Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1473003 > unrolled thread

[PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information

Started byAnshuman Khandual <khandual@linux.vnet.ibm.com>
First post2016-08-31 05:30 +0200
Last post2016-09-02 06:40 +0200
Articles 3 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.


Contents

  [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2016-08-31 05:30 +0200
    Re: [PATCH 2/2] mm: Add sysfs interface to dump each node's  zonelist information Andrew Morton <akpm@linux-foundation.org> - 2016-08-31 23:20 +0200
      Re: [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist  information Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2016-09-02 06:40 +0200

#1473003 — [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information

FromAnshuman Khandual <khandual@linux.vnet.ibm.com>
Date2016-08-31 05:30 +0200
Subject[PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information
Message-ID<sc4nZ-5qM-5@gated-at.bofh.it>
Each individual node in the system has a ZONELIST_FALLBACK zonelist
and a ZONELIST_NOFALLBACK zonelist. These zonelists decide fallback
order of zones during memory allocations. Sometimes it helps to dump
these zonelists to see the priority order of various zones in them.
This change just adds a sysfs interface for doing the same.

Example zonelist information from a KVM guest.

[NODE (0)]
        ZONELIST_FALLBACK
        (0) (node 0) (zone DMA c00000000140c000)
        (1) (node 1) (zone DMA c000000100000000)
        (2) (node 2) (zone DMA c000000200000000)
        (3) (node 3) (zone DMA c000000300000000)
        ZONELIST_NOFALLBACK
        (0) (node 0) (zone DMA c00000000140c000)
[NODE (1)]
        ZONELIST_FALLBACK
        (0) (node 1) (zone DMA c000000100000000)
        (1) (node 2) (zone DMA c000000200000000)
        (2) (node 3) (zone DMA c000000300000000)
        (3) (node 0) (zone DMA c00000000140c000)
        ZONELIST_NOFALLBACK
        (0) (node 1) (zone DMA c000000100000000)
[NODE (2)]
        ZONELIST_FALLBACK
        (0) (node 2) (zone DMA c000000200000000)
        (1) (node 3) (zone DMA c000000300000000)
        (2) (node 0) (zone DMA c00000000140c000)
        (3) (node 1) (zone DMA c000000100000000)
        ZONELIST_NOFALLBACK
        (0) (node 2) (zone DMA c000000200000000)
[NODE (3)]
        ZONELIST_FALLBACK
        (0) (node 3) (zone DMA c000000300000000)
        (1) (node 0) (zone DMA c00000000140c000)
        (2) (node 1) (zone DMA c000000100000000)
        (3) (node 2) (zone DMA c000000200000000)
        ZONELIST_NOFALLBACK
        (0) (node 3) (zone DMA c000000300000000)

Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
---
 drivers/base/memory.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)

diff --git a/drivers/base/memory.c b/drivers/base/memory.c
index dc75de9..8c9330a 100644
--- a/drivers/base/memory.c
+++ b/drivers/base/memory.c
@@ -442,7 +442,52 @@ print_block_size(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "%lx\n", get_memory_block_size());
 }
 
+static ssize_t dump_zonelist(char *buf, struct zonelist *zonelist)
+{
+	unsigned int i;
+	ssize_t count = 0;
+
+	for (i = 0; zonelist->_zonerefs[i].zone; i++) {
+		count += sprintf(buf + count,
+			"\t\t(%d) (node %d) (%-10s %lx)\n", i,
+			zonelist->_zonerefs[i].zone->zone_pgdat->node_id,
+			zone_names[zonelist->_zonerefs[i].zone_idx],
+			(unsigned long) zonelist->_zonerefs[i].zone);
+	}
+	return count;
+}
+
+static ssize_t dump_zonelists(char *buf)
+{
+	struct zonelist *zonelist;
+	unsigned int node;
+	ssize_t count = 0;
+
+	for_each_online_node(node) {
+		zonelist = &(NODE_DATA(node)->
+				node_zonelists[ZONELIST_FALLBACK]);
+		count += sprintf(buf + count, "[NODE (%d)]\n", node);
+		count += sprintf(buf + count, "\tZONELIST_FALLBACK\n");
+		count += dump_zonelist(buf + count, zonelist);
+
+		zonelist = &(NODE_DATA(node)->
+				node_zonelists[ZONELIST_NOFALLBACK]);
+		count += sprintf(buf + count, "\tZONELIST_NOFALLBACK\n");
+		count += dump_zonelist(buf + count, zonelist);
+	}
+	return count;
+}
+
+static ssize_t
+print_system_zone_details(struct device *dev, struct device_attribute *attr,
+		 char *buf)
+{
+	return dump_zonelists(buf);
+}
+
+
 static DEVICE_ATTR(block_size_bytes, 0444, print_block_size, NULL);
+static DEVICE_ATTR(system_zone_details, 0444, print_system_zone_details, NULL);
 
 /*
  * Memory auto online policy.
@@ -783,6 +828,7 @@ static struct attribute *memory_root_attrs[] = {
 #endif
 
 	&dev_attr_block_size_bytes.attr,
+	&dev_attr_system_zone_details.attr,
 	&dev_attr_auto_online_blocks.attr,
 	NULL
 };
-- 
1.9.3

[toc] | [next] | [standalone]


#1473891 — Re: [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information

FromAndrew Morton <akpm@linux-foundation.org>
Date2016-08-31 23:20 +0200
SubjectRe: [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information
Message-ID<scl5o-7Av-17@gated-at.bofh.it>
In reply to#1473003
On Wed, 31 Aug 2016 08:55:50 +0530 Anshuman Khandual <khandual@linux.vnet.ibm.com> wrote:

