1.8 release cd ACPI message loop

[YONETANI Tomokazu]

On Mon, Jan 29, 2007 at 09:09:36PM -0500, Justin C. Sherrill wrote:
> On Mon, January 29, 2007 1:34 pm, Matthew Dillon wrote:
> 
> >    I would appreciate a few people downloading and testing the release ISO
> >    that is now on the main FTP site.  dfly-1.8.0_REL.iso.gz.
> 
> Booting 1.8 REL on a Thinkpad 600E gives an endless loop of ACPI errors:
> (by hand)

Does this message show up very quickly?  If not, can you try to drop
into the boot loader(press "6" on Fred menu):
  set debug.acpi_semaphore_debug=1
then boot, and write down some messages beginning with "AcpiOsWaitSemaphore"
and "AcpiOsSignalSemaphore".

> ACPI Exception (evregion-0529): AE_TIME, Returned by Handler for
> [EmbeddedControl] [20061109]
> ACPI Error (psparse-0638): Method parse/execution failed
> [\_SB_.PCI0.ISA0.EC0_.I2RB] (Node 0xc1084e78), AE_TIME
> ACPI Error (psparse-0638): Method parse/execution failed
> [\_SB_.PCI0.ISA0.EC0_.UPDT] (Node 0xc1084dd0), AE_TIME
> ACPI Error (psparse-0638): Method parse/execution failed [\_TZ_.THM0._TMP]
> (Node 0xc1084a10), AE_TIME

I couldn't find the DSDT for 600E, but only for 600 on the Internet:
  http://acpi.sourceforge.net/dsdt/view.php?id=457

according to this one, the trace is something like:
  \_TZ.THM0._TMP
  --> \_SB.PCI0,ISA0,EC0.UPDT
  --> \_SB.PCI0,ISA0,EC0.I2RB
  --> (access to one of members of ECOR, trying to acquire a semaphore)

The only place where AE_TIME is returned is AcpiOsWaitSemaphore(),
so it's either
- there's four or more threads waiting for the same semaphore,
- msleep(9) failed for some reason, or
- AcpiOsWaitSemaphore() was instructed to return immediately
  (but it's only in AcpiExSystemWaitSemaphore() and no warnings
   should show up)

There's a mechanism to show you extra debugging information, but
unfortunately acpi.ko gets compiled without ACPI_DEBUG defined,
so it's not available via release ISO (but in this case,
debug.acpi_semaphore_debug should help).  I think making ACPI_DEBUG
enabled for snapshot ISOs is a good idea.

> These may be out of order, since it's looping on these messages and I
> can't tell where one starts and the other stops.  Hitting ctrl-alt-del
> started the normal shutdown process, which makes me think it was the
> console getting spammed with messages.
> 
> Booting with ACPI off removes the messages, not surprisingly.  This laptop
> booted 1.4, I think it was, with ACPI on.

I think debug.acpi.disabled="thermal" (or "ec") in the boot loader can be
used as a workaround for now.

Cheers.