CORE-20343
The 3-second timeout FIXME (instead of waiting indefinitely), made in the
COMPBATT worker thread for the `IOCTL_BATTERY_QUERY_STATUS` case, was done
because _*our*_ BATTC handler expects the batteries to always support the
BTP (Battery Trip Point) feature for signaling a change of battery status,
but in the cases where it isn't supported, any waits it tried for the
battery to notify about a status change would never happen.
Furthermore, following commit 3a6e0d4b65, the `SetStatusNotify()` call
(_which always fails if the battery doesn't support the BTP feature_)
would now always exit the `IOCTL_BATTERY_QUERY_STATUS` handling without
any waiting nor battery polling [^1], and this would cause the COMPBATT
worker thread to busy-poll again forever.
The timeout FIXME is now moved to BATTC, instead of COMPBATT, since the
actual fixes should be in BATTC. In particular, it should queue all the
query IOCTLs and then serve them, either using the StatusNotify (BTP)
functionality if the battery supports it, or if not, do a very-slow
battery polling.
I've also increased the timeout a little bit more (5 seconds and not 3).
[^1]: Per the ACPI specification, it is expected that the operating system
performs battery polling if the battery doesn't support BTP, see:
https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/10_Power_Source_and_Power_Meter_Devices/Power_Source_and_Power_Meter_Devices.html#btp-battery-trip-point
Previously the function waited for the conditions *after* querying the status, and then returned the old status. Also, if querying failed, it waited and when the wait timed out it returned STATUS_SUCCESS without returning any data. If the call to SetStatusNotify failed and there was no timeout, it would wait forever.
This is all fixed now.
Move these out of ACPI in preparation for ACPI_NEW so they keep their juicy commit history.
Also both ACPIs can coexist for awhile as I don't see the old driver getting removed anytime soon.
BatteryClassStatusNotify is used by battery miniport drivers to notify the Battery Class of a status change. This can either be a battery status change or battery tag assignation, depending on what the device extension (namely the composite battery) waits for.
We do have implementation for EVENT_BATTERY_STATUS but not for EVENT_BATTERY_TAG. What happens is when BatteryClassIoctl fails to query the battery tag because it has not yet been assigned, the thread is stuck on waiting for the event object to be signaled, forever.
This tipically happens when a timeout of -1 (meaning the calling thread must wait indefinitely) is supplied. The composite battery driver (COMPBATT) is responsible to signal the Battery Class when a CM (Control Method) ACPI battery receives a tag, which then this function will signal the event.
CORE-18969
CORE-19452
KeWaitForSingleObject takes 100ns unit for timeout. Both IOCTL_BATTERY_QUERY_TAG and IOCTL_BATTERY_QUERY_STATUS take a wait for the timeout in milliseconds.
Supposedly a miniport driver wants to supply a wait of 5000 ms (which is equivalent to 5 s), the miniport driver WON'T BE WAITING 5 seconds but 0.5!!!
CORE-18969
CORE-19452
They were reversed, which meant to immediately complete the IRP
if conditions were NOT satisfied.
Remove a now useless/invalid call to QueryStatus, to return the
status which lead to IRP completion.
At first QueryStatus call, output is written to Irp->AssociatedIrp.SystemBuffer.
Unfortunately, this is also the BattWait buffer.
At return of QueryStatus call, BattWait buffer has been modified, so
following code can't use it anymore.
Fix memory overwrite by copying BattWait buffer on the stack.
There is no need to compile our DLLs as shared libraries since we are
managing symbols exports and imports through spec files.
On my system, this reduces the configure-time by a factor of two.
- Add missing parameter to a DPRINT.
- Print pointers using %p instead of %x.
- Move all executable code behind a functions initial DPRINT.
- Do not use non-trivial initializers.
- Check for NULL instead of using !.
- Add missing parentheses.
Hermès Bélusca-Maïto