> Each individual node in the system has a ZONELIST_FALLBACK zonelist
> and a ZONELIST_NOFALLBACK zonelist. These zonelists decide fallback
> order of zones during memory allocations. Sometimes it helps to dump
> these zonelists to see the priority order of various zones in them.
> This change just adds a sysfs interface for doing the same.
> 
> Example zonelist information from a KVM guest.
> 
> [NODE (0)]
>         ZONELIST_FALLBACK
>         (0) (node 0) (zone DMA c00000000140c000)
>         (1) (node 1) (zone DMA c000000100000000)
>         (2) (node 2) (zone DMA c000000200000000)
>         (3) (node 3) (zone DMA c000000300000000)
>         ZONELIST_NOFALLBACK
>         (0) (node 0) (zone DMA c00000000140c000)
> [NODE (1)]
>         ZONELIST_FALLBACK
>         (0) (node 1) (zone DMA c000000100000000)
>         (1) (node 2) (zone DMA c000000200000000)
>         (2) (node 3) (zone DMA c000000300000000)
>         (3) (node 0) (zone DMA c00000000140c000)
>         ZONELIST_NOFALLBACK
>         (0) (node 1) (zone DMA c000000100000000)
> [NODE (2)]
>         ZONELIST_FALLBACK
>         (0) (node 2) (zone DMA c000000200000000)
>         (1) (node 3) (zone DMA c000000300000000)
>         (2) (node 0) (zone DMA c00000000140c000)
>         (3) (node 1) (zone DMA c000000100000000)
>         ZONELIST_NOFALLBACK
>         (0) (node 2) (zone DMA c000000200000000)
> [NODE (3)]
>         ZONELIST_FALLBACK
>         (0) (node 3) (zone DMA c000000300000000)
>         (1) (node 0) (zone DMA c00000000140c000)
>         (2) (node 1) (zone DMA c000000100000000)
>         (3) (node 2) (zone DMA c000000200000000)
>         ZONELIST_NOFALLBACK
>         (0) (node 3) (zone DMA c000000300000000)

Can you please sell this a bit better?  Why does it "sometimes help"? 
Why does the benefit of this patch to our users justify the overhead
and cost?

Please document the full path to the sysfs file(s) within the changelog.

Please find somewhere in Documentation/ to document the new interface.

[toc] | [prev] | [next] | [standalone]


#1474885 — Re: [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information

FromAnshuman Khandual <khandual@linux.vnet.ibm.com>
Date2016-09-02 06:40 +0200
SubjectRe: [PATCH 2/2] mm: Add sysfs interface to dump each node's zonelist information
Message-ID<scOqK-36G-1@gated-at.bofh.it>
In reply to#1473891
On 09/01/2016 02:42 AM, Andrew Morton wrote:
> On Wed, 31 Aug 2016 08:55:50 +0530 Anshuman Khandual <khandual@linux.vnet.ibm.com> wrote:
> 
>> Each individual node in the system has a ZONELIST_FALLBACK zonelist
>> and a ZONELIST_NOFALLBACK zonelist. These zonelists decide fallback
>> order of zones during memory allocations. Sometimes it helps to dump
>> these zonelists to see the priority order of various zones in them.
>> This change just adds a sysfs interface for doing the same.
>>
>> Example zonelist information from a KVM guest.
>>
>> [NODE (0)]
>>         ZONELIST_FALLBACK
>>         (0) (node 0) (zone DMA c00000000140c000)
>>         (1) (node 1) (zone DMA c000000100000000)
>>         (2) (node 2) (zone DMA c000000200000000)
>>         (3) (node 3) (zone DMA c000000300000000)
>>         ZONELIST_NOFALLBACK
>>         (0) (node 0) (zone DMA c00000000140c000)
>> [NODE (1)]
>>         ZONELIST_FALLBACK
>>         (0) (node 1) (zone DMA c000000100000000)
>>         (1) (node 2) (zone DMA c000000200000000)
>>         (2) (node 3) (zone DMA c000000300000000)
>>         (3) (node 0) (zone DMA c00000000140c000)
>>         ZONELIST_NOFALLBACK
>>         (0) (node 1) (zone DMA c000000100000000)
>> [NODE (2)]
>>         ZONELIST_FALLBACK
>>         (0) (node 2) (zone DMA c000000200000000)
>>         (1) (node 3) (zone DMA c000000300000000)
>>         (2) (node 0) (zone DMA c00000000140c000)
>>         (3) (node 1) (zone DMA c000000100000000)
>>         ZONELIST_NOFALLBACK
>>         (0) (node 2) (zone DMA c000000200000000)
>> [NODE (3)]
>>         ZONELIST_FALLBACK
>>         (0) (node 3) (zone DMA c000000300000000)
>>         (1) (node 0) (zone DMA c00000000140c000)
>>         (2) (node 1) (zone DMA c000000100000000)
>>         (3) (node 2) (zone DMA c000000200000000)
>>         ZONELIST_NOFALLBACK
>>         (0) (node 3) (zone DMA c000000300000000)
> 
> Can you please sell this a bit better?  Why does it "sometimes help"?
> Why does the benefit of this patch to our users justify the overhead
> and cost?

On platforms which support memory hotplug into previously non existing
(at boot) zones, this interface helps in visualizing which zonelists
of the system, the new hot added memory ends up in. POWER is such a
platform where all the memory detected during boot time remains with
ZONE_DMA for good but then hot plug process can actually get new memory
into ZONE_MOVABLE. So having a way to get the snapshot of the zonelists
on the system after memory or node hot[un]plug is a good thing, IMHO.

> 
> Please document the full path to the sysfs file(s) within the changelog.

Sure, will do.

> 
> Please find somewhere in Documentation/ to document the new interface.
> 

Sure, will create this following file describing the interface.

Documentation/ABI/testing/sysfs-system-zone-details

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web