2.4.27-pre3 backout acpi_fixed_pwr_button

Marcelo:

2.4.27-pre3 still seems to have the acpi_fixed_pwr_button and acpi_fixed_sleep_button changes in it, which oopses for me upon module unload. Sergey Vlasov's response to my report is attached. The original oops report is here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=108405180820535&w=2

If you don't have another fix for it, please apply the included patch in order to back that out.

Regards,
Özkan Sezer

X-UIDL: 1084210475.26274.qsol01,S=4197
X-Mozilla-Status: 0011
X-Mozilla-Status2: 00000000
Return-Path: <[email protected]>
Delivered-To: [email protected]
Received: (qmail 26269 invoked from network); 10 May 2004 17:34:35 -0000
Received: from unknown ([212.252.122.201])
(envelope-sender <>)
by qsol01.superonline.com (qmail-ldap-1.03) with QMQP for <>; 10 May 2004 17:34:35 -0000
Delivered-To: CLUSTERHOST vfep01.superonline.com [email protected]
Received: (qmail 32592 invoked from network); 10 May 2004 17:34:15 -0000
Received: from unknown (HELO ns1.murom.ru) ([213.177.124.6])
(envelope-sender <[email protected]>)
by vfep01.superonline.com (qmail-ldap-1.03) with SMTP for <[email protected]>; 10 May 2004 17:34:14 -0000
Received: from [172.16.7.8] (helo=sirius.home)

	by ns1.murom.ru with smtp (Exim 4.30)
	id 1BNEfC-0002tR-7Z
	for [email protected]; Mon, 10 May 2004 21:34:02 +0400

From: Sergey Vlasov <[email protected]>
Subject: Re: OOPS : 2.4.27-pre2 + latest ACPI
Date: Mon, 10 May 2004 21:34:02 +0400
User-Agent: Pan/0.14.2 (This is not a psychotic episode. It's a cleansing moment of clarity.)
Message-Id: <[email protected]>
References: <[email protected]> <[email protected]> <[email protected]>
To: "O.Sezer" <[email protected]>
Cc: [email protected]
MIME-Version: 1.0
Content-Type: text/plain; charset=KOI8-R
Content-Transfer-Encoding: 8bit
X-MailScanner-Information: Please contact the ISP for more information
X-MailScanner: Found to be clean
X-MailScanner-From: [email protected]
X-Spam-Checker-Version: SpamAssassin 2.60-superonline (1.212-2003-09-23-exp)
on vfep03
X-Spam-Version: SpamAssassin Superonline
X-Spam-Status: No, hits=0.0 required=40.0 tests=BAYES_50 autolearn=no
version=2.60-superonline
X-Spam-Level:

Looks like a free memory access - at acpi_button_remove() button->handle was trashed by 0x5a5a5a5a.

Does the patch below (on top of the new ACPI changes) fix this? Seems that the special handling for fixed-feature buttons is unneeded at least for 2.4.x - acpi_bus_unregister_driver() works for them.

@@ -69,8 +69,6 @@
    -------------------------------------------------------------------------- */
 
 static struct proc_dir_entry	*acpi_button_dir;
-extern struct acpi_device 	*acpi_fixed_pwr_button;
-extern struct acpi_device	*acpi_fixed_sleep_button;
 
 static int
 acpi_button_read_info (

@@ -514,12 +512,6 @@
 {
 	ACPI_FUNCTION_TRACE("acpi_button_exit");
 
-	if(acpi_fixed_pwr_button) 
-		acpi_button_remove(acpi_fixed_pwr_button, ACPI_BUS_TYPE_POWER_BUTTON);
-
-	if(acpi_fixed_sleep_button)
-		acpi_button_remove(acpi_fixed_sleep_button, ACPI_BUS_TYPE_SLEEP_BUTTON);
-
 	acpi_bus_unregister_driver(&acpi_button_driver);
 
 	remove_proc_entry(ACPI_BUTTON_CLASS, acpi_root_dir);

@@ -1769,23 +1769,15 @@
 }
 
 
-struct acpi_device *acpi_fixed_pwr_button;
-struct acpi_device *acpi_fixed_sleep_button;
-
-EXPORT_SYMBOL(acpi_fixed_pwr_button);
-EXPORT_SYMBOL(acpi_fixed_sleep_button);
-
 static int
 acpi_bus_scan_fixed (
 	struct acpi_device	*root)
 {
 	int			result = 0;
+	struct acpi_device	*device = NULL;
 
 	ACPI_FUNCTION_TRACE("acpi_bus_scan_fixed");
 
-	acpi_fixed_pwr_button = NULL;
-	acpi_fixed_sleep_button = NULL;
-
 	if (!root)
 		return_VALUE(-ENODEV);
 

@@ -1793,11 +1785,11 @@
 	 * Enumerate all fixed-feature devices.
 	 */
 	if (acpi_fadt.pwr_button == 0)
-		result = acpi_bus_add(&acpi_fixed_pwr_button, acpi_root, 
+		result = acpi_bus_add(&device, acpi_root, 
 			NULL, ACPI_BUS_TYPE_POWER_BUTTON);
 
 	if (acpi_fadt.sleep_button == 0)
-		result = acpi_bus_add(&acpi_fixed_sleep_button, acpi_root, 
+		result = acpi_bus_add(&device, acpi_root, 
 			NULL, ACPI_BUS_TYPE_SLEEP_BUTTON);
 
 	return_VALUE(result);

@@ -14,7 +14,7 @@
 
 EXTRA_CFLAGS	+= $(ACPI_CFLAGS)
 
-export-objs 	:= acpi_ksyms.o processor.o bus.o
+export-objs 	:= acpi_ksyms.o processor.o
 
 obj-$(CONFIG_ACPI)	:= acpi_ksyms.o 
 

O.Sezer,

I prefer it to come through Len, who is maintaining on ACPI.

Thanks for pushing this.