Bug 1622

Summary:	Dell PE4600 requires SCI w/ level polarity w/o override -- IO-APIC
Product:	ACPI	Reporter:	Shaohua (shaohua.li)
Component:	Config-Interrupts	Assignee:	Len Brown (lenb)
Status:	CLOSED CODE_FIX
Severity:	normal	CC:	acpi-bugzilla, namsh, suresh.b.siddha
Priority:	P2
Hardware:	i386
OS:	Linux
Kernel Version:	2.6.0-test9	Subsystem:
Regression:	---	Bisected commit-id:
Attachments:	interrupt when SCI is edge interrupt when SCI is level dmesg when SCI is edge dmesg when SCI is level acpidmp Dmesg on SCB2 system, which exhibits the same problem i386 patch on top of linux-acpi-test-2.6.4 tree updated i386 patch handles non-identity mappings

Description Shaohua 2003-12-01 18:45:57 UTC

Distribution:
Hardware Environment:
Dell poweredge 3600


Steps to reproduce:
the box lacks SCI overriding. Current code will set SCI to edge state in the 
box. But SCI must be level in the box.
related Bug:Bug 774

Comment 1 Shaohua 2003-12-01 18:54:01 UTC

if we set default SCI polarity to edge, my box will break. if verbose, Tyan 
board(bug 774) will break.

Comment 2 Shaohua 2003-12-01 22:32:47 UTC

Created attachment 1587 [details]
interrupt when SCI is edge

Comment 3 Shaohua 2003-12-01 22:33:39 UTC

Created attachment 1588 [details]
interrupt when SCI is level

Comment 4 Shaohua 2003-12-01 22:34:29 UTC

Created attachment 1589 [details]
dmesg when SCI is edge

Comment 5 Shaohua 2003-12-01 22:35:11 UTC

Created attachment 1590 [details]
dmesg when SCI is level

Comment 6 Shaohua 2003-12-01 23:22:28 UTC

Created attachment 1593 [details]
acpidmp

Comment 7 Andy Grover 2004-02-24 16:38:08 UTC

Created attachment 2225 [details]
Dmesg on SCB2 system, which exhibits the same problem

This machine is why we added the SCI level force way back when. As you can see
it has no overrides for IRQ 9, but yet the SCI only works if it is forced level
low. One other note about this dmesg - the dswload warnings and the global lock
warnings are not the problem. We looked at these when the problem originally
came up.

Comment 8 Len Brown 2004-03-21 23:45:02 UTC

Created attachment 2384 [details]
i386 patch on top of linux-acpi-test-2.6.4 tree

Comment 9 Len Brown 2004-03-22 13:12:42 UTC

Created attachment 2388 [details]
updated i386 patch handles non-identity mappings

Comment 10 Len Brown 2004-03-22 22:38:47 UTC

2.6 patch to share i386/kernel/acpi/boot.c with x86_64, thus bringing that arch 
in sync with i386 and this fix: 
 
http://ftp.kernel.org/pub/linux/kernel/people/lenb/acpi/patches/test/2.6.4/20040323013217-boot.c-x86_64.patch

Comment 11 Len Brown 2004-03-23 01:44:36 UTC

i386 and x86_64 patch against 2.4.22: 
http://ftp.kernel.org/pub/linux/kernel/people/lenb/acpi/patches/test/2.4.22/20040323042335-SCI.patch

Comment 12 Len Brown 2004-04-12 20:58:40 UTC

in 2.4.26 and 2.6.5 -- closing.

Comment 13 Len Brown 2004-05-14 23:12:59 UTC

*** Bug 1689 has been marked as a duplicate of this bug. ***