0xAF is missing. The rest (FullCap, > DesignCap, ICHGTerm, ..) are set to the same values as with vendor > kernel. Are you sure? I could not find setting of these (e.g. MAX17042_FullCAP/config->fullcap) for a DT platform. Best regards, Krzysztof
On 19/09/2021 22:07, Henrik Grimler wrote: > On Galaxy S3 (i9300/i9305), which has the max17047 fuel gauge and no > current sense resistor (rsns), the RepSOC register does not provide an > accurate state of charge value. The reported value is wrong, and does > not change over time. VFSOC however, which uses the voltage fuel gauge > to determine the state of charge, always shows an accurate value. > > At least one max170xx driver, found in Asus' Z00D kernel [1], chooses > how to get the capacity based on if current sense is available or not. > Lets change the mainline driver to match the Asus Z00D driver's > behaviour and thereby fix so that correct state of charge values are > obtained on Galaxy S3. > > [1] https://github.com/LineageOS/android_kernel_asus_Z00D/blob/c7ab0e3ec5b5/drivers/power/max17042_battery.c#L1103-L1105 > > Suggested-by: Wolfgang Wiedmeyer <wolfgit@wiedmeyer.de> > Signed-off-by: Henrik Grimler <henrik@grimler.se> > --- > drivers/power/supply/max17042_battery.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > After explanation and double-checking of driver this makes sense. The driver already uses VFSoc for alerts (MiscCFG register) if !enable_current_sense. Fixes: 359ab9f5b154 ("power_supply: Add MAX17042 Fuel Gauge Driver") Cc: <stable@vger.kernel.org> These could be added when applying but maybe Sebastian wants a v2 with it? Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com> Best regards, Krzysztof
On Tue, Sep 28, 2021 at 11:18:27AM +0200, Krzysztof Kozlowski wrote: > On 27/09/2021 21:05, Henrik Grimler wrote: > > On Fri, Sep 24, 2021 at 01:45:29PM +0200, Krzysztof Kozlowski wrote: > >> If you switch to VSSoc, I think you need to modify the SOC Alert Config > >> in MiscCFG register (bits 0:1 to 0x1). Otherwise the alerts will be > >> generated on different value. > > > > So, 0x1 should correspond to AvSOC (i.e. non-filtered RepSOC), while > > right now we write 0x3 (VFSOC) to MiscCFG for devices without current > > sense [1]. Could you elaborate on why AvSOC should be used for alert > > if we use VFSOC to get PROP_CAPACITY? > > I meant that same measurement should be used for both: for PROP_CAPACITY > and for alerts. > > I double checked the driver and your change is actually aligned with it. > If !enable_current_sense, the driver will set MiscCFG to 0x3 to use > VFSOC for alerts. I think you can ignore that part of my comment before. Makes sense, thanks! > However still remaining issue is that switching to VFSoc should happen > not only if !enable_current_sense but also if ModelGauge m3 is not > configured. If I manage to get ModelGauge working on this device in the future I can address this. > > On this particular device it does not seem to make a difference what I > > use for the SOC alert, the alert triggers all the time in any case > > since RepSOC does not give an accurate value. Supposedly this happens > > because ModelGauge configuration is incomplete, as you said. Looking > > at the registers used by the ModelGauge it seems that only the > > "characterization table" at 0x80 - 0xAF is missing. The rest (FullCap, > > DesignCap, ICHGTerm, ..) are set to the same values as with vendor > > kernel. > > Are you sure? I could not find setting of these (e.g. > MAX17042_FullCAP/config->fullcap) for a DT platform. Actually, it seems that the registers are set to the default Power-On Reset (POR) values in both mainline and vendor kernel. Printing all the Cell Characterization Information Registers (given in Table 1 in the MAX17047 datasheet) with something like: u32 tmp; regmap_read(chip->regmap, MAX17042_FullCAP, &tmp); dev_err(&chip->client->dev, "Fullcap %04x\n", tmp); in both vendor kernel and mainline gives the same values: Fullcap 07d0 DesignCap 07d0 ICHGTerm 03c0 FullCapNom 0667 RCOMP0 0065 Iavg_empty 0780 TempCo 0930 QRes 00 1e2f QRes 10 1e00 QRes 20 1306 QRes 30 0c00 and these are the POR values (seen in Table 5 of the datasheet). Only difference between vendor and mainline is that MAX17042_MODELChrTbl is all zeros on mainline, while with vendor