2002-08-13 05:48:26

by Albert Cranford

Subject: [PATCH]2.4.20 ARCH=i386 create dmi_scan.h and move decl from dmi_scan.c

Hello Marcelo,
Could you apply the following patch against 2.4.20-pre2.
Alan C. thought this would be OK to support i2c/sensors.
Thanks,
Albert
--- linux/arch/i386/kernel/dmi_scan.c.orig 2002-07-31 23:10:21.000000000 -0400
+++ linux/arch/i386/kernel/dmi_scan.c 2002-07-31 23:13:52.000000000 -0400
@@ -9,6 +9,7 @@
#include <linux/pm.h>
#include <asm/keyboard.h>
#include <asm/system.h>
+#include <asm/dmi_scan.h>
#include <linux/bootmem.h>

unsigned long dmi_broken;
@@ -127,22 +128,7 @@
return -1;
}

-
-enum
-{
- DMI_BIOS_VENDOR,
- DMI_BIOS_VERSION,
- DMI_BIOS_DATE,
- DMI_SYS_VENDOR,
- DMI_PRODUCT_NAME,
- DMI_PRODUCT_VERSION,
- DMI_BOARD_VENDOR,
- DMI_BOARD_NAME,
- DMI_BOARD_VERSION,
- DMI_STRING_MAX
-};
-
-static char *dmi_ident[DMI_STRING_MAX];
+char *dmi_ident[DMI_STRING_MAX];

/*
* Save a DMI string
--- /dev/null 1994-07-17 19:46:18.000000000 -0400
+++ linux/include/asm-i386/dmi_scan.h 2002-07-31 23:12:49.000000000 -0400
@@ -0,0 +1,19 @@
+#ifndef __i386_DMI_SCAN_H
+#define __i386_DMI_SCAN_H
+enum
+{
+ DMI_BIOS_VENDOR,
+ DMI_BIOS_VERSION,
+ DMI_BIOS_DATE,
+ DMI_SYS_VENDOR,
+ DMI_PRODUCT_NAME,
+ DMI_PRODUCT_VERSION,
+ DMI_BOARD_VENDOR,
+ DMI_BOARD_NAME,
+ DMI_BOARD_VERSION,
+ DMI_STRING_MAX
+};
+
+extern char *dmi_ident[DMI_STRING_MAX];
+
+#endif

On Fri, 2002-08-02 at 15:58, Albert Cranford wrote:
> Hello Linus,
> Alan suggested that sensors group use a dmi scanner to
> manage allow/blacklist products. In order to do this
> we need to use arch/i386/kernel/dmi_scan.c components.
>
> Could you apply the following patch to facilitate this?
> Its been tested in linux-2.5.30 with no negative impact on
> kernel and may be useful for others.


Ok by me

2002-08-13 12:11:50

by Alan

Subject: Re: [PATCH]2.4.20 ARCH=i386 create dmi_scan.h and move decl from dmi_scan.c

On Tue, 2002-08-13 at 06:50, Albert Cranford wrote:
> -
> -static char *dmi_ident[DMI_STRING_MAX];
> +char *dmi_ident[DMI_STRING_MAX];

Be very careful with this change btw. The dmi_ident strings are unmapped
at the point dmi_table() returns. That means you can only use them in
the decode callback, and so it seems odd to export them. It certainly
wants clear documentation if you are doing it that way.

What you probably should do is create some flag bits for i2c/smbus and
set those then export the flag data (we do the same for APM), rather
than export them as globals