Re: [PATCH] Disable useless bootmem warning

Re: [PATCH] Disable useless bootmem warning

[Ben Collins]

> I've never seen a bug uncovered by this warning too. I considered to disable it 
> by passing a special array of "ok to reserve twice" regions, but on second thought 
> it is just best to remove it completely. Reserving things twice is not usually
> an error.

I have. When I was working to get sparc64 booting from alternate memory
(other than 0x0 physical), those messages helped me a lot.

Maybe make it ifdef'd by CONFIG_DEBUG_BOOTMEM (which is an option that I
know sparc and sparc64 already have).

-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/
WatchGuard - http://www.watchguard.com/

[PATCH] Disable useless bootmem warning

[Andi Kleen]

This bootmem warning has been annoying me forever:

hm, page 000f5000 reserved twice.
hm, page 000f6000 reserved twice.
hm, page 000f1000 reserved twice.
hm, page 000f2000 reserved twice.

It happens because the i386/x86-64 boot code prereserves the mptable and then later bootmem
tries to reserve it again because it's marked reserved in the e820 map.

I've never seen a bug uncovered by this warning too. I considered to disable it 
by passing a special array of "ok to reserve twice" regions, but on second thought 
it is just best to remove it completely. Reserving things twice is not usually
an error.

This patch does this.

-Andi

--- linux-2.6.3rc2-amd64/mm/bootmem.c-o	2004-02-11 22:06:58.000000000 +0100
+++ linux-2.6.3rc2-amd64/mm/bootmem.c	2004-02-16 17:52:05.000000000 +0100
@@ -91,8 +91,7 @@
 	if (end > bdata->node_low_pfn)
 		BUG();
 	for (i = sidx; i < eidx; i++)
-		if (test_and_set_bit(i, bdata->node_bootmem_map))
-			printk("hm, page %08lx reserved twice.\n", i*PAGE_SIZE);
+		set_bit(i, bdata->node_bootmem_map);
 }
 
 static void __init free_bootmem_core(bootmem_data_t *bdata, unsigned long addr, unsigned long size)

Re: [PATCH] Disable useless bootmem warning

[Andi Kleen]

On Mon, 16 Feb 2004 11:52:11 -0500
Ben Collins <bcollins@debian.org> wrote:

> > I've never seen a bug uncovered by this warning too. I considered to disable it 
> > by passing a special array of "ok to reserve twice" regions, but on second thought 
> > it is just best to remove it completely. Reserving things twice is not usually
> > an error.
> 
> I have. When I was working to get sparc64 booting from alternate memory
> (other than 0x0 physical), those messages helped me a lot.
> 
> Maybe make it ifdef'd by CONFIG_DEBUG_BOOTMEM (which is an option that I
> know sparc and sparc64 already have).

That would be fine by me too.  Anything, as long as I don't have to see them ;-)

Here's a new patch.

-Andi

diff -u linux-2.6.2-work32/mm/bootmem.c-o linux-2.6.2-work32/mm/bootmem.c
--- linux-2.6.2-work32/mm/bootmem.c-o	2004-02-11 22:06:58.000000000 +0100
+++ linux-2.6.2-work32/mm/bootmem.c	2004-02-16 20:11:10.000000000 +0100
@@ -91,8 +91,11 @@
 	if (end > bdata->node_low_pfn)
 		BUG();
 	for (i = sidx; i < eidx; i++)
-		if (test_and_set_bit(i, bdata->node_bootmem_map))
+		if (test_and_set_bit(i, bdata->node_bootmem_map)) { 
+#ifdef CONFIG_DEBUG_BOOTMEM
 			printk("hm, page %08lx reserved twice.\n", i*PAGE_SIZE);
+#endif
+		}
 }
 
 static void __init free_bootmem_core(bootmem_data_t *bdata, unsigned long addr, unsigned long size)