Skip to contentSkip to conversation

x86: Add nodmi command line option to turn off DMI parsing.

PATCH In review Ian Campbell 2 participants 6 messages Feb 28, 2008

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 x86: Add nodmi command line option to turn off DMI parsing.

Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc H. Peter Anvin <[email protected]>
Cc Jeremy Fitzhardinge <[email protected]>
Cc Mark McLoughlin <[email protected]>
---
 drivers/firmware/dmi_scan.c |   12 ++++++++++++
 1 files changed, 12 insertions(+), 0 deletions(-)
drivers/firmware/dmi_scan.c +12 -0
index 4072449..62b6200 100644
@@ -117,6 +117,12 @@ static int __init dmi_checksum(const u8 *buf)
 static char *dmi_ident[DMI_STRING_MAX];
 static LIST_HEAD(dmi_devices);
 int dmi_available;
+static int __init setup_disable_dmi(char *str)
+{
+	dmi_available = -1;
+	return 0;
+}
+early_param("nodmi", setup_disable_dmi);
 
 /*
  *	Save a DMI string
@@ -359,6 +365,12 @@ void __init dmi_scan_machine(void)
 	char __iomem *p, *q;
 	int rc;
 
+	if (dmi_available < 0) {
+		printk(KERN_INFO "DMI disabled.\n");
+		dmi_available = 0;
+		return;
+	}
+
 	if (efi_enabled) {
 		if (efi.smbios == EFI_INVALID_TABLE_ADDR)
 			goto out;
Signature · 1 line
1.5.4.2
#1 lore raw

PATCH 2/2 x86/xen: Disable DMI parsing in Xen kernel.

Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc H. Peter Anvin <[email protected]>
Cc Jeremy Fitzhardinge <[email protected]>
Cc Mark McLoughlin <[email protected]>
---
 arch/x86/xen/enlighten.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
arch/x86/xen/enlighten.c +4 -0
index 49e5358..69e11ef 100644
@@ -25,6 +25,7 @@
 #include <linux/mm.h>
 #include <linux/page-flags.h>
 #include <linux/highmem.h>
+#include <linux/dmi.h>
 
 #include <xen/interface/xen.h>
 #include <xen/interface/physdev.h>
@@ -1209,6 +1210,9 @@ asmlinkage void __init xen_start_kernel(void)
 		? __pa(xen_start_info->mod_start) : 0;
 	boot_params.hdr.ramdisk_size = xen_start_info->mod_len;
 
+	/* Disable DMI parsing */
+	dmi_available = 1;
+
 	/* Start the world */
 	start_kernel();
 }
Signature · 1 line
1.5.4.2
#2 lore raw

Re: [PATCH 2/2] x86/xen: Disable DMI parsing in Xen kernel.

Actually scratch this patchset -- its not sufficient. For example ddcprobe in userspace causes the exact same issue vs 0xa0000.

Looks like the only sane solution will be to mark regions between 640K-1M reserved early on in boot.

Ian.

Signature · 3 lines
Ian Campbell

Put no trust in cryptic comments.
#3 lore raw

Re: [PATCH 2/2] x86/xen: Disable DMI parsing in Xen kernel.

Quoted 5 lines · Ian Campbell

Actually scratch this patchset -- its not sufficient. For example ddcprobe in userspace causes the exact same issue vs 0xa0000.

Looks like the only sane solution will be to mark regions between 640K-1M reserved early on in boot.

and as luck has it, such a patch from Alexander van Heukelum has been put into x86.git#testing just yesterday.

So ... could you try x86.git#testing - does it work out of box? Is the reservation early enough to prevent pagetables be allocated in weird places? [or am i missing something - this is about the guest kernel reserving th 640k..1M area, not the host kernel - right?]

Ingo

#4 lore raw

Re: [PATCH 2/2] x86/xen: Disable DMI parsing in Xen kernel.

Quoted 8 lines · Ingo Molnar
Quoted 5 lines · Ian Campbell

Actually scratch this patchset -- its not sufficient. For example ddcprobe in userspace causes the exact same issue vs 0xa0000.

Looks like the only sane solution will be to mark regions between 640K-1M reserved early on in boot.

Quoted 6 lines

and as luck has it, such a patch from Alexander van Heukelum has been put into x86.git#testing just yesterday.

So ... could you try x86.git#testing - does it work out of box? Is the reservation early enough to prevent pagetables be allocated in weird places?

That patch is 64 bit only and also I'm not sure what EBDA looks like to a Xen domU (I expect it doesn't exist). I'll give it a go but I'm not hopeful.

Quoted 2 lines

[or am i missing something - this is about the guest kernel
reserving th 640k..1M area, not the host kernel - right?]

Correct, the area between 640k and 1M is just normal memory for a Xen guest and the memory map reflects that so page tables and other stuff get allocated in regions which would be magic data structures on native.

Ian.

Signature · 5 lines
Ian Campbell

The woman you buy -- and she is the least expensive -- takes a great
deal of money.  The woman who gives herself takes all your time.
		-- Balzac
#5 lore raw

Re: [PATCH 2/2] x86/xen: Disable DMI parsing in Xen kernel.

Quoted 6 lines · Ian Campbell
Quoted 6 lines
Quoted 2 lines

Looks like the only sane solution will be to mark regions between 640K-1M reserved early on in boot.

and as luck has it, such a patch from Alexander van Heukelum has been put into x86.git#testing just yesterday.

Quoted 7 lines
Quoted 3 lines

So ... could you try x86.git#testing - does it work out of box? Is the reservation early enough to prevent pagetables be allocated in weird places?

That patch is 64 bit only and also I'm not sure what EBDA looks like to a Xen domU (I expect it doesn't exist). I'll give it a go but I'm not hopeful.

it was requested to be unified to 32 bit as well already - no patch yet. I'd gladly take a patch that does it - it fixing 32-bit Xen guests would be an added bonus. Perhaps this all should be done by injecting a very eary e820 filter instead of reserving it later on? 640k..1M being special is a deep PC legacy, i doubt we'll ever recover from it ;-)

Ingo