Skip to contentSkip to conversation

2.6.11-mm1 (x86-abstract-discontigmem-setup.patch)

DISCUSSION Alexey Dobriyan 2 participants 5 messages Mar 5, 2005 – Mar 7, 2005

lore
Share & tools
lore
#0 lore raw
Quoted 4 lines
arch/i386/mm/discontig.c renamed from arch/i386/mm/discontig.c~x86-abstract-discontigmem-setup +1 -0
 
+void memory_present(int nid, unsigned long start, unsigned long end)

arch/i386/mm/discontig.c

Quoted 10 lines
arch/i386/mm/discontig.c renamed from arch/i386/mm/discontig.c~x86-abstract-discontigmem-setup +10 -0
+{
+	unsigned long pfn;
+
+	printk(KERN_INFO "Node: %d, start_pfn: %ld, end_pfn: %ld\n",
+			nid, start, end);
+	printk(KERN_DEBUG "  Setting physnode_map array to node %d for pfns:\n", nid);
+	printk(KERN_DEBUG "  ");
+	for (pfn = start; pfn < end; pfn += PAGES_PER_ELEMENT) {
+		physnode_map[pfn / PAGES_PER_ELEMENT] = nid;
+		printk(KERN_DEBUG "%ld ", pfn);
		       ^^^^^^^^^^

arch/i386/mm/discontig.c

Quoted 2 lines
arch/i386/mm/discontig.c renamed from arch/i386/mm/discontig.c~x86-abstract-discontigmem-setup +2 -0
+	}
+	printk(KERN_DEBUG "\n");
	       ^^^^^^^^^^
Quoted 1 line

+}

Too much KERN_DEBUG.

include/linux/mmzone.h

Quoted 12 lines
include/linux/mmzone.h renamed from include/linux/mmzone.h~x86-abstract-discontigmem-setup +8 -0
 
+#ifdef CONFIG_HAVE_MEMORY_PRESENT
+void memory_present(int nid, unsigned long start, unsigned long end);
+#else
+static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
+#endif
 
+#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
+unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
+#endif
	#else
	static inline unsigned long node_memmap_size_bytes(...);
	#endif

Is this needed?

Alexey

#1 lore raw
Quoted 6 lines · Alexey Dobriyan

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+	printk(KERN_DEBUG "\n");
	       ^^^^^^^^^^
Quoted 1 line

+}

Too much KERN_DEBUG.

On my system, that ends up printing out 4 or 5 lines of output per node, but it's quite invaluable if you're debugging early memory setup issues. It is KERN_DEBUG after all. What does it do on your system?

I'm not horribly opposed to removing some of this output, let's just make sure...

Quoted 17 lines

include/linux/mmzone.h

Quoted 12 lines
include/linux/mmzone.h renamed from include/linux/mmzone.h~x86-abstract-discontigmem-setup +8 -0
 
+#ifdef CONFIG_HAVE_MEMORY_PRESENT
+void memory_present(int nid, unsigned long start, unsigned long end);
+#else
+static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
+#endif
 
+#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
+unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
+#endif
	#else
	static inline unsigned long node_memmap_size_bytes(...);
	#endif
Quoted 1 line

Is this needed?

It's really only used for the i386 NUMA architectures, but it is necessary. We'll be overriding that discontigmem version for sparsemem, which I'll be submitting soon:

http://www.sr71.net/patches/2.6.11/2.6.11-mhp1/broken-out/B-sparse-150-sparsemem.patch

-- Dave

#2 lore raw
Quoted 6 lines · Dave Hansen
Quoted 5 lines · Alexey Dobriyan

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+	printk(KERN_DEBUG "\n");
	       ^^^^^^^^^^
Quoted 1 line

+}

Quoted 8 lines
Quoted 1 line

Too much KERN_DEBUG.

On my system, that ends up printing out 4 or 5 lines of output per node, but it's quite invaluable if you're debugging early memory setup issues. It is KERN_DEBUG after all. What does it do on your system?

I'm not horribly opposed to removing some of this output, let's just make sure...

You misundestood. I'm not proposing to remove these printk's altogether. I'm for removing KERN_DEBUG solely in the middle of the line.

Try the following program with and without 3-rd and 4-th KERN_DEBUG.

Alexey
============================================================================
#include <stdio.h>

#define KERN_DEBUG "<7>"

int main(void)

{
        int i;

        printf(KERN_DEBUG "  Setting physnode_map array to node:\n");
        printf(KERN_DEBUG "  ");
        for (i = 0; i < 10; i++)
                printf(KERN_DEBUG "%d ", i);
        printf(KERN_DEBUG "\n");

        return 0;
}
#3 lore raw

I hit send on that other one accidentally...

I believe the attached patch should fix the extra output. Compiles on my normal summit discontig configuration. Andrew, please apply some time after x86-abstract-discontigmem-setup-fix.patch in your series.

-- Dave

---

 clean-dave/arch/i386/mm/discontig.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
arch/i386/mm/discontig.c renamed from arch/i386/mm/discontig.c~x86-abstract-discontigmem-setup-fix-printk +2 -2
diff -puN arch/i386/mm/discontig.c~x86-abstract-discontigmem-setup-fix-printk arch/i386/mm/discontig.c
@@ -70,9 +70,9 @@ void memory_present(int nid, unsigned lo
 	printk(KERN_DEBUG "  ");
 	for (pfn = start; pfn < end; pfn += PAGES_PER_ELEMENT) {
 		physnode_map[pfn / PAGES_PER_ELEMENT] = nid;
-		printk(KERN_DEBUG "%ld ", pfn);
+		printk("%ld ", pfn);
 	}
-	printk(KERN_DEBUG "\n");
+	printk("\n");
 }
 
 unsigned long node_memmap_size_bytes(int nid, unsigned long start_pfn,

_

#4 lore raw
In reply to Alexey Dobriyan
Quoted 13 lines · Alexey Dobriyan
Quoted 12 lines · Dave Hansen
Quoted 6 lines · Alexey Dobriyan

Review

Quoted 2 lines
(quoted diff) +2 -0
+	}
+	printk(KERN_DEBUG "\n");
	       ^^^^^^^^^^
Quoted 1 line

+}

Too much KERN_DEBUG.

On my system, that ends up printing out 4 or 5 lines of output per node, but it's quite invaluable if you're debugging early memory setup issues. It is KERN_DEBUG after all. What does it do on your system?

Quoted 7 lines
Quoted 2 lines

I'm not horribly opposed to removing some of this output, let's just make sure...

You misundestood. I'm not proposing to remove these printk's altogether. I'm for removing KERN_DEBUG solely in the middle of the line.

Try the following program with and without 3-rd and 4-th KERN_DEBUG.

Yep, I misunderstood :)

-- Dave