Re: [PATCH] memory-hotplug: fix not enough check of valid_zones

From: Yasuaki Ishimatsu
Date: Tue Aug 26 2014 - 23:24:27 EST


(2014/08/27 10:55), Zhang Zhen wrote:
On 2014/8/26 18:23, Yasuaki Ishimatsu wrote:
(2014/08/26 18:57), Zhang Zhen wrote:
As Yasuaki Ishimatsu described the check here is not enough
if memory has hole as follows:

PFN 0x00 0xd0 0xe0 0xf0
+-------------+-------------+-------------+
zone type | Normal | hole | Normal |
+-------------+-------------+-------------+
In this case, the check can't guarantee that this is "the last
block of memory".
The check of ZONE_MOVABLE has the same problem.

Change the interface name to valid_zones according to most pepole's
suggestion.

Sample output of the sysfs files:
memory0/valid_zones: none
memory1/valid_zones: DMA32
memory2/valid_zones: DMA32
memory3/valid_zones: DMA32
memory4/valid_zones: Normal
memory5/valid_zones: Normal
memory6/valid_zones: Normal Movable
memory7/valid_zones: Movable Normal
memory8/valid_zones: Movable

The patch has two changes:
- change sysfs interface name
- change check of ZONE_MOVABLE
So please separate them.

Ok, i will separate them.

Thanks!
Signed-off-by: Zhang Zhen <zhenzhang.zhang@xxxxxxxxxx>
---
Documentation/ABI/testing/sysfs-devices-memory | 8 ++---
Documentation/memory-hotplug.txt | 4 +--
drivers/base/memory.c | 42 ++++++--------------------
3 files changed, 15 insertions(+), 39 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-devices-memory b/Documentation/ABI/testing/sysfs-devices-memory
index 2b2a1d7..deef3b5 100644
--- a/Documentation/ABI/testing/sysfs-devices-memory
+++ b/Documentation/ABI/testing/sysfs-devices-memory
@@ -61,13 +61,13 @@ Users: hotplug memory remove tools
http://www.ibm.com/developerworks/wikis/display/LinuxP/powerpc-utils


-What: /sys/devices/system/memory/memoryX/zones_online_to
+What: /sys/devices/system/memory/memoryX/valid_zones
Date: July 2014
Contact: Zhang Zhen <zhenzhang.zhang@xxxxxxxxxx>
Description:
- The file /sys/devices/system/memory/memoryX/zones_online_to
- is read-only and is designed to show which zone this memory block can
- be onlined to.
+ The file /sys/devices/system/memory/memoryX/valid_zones is
+ read-only and is designed to show which zone this memory
+ block can be onlined to.

What: /sys/devices/system/memoryX/nodeY
Date: October 2009
diff --git a/Documentation/memory-hotplug.txt b/Documentation/memory-hotplug.txt
index 5b34e33..947229c 100644
--- a/Documentation/memory-hotplug.txt
+++ b/Documentation/memory-hotplug.txt
@@ -155,7 +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/zones_online_to
+/sys/devices/system/memory/memoryXXX/valid_zones

'phys_index' : read-only and contains memory block id, same as XXX.
'state' : read-write
@@ -171,7 +171,7 @@ 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.
-'zones_online_to' : read-only: designed to show which zone this memory block
+'valid_zones' : read-only: designed to show which zone this memory block
can be onlined to.

NOTE:
diff --git a/drivers/base/memory.c b/drivers/base/memory.c
index ccaf37c..efd456c 100644
--- a/drivers/base/memory.c
+++ b/drivers/base/memory.c
@@ -374,21 +374,7 @@ static ssize_t show_phys_device(struct device *dev,
}

