There is a scenario where the status change work can hold
a mutex lock for ~1.5 seconds which can race with the FG resume
callback. Avoid this condition by adding a state variable
to track the suspend/resume state and skip executing the
status change work while suspended.
CRs-Fixed: 2101514
Change-Id: Ib5300a5dfce30c4c6bcc8d8428b664c184a83fb4
Signed-off-by: Anirudh Ghayal <aghayal@codeaurora.org>
Currently, all the values of raw monotonic SOC (0-255) gets
rounded off to 0-100. This can show up monotonic SOC hitting 0%
earlier when the SOC hadn't really hit zero yet. Improve the
SOC round off calculation so that 0 and 100 % can be shown when
it reaches the exact point.
Change-Id: I5bd9ebc8667a5beed9e1e97ff492aa1350f4d0f7
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
When both sides are doing DRP and the other side applies Rd near the end
of PMIC DFP cycle, sometimes, PMIC does not lock on into AttachedWait.SRC
and transitions back into Unattached.SNK.
If a resistance ranging (5.1k,300k) pulldown is applied and then 5.1kohms
pulldown is applied, then it takes 10-20ms to detect 5.1kohms.
Before this, if DRP.DFP time expires, then PMI transitions to UFP mode to
avoid extending DRP cycle time.
Fix it by extending DFP time. Note the total DRP time should still be
within max limit (100ms).
Change-Id: I1dd180229bc771542b4f6ca7850aa457478ae5cd
Signed-off-by: Harry Yang <harryy@codeaurora.org>
In Try.SNK software implementation, after forcing sink, there
is a delay of tDRPTRY + Tccdebounce time, giviing opportunity
to the other side to be a SRC.
The current delay is 100ms < tDRPTRY (90ms) + Tccdebounce (20ms),
too short, which sometimes results in checking Tccdebounce_done
status too early.
Change the delay to 120ms.
Change-Id: If9fe9aa7c405a5ae6416e9373449c4a76295e4d8
Signed-off-by: Harry Yang <harryy@codeaurora.org>
Try.SNK allows a DRP that has a policy based preference to be a Sink
when connecting to another DRP to effect a transition from a destined
Source role to the Sink role.
This is a Software implementation for PMI8998.
Exiting from AttachWait.SRC State, a DRP that strongly prefers the
Sink role may optionally transition to Try.SNK instead of Attached.SRC.
When in the Try.SNK state, the port is querying to determine if the
port partner supports the Source role.
The port shall then transition to Attached.SNK when the SNK.Rp state
is detected for at least tTryCCDebounce and VBUS is detected
Alternatively, the port shall transition to TryWait.SRC if SNK.Rp state
is not detected tTryCCDebounce.
The port shall transition to Attached.SRC when SRC.Rd state is detected
on exactly one of the CC pins.
The port shall transition to Unattached.SNK after tDRPTry if neither of
the CC1 or CC2 pins is in the SRC.Rd state.
CRs-Fixed: 1114313
Change-Id: I777332e87df65082333f4e24a61c5ea834784d09
Signed-off-by: Harry Yang <harryy@codeaurora.org>
This reverts commit 941bfaf2dc.
Bring back the charger and fuel gauge drivers supported on
8996.
Change-Id: Ibba07ece69e29a8c7bf26acdc9dda30ac755fb83
Signed-off-by: Anirudh Ghayal <aghayal@codeaurora.org>
Currently, charge_counter is based off CC_SOC_SW which is based
off coulomb counter. However, there could be some accumulation
of error due to inaccuracy in ADC over time. This can potentially
affect the accuracy of charge_counter. To overcome this, prime
CC_SOC_SW during discharging based off battery SOC and to a full
value during charge termination.
While at it, expose the charge_counter_shadow property based off
battery SOC for comparison.
CRs-Fixed: 2109421
Change-Id: I50de44afbdbd747db95946416a09062df205fd6c
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
When SOC is held @ 100% after charge termination, charge_full
flag is cleared only when the monotonic SOC drops below recharge
SOC threshold. This can cause the flag to be held for a long time
when the charger is removed. Fix this by clearing charge_full
flag when charging status had changed from FULL.
