Re: [PATCH V7] mm: memory hot-add: memory can not be added to movable zone defaultly

From: Tang Chen
Date: Wed Oct 21 2015 - 05:22:40 EST



On 10/21/2015 05:18 PM, Tang Chen wrote:

On 10/21/2015 05:00 PM, Vlastimil Babka wrote:
On 10/21/2015 04:20 AM, Changsheng Liu wrote:


å 2015/10/15 0:18, Vlastimil Babka åé:
On 10/12/2015 08:58 AM, Changsheng Liu wrote:
From: Changsheng Liu <liuchangcheng@xxxxxxxxxx>

After the user config CONFIG_MOVABLE_NODE,
When the memory is hot added, should_add_memory_movable() return 0
because all zones including ZONE_MOVABLE are empty,
so the memory that was hot added will be assigned to ZONE_NORMAL
and ZONE_NORMAL will be created firstly.
But we want the whole node to be added to ZONE_MOVABLE by default.

So we change should_add_memory_movable(): if the user config
CONFIG_MOVABLE_NODE and sysctl parameter hotadd_memory_as_movable is 1
and the ZONE_NORMAL is empty or the pfn of the hot-added memory
is after the end of the ZONE_NORMAL it will always return 1
and then the whole node will be added to ZONE_MOVABLE by default.
If we want the node to be assigned to ZONE_NORMAL,
we can do it as follows:
"echo online_kernel > /sys/devices/system/memory/memoryXXX/state"

By the patch, the behavious of kernel is changed by sysctl,
user can automatically create movable memory
by only the following udev rule:
SUBSYSTEM=="memory", ACTION=="add",
ATTR{state}=="offline", ATTR{state}="online"
I'm sorry for replying you so late due to the busy business trip.
So just to be clear, we are adding a new sysctl, because the existing
movable_node kernel option, which is checked by movable_node_is_enabled(), and
does the same thing for non-hot-added-memory (?) cannot be reused for hot-added
memory, as that would be a potentially surprising behavior change? Correct? Then
this should be mentioned in the changelog too, and wherever "movable_node" is
documented should also mention the new sysctl. Personally, I would expect
movable_node to affect hot-added memory as well, and would be surprised that it
doesn't...
I think it can let the user decides when to use this feature.
The user can enable the feature when making the hot_added memory
of a node movable and
make the feature disable to assign the hot_added memory of the next
node to ZONE_NORMAL .

So you mean sysctl is more flexible than boot option. OK, but wasn't such flexibility already provided by "echo online_kernel" vs "echo online_movable"? It doesn't sound like a strong reason for a new sysctl? Not doing surprising behavior change maybe does...
.

Hi Vlastimil,

The current kernel will add hot-added memory to ZONE_NORMAL by default. If users use a udev rule as below:

SUBSYSTEM=="memory", ACTION=="add", ATTR{state}=="offline", ATTR{state}="online"

it will online the memory as normal memory, which will not be hotpluggable.

Please refer to: https://lkml.org/lkml/2015/10/9/58

I think this is the root motivation of the patch.

But BTW, I'm quite familiar with udev rules, but can something like this work ?

sorry, I'm NOT quite familiar with udev rules, ......


SUBSYSTEM=="memory", ACTION=="add", ATTR{state}=="offline", ATTR{state}="online_movable"

I'm not sure. I added Ishimatu in.

For now, I think, if the above rule works, we don't need this patch. If not, maybe we should just change the kernel behavior to make the hot-added memory be added to ZONE_MOVABLE by default.

I don't have objection. But a sysctl doesn't sound necessary.

Thanks.

--
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/
.


--
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/