#ifdef CONFIG_MEMORY_HOTREMOVE
-static int __zones_online_to(unsigned long end_pfn,
- struct page *first_page, unsigned long nr_pages)
-{
- struct zone *zone_next;
-
- /* The mem block is the last block of memory. */
- if (!pfn_valid(end_pfn + 1))
- return 1;
- zone_next = page_zone(first_page + nr_pages);
- if (zone_idx(zone_next) == ZONE_MOVABLE)
- return 1;
- return 0;
-}
-
-static ssize_t show_zones_online_to(struct device *dev,
+static ssize_t show_valid_zones(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct memory_block *mem = to_memory_block(dev);
@@ -407,33 +393,23 @@ static ssize_t show_zones_online_to(struct device *dev,

zone = page_zone(first_page);

-#ifdef CONFIG_HIGHMEM
- if (zone_idx(zone) == ZONE_HIGHMEM) {
- if (__zones_online_to(end_pfn, first_page, nr_pages))
+ if (zone_idx(zone) == ZONE_MOVABLE - 1) {
+ /*The mem block is the last memoryblock of this zone.*/
+ if (end_pfn == zone_end_pfn(zone))
return sprintf(buf, "%s %s\n",
zone->name, (zone + 1)->name);
}
-#else
- if (zone_idx(zone) == ZONE_NORMAL) {
- if (__zones_online_to(end_pfn, first_page, nr_pages))
- return sprintf(buf, "%s %s\n",
- zone->name, (zone + 1)->name);
- }
-#endif

if (zone_idx(zone) == ZONE_MOVABLE) {
- if (!pfn_valid(start_pfn - nr_pages))
- return sprintf(buf, "%s %s\n",
- zone->name, (zone - 1)->name);
- zone_prev = page_zone(first_page - nr_pages);
- if (zone_idx(zone_prev) != ZONE_MOVABLE)
+ /*The mem block is the first memoryblock of ZONE_MOVABLE.*/

+ if (start_pfn == zone->zone_start_pfn)
return sprintf(buf, "%s %s\n",
- zone->name, (zone - 1)->name);
+ zone->name, (zone - 1)->name);

How about swap zone->name and (zone - 1)->name.

If swapping them, sample output of the sysfs files shows as follows:
memory0/valid_zones: none
memory1/valid_zones: DMA32
memory2/valid_zones: DMA32
memory3/valid_zones: DMA32
memory4/valid_zones: Normal
memory5/valid_zones: Normal
memory6/valid_zones: Normal Movable
memory7/valid_zones: Normal Movable

memory6/valid_zones: Normal Movable
memory7/valid_zones: Movable Normal
Here can better show the dividing line between ZONE_MOVABLE and ZONE_NORMAL.


The first column shows it's default zone,

If so, please write the information in 'valid_zones' term
of memory-hotplug.txt. Nobody know it.

Thanks,
Yasuaki Ishimatasu

for memory6:
the first column Normal shows that it can be onlined to ZONE_NORMAL by default.
echo offline > memory6/state
echo online > memory6/state
the second column Movable shows that it can be onlined to ZONE_MOVABLE by online_movable.
echo offline > memory6/state
echo online_movable > memory6/state
for memory7:
the first column Movable shows that it can be onlined to ZONE_MOVABLE by default.
echo offline > memory7/state
echo online > memory7/state
the second column Normal shows that it can be onlined to ZONE_NORMAL by online_kernel.
echo offline > memory7/state
echo online_kernel > memory7/state

And it is more convenient for script to work.
So i think we should leave it as it is.

Thanks!
~~~~~~~~~~~~~~
memory8/valid_zones: Movable

Thanks,
Yasuaki Ishimatsu

}

return sprintf(buf, "%s\n", zone->name);
}
-static DEVICE_ATTR(zones_online_to, 0444, show_zones_online_to, NULL);
+static DEVICE_ATTR(valid_zones, 0444, show_valid_zones, NULL);
#endif

static DEVICE_ATTR(phys_index, 0444, show_mem_start_phys_index, NULL);
@@ -587,7 +563,7 @@ static struct attribute *memory_memblk_attrs[] = {
&dev_attr_phys_device.attr,
&dev_attr_removable.attr,
#ifdef CONFIG_MEMORY_HOTREMOVE
- &dev_attr_zones_online_to.attr,
+ &dev_attr_valid_zones.attr,
#endif
NULL
};




.





--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/