memory-hotplug: add sysfs zone_index attribute

PATCH In review Zhang Zhen 2 participants 4 messages Jul 24, 2014 – Jul 25, 2014 1 message not in the archive

Subsystems (4)

Memory management Documentation Driver core Memory hot(un)plug

lore

Share & tools

lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

Zhang Zhen

author

2014-07-24 07:42 UTC #1 lore raw

PATCH 1/1 memory-hotplug: add sysfs zone_index attribute

Currently memory-hotplug has two limits:
1. If the memory block is in ZONE_NORMAL, you can change it to ZONE_MOVABLE, but this memory block must be adjacent to ZONE_MOVABLE.
2. If the memory block is in ZONE_MOVABLE, you can change it to ZONE_NORMAL, but this memory block must be adjacent to ZONE_NORMAL.

Without this patch, we don't know which zone a memory block is in. So we don't know which memory block is adjacent to ZONE_MOVABLE or ZONE_NORMAL.

On the other hand, with this patch, we can easy to know newly added memory is added as ZONE_NORMAL (for powerpc, ZONE_DMA, for x86_32, ZONE_HIGHMEM).

Updated the related Documentation.

Signed-off-by Zhang Zhen <[email protected]>

---
 Documentation/ABI/testing/sysfs-devices-memory |  9 +++++++++
 Documentation/memory-hotplug.txt               |  4 +++-
 drivers/base/memory.c                          | 15 ++++++++++++++-
 3 files changed, 26 insertions(+), 2 deletions(-)

Documentation/ABI/testing/sysfs-devices-memory +9 -0

index 7405de2..39d3423 100644
@@ -61,6 +61,15 @@ Users:		hotplug memory remove tools
 		http://www.ibm.com/developerworks/wikis/display/LinuxP/powerpc-utils
 
 
+What:           /sys/devices/system/memory/memoryX/zone_index
+Date:           July 2014
+Contact:	Zhang Zhen <[email protected]>
+Description:
+		The file /sys/devices/system/memory/memoryX/zone_index
+		is read-only and is designed to show which zone this memory block is in.
+Users:		hotplug memory remove tools
+		http://www.ibm.com/developerworks/wikis/display/LinuxP/powerpc-utils			
+
 What:		/sys/devices/system/memoryX/nodeY
 Date:		October 2009
 Contact:	Linux Memory Management list <[email protected]>

Documentation/memory-hotplug.txt +3 -1

index 45134dc..07019133 100644
@@ -155,6 +155,7 @@ Under each memory block, you can see 4 files:
 /sys/devices/system/memory/memoryXXX/phys_device
 /sys/devices/system/memory/memoryXXX/state
 /sys/devices/system/memory/memoryXXX/removable
+/sys/devices/system/memory/memoryXXX/zone_index
 
 'phys_index'      : read-only and contains memory block id, same as XXX.
 'state'           : read-write

@@ -170,6 +171,8 @@ Under each memory block, you can see 4 files:
                     block is removable and a value of 0 indicates that
                     it is not removable. A memory block is removable only if
                     every section in the block is removable.
+'zone_index' 	  : read-only: designed to show which zone this memory block
+		    is in.
 
 NOTE:
   These directories/files appear after physical memory hotplug phase.

@@ -408,7 +411,6 @@ node if necessary.
   - allowing memory hot-add to ZONE_MOVABLE. maybe we need some switch like
     sysctl or new control file.
   - showing memory block and physical device relationship.
-  - showing memory block is under ZONE_MOVABLE or not
   - test and make it better memory offlining.
   - support HugeTLB page migration and offlining.
   - memmap removing at memory offline.

drivers/base/memory.c +14 -1

index 89f752d..3434d97 100644
@@ -373,11 +373,23 @@ static ssize_t show_phys_device(struct device *dev,
 	return sprintf(buf, "%d\n", mem->phys_device);
 }
 