Change-Id: I35b52ddc45f314347f0e4d8433d5fb550663267c
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Delta battery SOC interrupt can fire during boot just after
registering the interrupts before disabling it through vote() on
delta_bsoc_irq_en_votable during driver probe. When this happens,
from the delta_bsoc interrupt handler, an attempt to disable the
interrupt can be made by fg_charge_full_update() if charge_done
is not set. This will in turn acquire the mutex lock of votable
while trying to do disable_irq() which needs synchronization.
Prevent this situation by using disable_irq_nosync() when
disabling the interrupts from votable callbacks.
CRs-Fixed: 2125269
Change-Id: I1b617c4afd39c544ce79f9396030b730c55445a7
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Currently, wait for soc_update completion is attempted once again
only when the wait is interrupted. Try waiting again even when it
is timed out. This should help IMA_ATOMIC transactions go through
reliably.
Change-Id: I8d31fc41020b8c05ff4d562ae205eaa278d7ffcb
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Handle notifications from parallel charger in GEN3 FG driver so
that ESR FCC control can be configured sooner based on parallel
charging status change. Without this, GEN3 FG driver can simply
wait on notifications from battery/USB power supply to move from
HW autonomous to SW controlled ESR FCC settings when parallel
charging is enabled which cause unwanted battery current spikes.
CRs-Fixed: 2119134
Change-Id: Ic4c473a6cc6062f2eaaebc68c30a1a51763c5dda
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Currently the code ends up holding a wakelock even while discharging.
There is no need to hold wakelock or handle stepcharging/jeita
situations while discharging.
Change-Id: I0edfd16cef88ec45b0f0ebb80533b62c6e775116
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Disable USB IRQs after high duty cycle IRQ triggered and re-enable
them in the delay worker. This helps to avoid the high duty cycle
IRQ storming.
CRs-Fixed: 2094725
Change-Id: Iba22a7402692b5ee5f738cc17ed9c214f6014ac0
Signed-off-by: Fenglin Wu <fenglinw@codeaurora.org>
Currently the driver skips invoking adc api's if usb is not
present i.e. vbus has fallen below 3.3V.
This presence check is not required for the vbus. Skip that
check.
Change-Id: I4a2dd91109a308b553b7407618ee8f3221127f61
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Currently the driver moves FCC to PMI as it steps down
from SMB. Instead we should step down just the SMB.
Change-Id: I7ad9ca4364a44bacb539f8b045cf97f7945155f5
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Currently the code only waits 100mS between steps. The hardware
could take longer to deglitch its signals and status. Extend the
wait to 500mS.
Change-Id: I4a037f85f43bc670c5a357e4ecb46962b0d33598
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Currently, ESR extraction control is configured from two paths.
First is invoked when Qnovo is enabled or disabled. Second is
invoked when the pulse train is done and ESR measurement is
requested. When the latter is run, there is a sleep time of 1.5
seconds that is essential for allowing FG to do ESR measurement.
However in that time window, if Qnovo is disabled for some reason
then ESR extraction control will be kept disabled unexpectedly.
Fix this by protecting both the paths using a mutex lock as well
as checking for qnovo enable status before disabling ESR
extraction control after measuring ESR.
Change-Id: Ib93539071724a1ce59d198074854a8a9c5fd3ab3
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Currently, when becoming inactive, PD leaves the HDC and input
limited interrupt enabled.
Correct it by disabling it.
Change-Id: I62a1d837f6126951cb428a982fe1d8876f1d5bdc
Signed-off-by: Harry Yang <harryy@codeaurora.org>
Change the OTG regulator ops to static for smb1351 and smb135x
charger drivers.
Change-Id: Ide0c6461275f3ba74e0420e1d1a36c1e1f2bd18d
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
In certain conditions, after the battery SOC reaches 100%, SOC is
not getting updated during discharging. Based on the hardware
recommendation, adjust the Ki coefficient for full SOC so that
SOC can be updated faster during discharging.
