node0 and node1 is DRAM node, node3 is a PMEM node.
$ cat /sys/devices/system/node/node1/numastat
numa_hit 646590
numa_miss 3963
numa_foreign 30651
interleave_hit 416
local_node 645252
other_node 5301
demote_src 200478
demote_dst 0
Of cousre, the userspace numastat will be extened to support these 2
new fields in the future like:
$ numastat
node0 node1 node3
numa_hit 741793 702460 364154
numa_miss 1759 8104 28893
numa_foreign 8105 30651 0
interleave_hit 653 416 0
local_node 741762 701115 0
other_node 1790 9449 393047
demote_src 163612 203828 0
demote_dst 0 0 367440
Signed-off-by: Li Zhijian <[email protected]>
---
drivers/base/node.c | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/drivers/base/node.c b/drivers/base/node.c
index 27e8502548a7..d3fc70599b6a 100644
--- a/drivers/base/node.c
+++ b/drivers/base/node.c
@@ -496,20 +496,32 @@ static DEVICE_ATTR(meminfo, 0444, node_read_meminfo, NULL);
static ssize_t node_read_numastat(struct device *dev,
struct device_attribute *attr, char *buf)
{
+ struct pglist_data *pgdat = NODE_DATA(dev->id);
+ unsigned long demote_src, demote_dst;
+
fold_vm_numa_events();
+ demote_src = node_page_state_pages(pgdat, PGDEMOTE_SRC_KSWAPD) +
+ node_page_state_pages(pgdat, PGDEMOTE_SRC_DIRECT) +
+ node_page_state_pages(pgdat, PGDEMOTE_SRC_KHUGEPAGED);
+ demote_dst = node_page_state_pages(pgdat, PGDEMOTE_DST_KSWAPD) +
+ node_page_state_pages(pgdat, PGDEMOTE_DST_DIRECT) +
+ node_page_state_pages(pgdat, PGDEMOTE_DST_KHUGEPAGED);
return sysfs_emit(buf,
"numa_hit %lu\n"
"numa_miss %lu\n"
"numa_foreign %lu\n"
"interleave_hit %lu\n"
"local_node %lu\n"
- "other_node %lu\n",
+ "other_node %lu\n"
+ "demote_src %lu\n"
+ "demote_dst %lu\n",
sum_zone_numa_event_state(dev->id, NUMA_HIT),
sum_zone_numa_event_state(dev->id, NUMA_MISS),
sum_zone_numa_event_state(dev->id, NUMA_FOREIGN),
sum_zone_numa_event_state(dev->id, NUMA_INTERLEAVE_HIT),
sum_zone_numa_event_state(dev->id, NUMA_LOCAL),
- sum_zone_numa_event_state(dev->id, NUMA_OTHER));
+ sum_zone_numa_event_state(dev->id, NUMA_OTHER),
+ demote_src, demote_dst);
}
static DEVICE_ATTR(numastat, 0444, node_read_numastat, NULL);
--
2.29.2
On Thu, Nov 02, 2023 at 10:56:48AM +0800, Li Zhijian wrote:
> node0 and node1 is DRAM node, node3 is a PMEM node.
>
> $ cat /sys/devices/system/node/node1/numastat
> numa_hit 646590
> numa_miss 3963
> numa_foreign 30651
> interleave_hit 416
> local_node 645252
> other_node 5301
> demote_src 200478
> demote_dst 0
>
> Of cousre, the userspace numastat will be extened to support these 2
> new fields in the future like:
> $ numastat
> node0 node1 node3
> numa_hit 741793 702460 364154
> numa_miss 1759 8104 28893
> numa_foreign 8105 30651 0
> interleave_hit 653 416 0
> local_node 741762 701115 0
> other_node 1790 9449 393047
> demote_src 163612 203828 0
> demote_dst 0 0 367440
>
> Signed-off-by: Li Zhijian <[email protected]>
> ---
> drivers/base/node.c | 16 ++++++++++++++--
> 1 file changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/base/node.c b/drivers/base/node.c
> index 27e8502548a7..d3fc70599b6a 100644
> --- a/drivers/base/node.c
> +++ b/drivers/base/node.c
> @@ -496,20 +496,32 @@ static DEVICE_ATTR(meminfo, 0444, node_read_meminfo, NULL);
> static ssize_t node_read_numastat(struct device *dev,
> struct device_attribute *attr, char *buf)
> {
> + struct pglist_data *pgdat = NODE_DATA(dev->id);
> + unsigned long demote_src, demote_dst;
> +
> fold_vm_numa_events();
> + demote_src = node_page_state_pages(pgdat, PGDEMOTE_SRC_KSWAPD) +
> + node_page_state_pages(pgdat, PGDEMOTE_SRC_DIRECT) +
> + node_page_state_pages(pgdat, PGDEMOTE_SRC_KHUGEPAGED);
> + demote_dst = node_page_state_pages(pgdat, PGDEMOTE_DST_KSWAPD) +
> + node_page_state_pages(pgdat, PGDEMOTE_DST_DIRECT) +
> + node_page_state_pages(pgdat, PGDEMOTE_DST_KHUGEPAGED);
> return sysfs_emit(buf,
> "numa_hit %lu\n"
> "numa_miss %lu\n"
> "numa_foreign %lu\n"
> "interleave_hit %lu\n"
> "local_node %lu\n"
> - "other_node %lu\n",
> + "other_node %lu\n"
> + "demote_src %lu\n"
> + "demote_dst %lu\n",
This sysfs file is already a total abuse of sysfs so please, do NOT make
it worse by adding more fields, that's just wrong and something I can
not take at all for obvious reasons.
thanks,
greg k-h
On 02/11/2023 13:40, Greg Kroah-Hartman wrote:
>> index 27e8502548a7..d3fc70599b6a 100644
>> --- a/drivers/base/node.c
>> +++ b/drivers/base/node.c
>> @@ -496,20 +496,32 @@ static DEVICE_ATTR(meminfo, 0444, node_read_meminfo, NULL);
>> static ssize_t node_read_numastat(struct device *dev,
>> struct device_attribute *attr, char *buf)
>> {
>> + struct pglist_data *pgdat = NODE_DATA(dev->id);
>> + unsigned long demote_src, demote_dst;
>> +
>> fold_vm_numa_events();
>> + demote_src = node_page_state_pages(pgdat, PGDEMOTE_SRC_KSWAPD) +
>> + node_page_state_pages(pgdat, PGDEMOTE_SRC_DIRECT) +
>> + node_page_state_pages(pgdat, PGDEMOTE_SRC_KHUGEPAGED);
>> + demote_dst = node_page_state_pages(pgdat, PGDEMOTE_DST_KSWAPD) +
>> + node_page_state_pages(pgdat, PGDEMOTE_DST_DIRECT) +
>> + node_page_state_pages(pgdat, PGDEMOTE_DST_KHUGEPAGED);
>> return sysfs_emit(buf,
>> "numa_hit %lu\n"
>> "numa_miss %lu\n"
>> "numa_foreign %lu\n"
>> "interleave_hit %lu\n"
>> "local_node %lu\n"
>> - "other_node %lu\n",
>> + "other_node %lu\n"
>> + "demote_src %lu\n"
>> + "demote_dst %lu\n",
> This sysfs file is already a total abuse of sysfs so please, do NOT make
> it worse by adding more fields, that's just wrong and something I can
> not take at all for obvious reasons.
>
Alright, thank you for your feedback. We will reconsider other options if necessary.
Thanks
Zhijian
> thanks,
>
> greg k-h