[maemo-developers] Information on the interface to hald-addon-bme (so it can be replaced with something that bypasses BME and talks to the kernel directly)

[From: Jonathan Wilson]

The following items are exposed by hald-addon-bme:
HAL items exposed under /org/freedesktop/Hal/devices/bme
maemo.charger.type (data type is string), valid values are
wall charger
host 500 mA
host 100 mA
none
unknown
error
initial value is none

maemo.charger.connection_status (data type is string), valid values are
connected
error
disconnected
unknown
initial value is disconnected

maemo.rechargeable.charging_status (data type is string), valid values are
error
unknown
on
continue
off
full
initial value is off

maemo.rechargeable.positive_rate (data type is bool). Set when the flow 
into the battery is positive as far as I can tell.
initial is false

battery.present (data type is bool)
Always set to true

battery.type (data type is string)
Always set to pda

battery.is_rechargeable (data type is bool)
Always set to true

battery.remaining_time (data type is int)
set to the remaining battery use time

battery.charge_level.capacity_state (data type is string), valid values are
low
ok
full
empty
unknown
initial value is ok

battery.charge_level.unit (data type is string)
Always set to bars

battery.charge_level.design (data type is int)
Set to the design charge level, initial value is 0

battery.charge_level.current (data type is int)
Set to the current charge level, initial value is 0

battery.charge_level.percentage (data type is int)
set to capacity / nominal capacity, initial value is 0

battery.charge_level.last_full (data type is int)
set to last full charge level, initial value is 0

battery.rechargeable.is_charging (data type is bool)
set when charging, initial value is false

battery.rechargeable.is_discharging (data type is bool)
set when discharging, initial value is false

battery.reporting.unit (data type is string)
Always set to mAh

battery.reporting.design (data type is int)
set to nominal capacity, initial value is 0

battery.reporting.last_full (data type is int)
set to last full capacity, initial value is 0

battery.reporting.current (data type is int)
set to current capacity, initial value is 0

battery.voltage.unit (data type is string)
Always set to mv

battery.voltage.design (data type is int)
initial value is 0, it will then be set to one of
4100
4200
9999

battery.voltage.current (data type is int)
set to current voltage, initial value is 0

battery.remaining_time.calculate_per_time (data type is bool)
Always set to false

Also, hald-addon-bme exposes the following signals under 
com.nokia.bme.signal /com/nokia/bme/signal (unless otherwise specified, 
these signals have no arguments)
charger_connected
charger_disconnected
charger_charging_failed
charger_charging_off
charger_charging_on
battery_full
battery_empty
battery_low
battery_state_changed (arguments for this are 2 uints, dbus type 'u', 
unknown exactly what they are for)
battery_timeleft (arguments for this are 2 uints, dbus type 'u', first one 
is for idle time, second is for active time)

The purpose of these signals should be self-explanatory for the most part.

Plus it exposes 2 dbus methods on com.nokia.bme.request. Both take no 
parameters. The status_info_req method causes a method to be run that will 
send the appropriate signals from the list above. The timeleft_info_req 
method also causes the signals to be sent but it also sends a message to 
BME to retrieve the relavent data for the timeout fields.

Also it should be possible to create a dummy libbmeipc that supports the 
bits needed by dsme-thermalobject-surface and modules-nokia-voice.so (the 2 
users of libbmeipc we need to care about), thus allowing a complete 
replacement of BME without loosing the functioning of the nokia pulseaudio 
bits.

Information that is relavent for said libbmeipc replacement: (all taken 
from various revisions of the LGPL MeeGo n900_libbme sources and all 
verified against the N900 Fremantle BME to ensure they match)
typedef __uint16_t uint16;
typedef __uint32_t uint32;

struct emsg_battery_info_req
{
     uint16      type, subtype;
     uint32      flags;
};

struct emsg_battery_info_reply
{
     uint32      zero;
     uint32      flags;
     uint16      bat_type;
     uint16      nominal_capa;
     uint16      temp;
     uint16      voltage;
     uint16      voltage_tx_on;
     uint16      voltage_tx_off;
     uint16      voltage_pwm_on;
     uint16      voltage_pwm_off;
     uint16      generation;
     uint16      voltage_sh_chk;
};

union emsg_battery_info
{
     struct emsg_battery_info_req        request;
     struct emsg_battery_info_reply      reply;
};

#define EM_BATTERY_INFO_REQ 0x06
#define EM_BATTERY_TEMP 0x0004

int bme_connect(void);
void bme_disconnect(void);
int bme_write(const void *msg, int bytes);
int bme_read(void *msg, int bytes);

Study the source code of dsme-thermalobject-surface to see how it is 
calling libbmeipc and what its sending to (and receiving from) BME. I can 
confirm from my analysis that module-nokia-voice is making the exact same 
libbmeipc calls (and, like dsme-thermalobject-surface, only cares about the 
temprature)

I hope this information is enough to allow someone to replace bme, 
libbmeipc and hald-addon-bme with new software that talks directly to the 
hardware whilst still keeping all of the nokia closed source components 
functioning 100% (and any 3rd party components that talk to hald-addon-bme)
Please let me know if you need any more information on this stuff or have 
any questions, I want to do whatever I can in terms of analysis to enable 
improvements to battery and power management on the N900.