While at it, remove prev_charge_status variable which is not
used really.
CRs-Fixed: 2089555
Change-Id: I54b7d86529c21c016ab777053895ccda9c7a2b35
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Currently, FULL_SOC and MONOTONIC_SOC registers are updated
multiple times when the recharge SOC threshold is close enough to
trigger the recharging frequently. This was leading to a SOC
fluctuation when SOC masking algorithm is enabled.
Rearrange charge_full_update used in SOC masking algorithm such
that FULL_SOC and MONOTONIC_SOC registers are updated just once
when the charge_full flag is cleared.
Also, when the device is rebooted/shutdown and if charge_full
flag is set, write to FULL_SOC with battery_soc and a full value
to MONOTONIC_SOC registers. This helps to maintain the same SOC
across a device reboot.
Change-Id: I95afa90f27e4c0f948f301ff8d6a1f84ae422ef0
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Provide means to suspend the DC path using the power supply
property.
Change-Id: I97f83f5f7c892fc078a92914af0a923b6aacf02e
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Currently, SOC linearization is done when SOC masking feature is
enabled to spread out the drop in SOC across a wider range and
for better user experience. Make this configurable through a DT
property "qcom,linearize-soc" so that the user can enable it as
and when required.
Change-Id: Ic9f5f04f4e872d7804152c17206a8b204de3a9c1
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Change the regulator_ops of vbus/vconn regulators in smb2 and
smb138x drivers to static. While at it, change the interrupt
handler smb138x_handle_slave_chg_state_change to static.
Change-Id: Iacd6b84adc6db1819b4a337aac9f6ccc52989dcd
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Currently, JEITA thresholds are configured through the device
tree properties. Allow them to be configured runtime through
power supply properties.
Change-Id: Ie93c7e447a4f8ac793e6e324f7a0269dbc14c7e5
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Add POWER_SUPPLY_PROP_COLD_TEMP and POWER_SUPPLY_PROP_HOT_TEMP
properties to represent Cold and Hot JEITA thresholds of the
battery.
Change-Id: Id5e16a4f7b20bb05d997277d20db05fc08db9f33
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
CHARGE_FULL property indicates the learnt capacity of the battery
from the last capacity learning cycle. However, in some rare
cases, if the learnt capacity is having an unexpected value, it
requires a reset to a good value which is not possible for an
user who cannot reinsert a battery or do an explicit write to FG
SRAM. Allowing CHARGE_FULL to be writable helps this.
Change-Id: I05aa8392f103685f8fc1ba5a3780122150be0ee6
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>
Currently the handler skips calling the update function
but uses the final result from the votable not_ok_to_qnovo_votable.
Fix it by calling the update function so it gets a chance to vote on
not_ok_to_qnovo_votable.
Moreover do not return early from ptrain_done handler without setting
QNI_PT_VOTER's vote on pt_dis_votable. This is a self clearing bit and
unless QNI_PT_VOTER's bit is removed, the next restart may not trickle
down to hw.
Change-Id: I0a860287b638d20b6a4aa05c7bb25e9b81a980d2
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Usually, DC path is connected to a wireless charger. It is advisable
to disable qnovo when charging wirelessly.
Disable qnovo for dc path unless a property in device tree is set to
specifically enable qnovo for dc path.
Change-Id: I4ca475f30ca0a564f533b4ed8bcd55cfb5da291f
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
If a charger was already present at bootup, we may not delay the start
of qnovo by 15 seconds.
Fix it by setting the chg_ready votable to false during init.
Change-Id: Icf185763576f65028cf2ef0d1c0bbcf6570460ba
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
While shutting down set the not_ok_to_qnovo votable to true, so that
we skip measuring ESR if a pulse train done interrupt triggers right
during the shutdown.
Change-Id: If3a0a8e9b1d60920cbbedc16af91c05caec8dcb2
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Sometimes we observe that the ptrain_en bit is reset back to 0 while
in restart work.
Set it again if so.
Change-Id: I15d46f96a8b751f0386677e3b6876eec8bba9a1f
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