+static ssize_t show_mem_zone_index(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct memory_block *mem = to_memory_block(dev);
+	struct page *first_page;
+	struct zone *zone;
+
+	first_page = pfn_to_page(mem->start_section_nr << PFN_SECTION_SHIFT);
+	zone = page_zone(first_page);
+	return sprintf(buf, "%s\n", zone->name);
+}
+
 static DEVICE_ATTR(phys_index, 0444, show_mem_start_phys_index, NULL);
 static DEVICE_ATTR(state, 0644, show_mem_state, store_mem_state);
 static DEVICE_ATTR(phys_device, 0444, show_phys_device, NULL);
 static DEVICE_ATTR(removable, 0444, show_mem_removable, NULL);
-
+static DEVICE_ATTR(zone_index, 0444, show_mem_zone_index, NULL);
 /*
  * Block size attribute stuff
  */

@@ -521,6 +533,7 @@ static struct attribute *memory_memblk_attrs[] = {
 	&dev_attr_state.attr,
 	&dev_attr_phys_device.attr,
 	&dev_attr_removable.attr,
+	&dev_attr_zone_index.attr,
 	NULL
 };

Signature · 4 lines

1.8.1.2


.

Dave Hansen

2014-07-24 17:59 UTC #2 lore raw

Quoted 6 lines · Zhang Zhen

Currently memory-hotplug has two limits:
1. If the memory block is in ZONE_NORMAL, you can change it to ZONE_MOVABLE, but this memory block must be adjacent to ZONE_MOVABLE.
2. If the memory block is in ZONE_MOVABLE, you can change it to ZONE_NORMAL, but this memory block must be adjacent to ZONE_NORMAL.

Quoted 7 lines

Without this patch, we don't know which zone a memory block is in. So we don't know which memory block is adjacent to ZONE_MOVABLE or ZONE_NORMAL.
On the other hand, with this patch, we can easy to know newly added memory is added as ZONE_NORMAL (for powerpc, ZONE_DMA, for x86_32, ZONE_HIGHMEM).

A section can contain more than one zone. This interface will lie about such sections, which is quite unfortunate.

I'd really much rather see an interface that has a section itself enumerate to which zones it may be changed. The way you have it now, any user has to know the rules that you've laid out above. If the kernel changed those restrictions, we'd have to teach every application about the change in restrictions.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Zhang Zhen

author

2014-07-25 02:39 UTC #3 lore raw

Quoted 11 lines · Dave Hansen

Quoted 10 lines · Zhang Zhen
Currently memory-hotplug has two limits:
1. If the memory block is in ZONE_NORMAL, you can change it to ZONE_MOVABLE, but this memory block must be adjacent to ZONE_MOVABLE.
2. If the memory block is in ZONE_MOVABLE, you can change it to ZONE_NORMAL, but this memory block must be adjacent to ZONE_NORMAL.
Without this patch, we don't know which zone a memory block is in. So we don't know which memory block is adjacent to ZONE_MOVABLE or ZONE_NORMAL.

Quoted 7 lines

Quoted 3 lines
On the other hand, with this patch, we can easy to know newly added memory is added as ZONE_NORMAL (for powerpc, ZONE_DMA, for x86_32, ZONE_HIGHMEM).
A section can contain more than one zone. This interface will lie about such sections, which is quite unfortunate.

1. In arch_add_memory(), x86_64 add the new pages of the new memory block default to ZONE_NORMAL (for powerpc, ZONE_DMA, for x86_32, ZONE_HIGHMEM).

2. In __offline_pages(), test_pages_in_a_zone() guaranteed the pages of a memory block we try to offline are in the same zone. If a section contains more than one zone, the memory block can not be offlined.

Based on the above two points, i think the pages of a memory block are in one zone, and the sections of a memory block are in one zone.

Could you please explain in detail what is the case a section can contain more than one zone ?

Thanks for your comments!

Quoted 6 lines

I'd really much rather see an interface that has a section itself enumerate to which zones it may be changed. The way you have it now, any user has to know the rules that you've laid out above. If the kernel changed those restrictions, we'd have to teach every application about the change in restrictions.

