acpi shows wrong battery state (fwd)

[Ian Smith]

On Thu, 2 Sep 2010, Jung-uk Kim wrote:
 > On Thursday 02 September 2010 06:02 pm, kuba wrote:
      [Andriy wrote:]
 > >   >  Take a look at these tunables:
 > >   >  debug.acpi.ec.timeout: 750
 > >   >  debug.acpi.ec.polled: 0
 > >   >  debug.acpi.ec.burst: 0
 > >   >
 > >   >  Perhaps increasing timeout would help?
 > >
 > > Increasing timeout didn't help at all :(
 > > Is there a chance that in the future release 8.2 this problem will
 > > be fixed? I don't have any idea left how to force acpi to show
 > > proper battery state, and I need this information to work with my
 > > laptop,  so I think that the best option is to remove or leave
 > > unused freebsd for some time :/ and use only linux.
 > 
 > As few people already have said, your EC is not responding and that's 
 > the root cause.  However, 8.0 also didn't work according to your 
 > dmesg output.  You said "battery indicator" is not working from 8.1 
 > in another e-mail.  If I understand it correctly, that's some sort of 
 > LED on the laptop, right?  Did "acpiconf -i batt" ever work with 8.0?  
 > I doubt it, though.

8.0 eventually did 'work' despite the hatfull of dmesg error messages, 
and kuba reported that 8.1 has worked (re returning battery info) a 
couple of times which is why I suspected a 'sometimes' ec.timeout issue.

And strange as it sounds, 'acpiconf -i batt' does return the same info 
as 'acpiconf -i0' on my Thinkpad at 8.0, as does 'acpiconf -i foo' :) I 
guess 'acpi_battinfo(atoi(optarg));' just returns 0 for 'foo'.  That is 
what kuba meant by "battery indicator", also the hw.acpi.bat* values.

acpiconf(8) should make clear that _batt_ refers to unit 0, 1 etc; it 
also still (at 8.1) shows -s5 (soft off) as a valid option, when it is 
not.  I'll try a PR with patch once I'm finished upgrading to 8.1-S.

But that's really an aside from the real issue you identify below, and 
perhaps also some TZ value/s being 'absurd' ..

cheers, Ian

 > Another problem is acpi_wmi(4) is not attaching because it cannot 
 > match the device, which is defined wrong in the AML:
 > 
 > 		Name (_HID, "pnp0c14")
 > 
 > It should never be lower cases although it seems we have a workaround 
 > for *some* models in sys/dev/acpi_support/acpi_wmi.c:
 > 
 > 201: static char *wmi_ids[] = {"PNP0C14", "PNP0c14", NULL};
 > 
 > Also, your acpidump output shows typical BIOS stupidity, i.e., it 
 > tries to detect installed MS Windows version and responds 
 > differently.  We have updated ACPICA in 8.1 and it matches Vista's 
 > OSI now.  That's probably the reason why the battery indicator 
 > stopped working.
 > 
 > To recap:
 > 
 > 1.  Add "pnp0c14" in sys/dev/acpi_support/acpi_wmi.c like this:
 > 
 > static char *wmi_ids[] = {"PNP0C14", "PNP0c14", "pnp0c14", NULL};
 > 
 > Recompile, install, and reboot.
 > 
 > 2.  If #1 does not help, remove the following three lines from 
 > sys/contrib/dev/acpica/utilities/uteval.c, which were added in 8.1:
 > 
 > 149:     {"Windows 2006.1",      ACPI_OSI_WINSRV_2008}, ...
 > 150:     {"Windows 2006 SP1",    ACPI_OSI_WIN_VISTA_SP1}, ...
 > 151:     {"Windows 2009",        ACPI_OSI_WIN_7}, ...
 > 
 > Recompile, install, and reboot.
 > 
 > 3.  If #2 still does not help, you should complain to HP.
 > 
 > Good luck.
 > 
 > Jung-uk Kim