kernel I get something like: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 50 05 c0 a0 85 a9 f1 e0 5d 84 f1 01 00 00 00 and the values here change over time as well, as the algorithm learns about the battery(?). Maybe this means that the ModelGauge algorithm is not configured with vendor kernel either, and that a full battery characterization (described in [1]) is needed if we are to use ModelGauge. ModelGauge is also not mentioned in the vendor kernel driver for max17047 [2], but it is mentioned in the very similar max17042 [3] and max17050 drivers in the same kernel. > Best regards, > Krzysztof [1] https://pdfserv.maximintegrated.com/en/an/AN4799.pdf [2] https://github.com/LineageOS/android_kernel_samsung_smdk4412/blob/2489007e7d74/drivers/battery/max17047_fuelgauge.c [3] https://github.com/LineageOS/android_kernel_samsung_smdk4412/blob/2489007e7d74/drivers/battery/max17042_fuelgauge.c Best regards, Henrik Grimler
On Tue, Sep 28, 2021 at 11:22:15AM +0200, Krzysztof Kozlowski wrote: > On 19/09/2021 22:07, Henrik Grimler wrote: > > On Galaxy S3 (i9300/i9305), which has the max17047 fuel gauge and no > > current sense resistor (rsns), the RepSOC register does not provide an > > accurate state of charge value. The reported value is wrong, and does > > not change over time. VFSOC however, which uses the voltage fuel gauge > > to determine the state of charge, always shows an accurate value. > > > > At least one max170xx driver, found in Asus' Z00D kernel [1], chooses > > how to get the capacity based on if current sense is available or not. > > Lets change the mainline driver to match the Asus Z00D driver's > > behaviour and thereby fix so that correct state of charge values are > > obtained on Galaxy S3. > > > > [1] https://github.com/LineageOS/android_kernel_asus_Z00D/blob/c7ab0e3ec5b5/drivers/power/max17042_battery.c#L1103-L1105 > > > > Suggested-by: Wolfgang Wiedmeyer <wolfgit@wiedmeyer.de> > > Signed-off-by: Henrik Grimler <henrik@grimler.se> > > --- > > drivers/power/supply/max17042_battery.c | 5 ++++- > > 1 file changed, 4 insertions(+), 1 deletion(-) > > > > After explanation and double-checking of driver this makes sense. The > driver already uses VFSoc for alerts (MiscCFG register) if > !enable_current_sense. > > Fixes: 359ab9f5b154 ("power_supply: Add MAX17042 Fuel Gauge Driver") > Cc: <stable@vger.kernel.org> > > These could be added when applying but maybe Sebastian wants a v2 with it? > > Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com> I can send a v2 with updated commit message to mention that we already use VFSOC for the alert on this device, and sneak in a patch to fix a typo in Iavg_empty parameter. > Best regards, > Krzysztof Best regards, Henrik Grimler
diff --git a/drivers/power/supply/max17042_battery.c b/drivers/power/supply/max17042_battery.c index 622bdae6182c..7233670978d0 100644 --- a/drivers/power/supply/max17042_battery.c +++ b/drivers/power/supply/max17042_battery.c @@ -317,7 +317,10 @@ static int max17042_get_property(struct power_supply *psy, val->intval = data * 625 / 8; break; case POWER_SUPPLY_PROP_CAPACITY: - ret = regmap_read(map, MAX17042_RepSOC, &data); + if (chip->pdata->enable_current_sense) + ret = regmap_read(map, MAX17042_RepSOC, &data); + else + ret = regmap_read(map, MAX17042_VFSOC, &data); if (ret < 0) return ret;
On Galaxy S3 (i9300/i9305), which has the max17047 fuel gauge and no current sense resistor (rsns), the RepSOC register does not provide an accurate state of charge value. The reported value is wrong, and does not change over time. VFSOC however, which uses the voltage fuel gauge to determine the state of charge, always shows an accurate value. At least one max170xx driver, found in Asus' Z00D kernel [1], chooses how to get the capacity based on if current sense is available or not. Lets change the mainline driver to match the Asus Z00D driver's behaviour and thereby fix so that correct state of charge values are obtained on Galaxy S3. [1] https://github.com/LineageOS/android_kernel_asus_Z00D/blob/c7ab0e3ec5b5/drivers/power/max17042_battery.c#L1103-L1105 Suggested-by: Wolfgang Wiedmeyer <wolfgit@wiedmeyer.de> Signed-off-by: Henrik Grimler <henrik@grimler.se> --- drivers/power/supply/max17042_battery.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)