This interface is designed to show which zone a memory block is in. If the kernel changed those restrictions, this interface doesn't need to change.
For a x86_64 machine booted with "mem=400M" and with 2GiB memory installed. Sample output of the sysfs files:

# cat block_size_bytes
8000000
# cat memory0/zone_index
DMA
# cat memory1/zone_index
DMA32
# cat memory2/zone_index
DMA32
# cat memory3/zone_index
DMA32
# echo 0x20000000 > probe
# cat memory4/zone_index
Normal
# echo online > memory4/state
# cat memory4/zone_index
Normal

# echo offline > memory4/state
# echo online_movable > memory4/state
# cat memory4/zone_index
Movable

Thanks!

Best regards!

Quoted 5 lines

.

Zhang Zhen

author

2014-07-25 07:19 UTC #4 lore raw

Quoted 12 lines · Zhang Zhen

Quoted 11 lines · Dave Hansen
Quoted 10 lines · Zhang Zhen
Currently memory-hotplug has two limits:
1. If the memory block is in ZONE_NORMAL, you can change it to ZONE_MOVABLE, but this memory block must be adjacent to ZONE_MOVABLE.
2. If the memory block is in ZONE_MOVABLE, you can change it to ZONE_NORMAL, but this memory block must be adjacent to ZONE_NORMAL.
Without this patch, we don't know which zone a memory block is in. So we don't know which memory block is adjacent to ZONE_MOVABLE or ZONE_NORMAL.

Quoted 6 lines

Quoted 6 lines
Quoted 3 lines
On the other hand, with this patch, we can easy to know newly added memory is added as ZONE_NORMAL (for powerpc, ZONE_DMA, for x86_32, ZONE_HIGHMEM).
A section can contain more than one zone. This interface will lie about such sections, which is quite unfortunate.

Hi Dave,

You are right, i only considered the memory block added after machine booted. For a x86_64 machine booted with "mem=400M" and with 2GiB memory installed. Sample output of the sysfs files:

# cat block_size_bytes
8000000
# cat memory0/zone_index
DMA

Here memory0 cantain DMA_ZONE and DMA32_ZONE.

Quoted 8 lines

Quoted 1 line
1. In arch_add_memory(), x86_64 add the new pages of the new memory block default to ZONE_NORMAL (for powerpc, ZONE_DMA, for x86_32, ZONE_HIGHMEM).
2. In __offline_pages(), test_pages_in_a_zone() guaranteed the pages of a memory block we try to offline are in the same zone. If a section contains more than one zone, the memory block can not be offlined.

Quoted 13 lines

Based on the above two points, i think the pages of a memory block are in one zone, and the sections of a memory block are in one zone.
Could you please explain in detail what is the case a section can contain more than one zone ?
Thanks for your comments!
Quoted 6 lines
I'd really much rather see an interface that has a section itself enumerate to which zones it may be changed. The way you have it now, any user has to know the rules that you've laid out above. If the kernel changed those restrictions, we'd have to teach every application about the change in restrictions.

Here you are right too, we should add an interface to show which zones a memory block may be changed to. So user doesn't need to know the rules above.
I will send a new version.

Thank you very much !

Quoted 44 lines

This interface is designed to show which zone a memory block is in. If the kernel changed those restrictions, this interface doesn't need to change.
For a x86_64 machine booted with "mem=400M" and with 2GiB memory installed. Sample output of the sysfs files:
# cat block_size_bytes
8000000
# cat memory0/zone_index
DMA
# cat memory1/zone_index
DMA32
# cat memory2/zone_index
DMA32
# cat memory3/zone_index
DMA32
# echo 0x20000000 > probe
# cat memory4/zone_index
Normal
# echo online > memory4/state
# cat memory4/zone_index
Normal
# echo offline > memory4/state
# echo online_movable > memory4/state
# cat memory4/zone_index
Movable
Thanks!
Best regards!
Quoted 5 lines
.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>