>Number:         20897
>Category:       port-i386
>Synopsis:       ACPI runs for two days and then gets stuck in the ACPI thread
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-i386-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Mar 26 16:34:00 PST 2003
>Closed-Date:
>Last-Modified:
>Originator:     Christos Zoulas
>Release:        NetBSD 1.6Q
>Organization:
	Mess, Inc.
>Environment:
System: NetBSD shiny.astron.com 1.6Q NetBSD 1.6Q (SHINY) #107: Wed Mar 19 21:28:05 EST 2003 christos@shiny.astron.com:/usr/src/sys/arch/i386/compile/SHINY i386
Architecture: i386
Machine: i386
>Description:
	The kernel runs fine for a couple of days then it gets stuck
	deallocating the ACPI tree.

The kernel prints:
ACPI-0714: *** Error: Looking up [BUFF] in namespace, AE_NOT_FOUND
ACPI-1286: *** Error: Method execution failed [\_SB_.VALD.GHCI] (Node 0xc086a580), AU_NOT_FOUND

then the thread gets stuck and no userland processes can run. The machine
is still pingable, but the fan gets stuck on high rpm and the only way to
recover is to hit the power switch.
>How-To-Repeat:
	Configure a kernel on a Libretto L2 with all the ACPI devices.
	Stay on the console and wait a couple of days. Then you can
	get into the debugger when the kernel get stuck with control-alt-escape.
>Fix:
	Probably fix the code that deallocates the tree so that it does
	skips the nodes that are not found when freeing.
>Release-Note:
>Audit-Trail:
>Unformatted: