The memory-notify-based approach aims to handle meory-less nodes, however, it just adds
the complexity of code as pointed by David in thread [1]. The handling of memory-less
nodes is introduced by commit 4faf8d950ec4 ("hugetlb: handle memory hot-plug events").
From its commit message, we cannot find any necessity of handling this case. So, we can
simply register/unregister sysfs entries in register_node/unregister_node to simlify the
code.
BTW, hotplug callback added because in hugetlb_register_all_nodes() we register sysfs
nodes only for N_MEMORY nodes, seeing commit 9b5e5d0fdc91, which said it was a preparation
for handling memory-less nodes via memory hotplug. Since we want to remove memory hotplug,
so make sure we only register per-node sysfs for online (N_ONLINE) nodes in
hugetlb_register_all_nodes().
https://lore.kernel.org/linux-mm/[email protected]/ [1]
Suggested-by: David Hildenbrand <[email protected]>
Signed-off-by: Muchun Song <[email protected]>
---
v3:
- Fix 'struct node' is not declared reported by LTP.
v2:
- Move declaration of function related to hugetlb to hugetlb.h (David).
- Introduce hugetlb_sysfs_initialized() and call it from hugetlb_sysfs_init() (David).
- Move hugetlb_register_all_nodes() into hugetlb_sysfs_init() (David).
- Fix implicit-function-declaration reported by LKP.
- Register per-node sysfs for online (N_ONLINE) nodes instead of N_MEMORY (Aneesh).
drivers/base/node.c | 8 +++--
include/linux/hugetlb.h | 14 +++++++++
mm/hugetlb.c | 81 ++++++++++++++++++++++---------------------------
3 files changed, 57 insertions(+), 46 deletions(-)
diff --git a/drivers/base/node.c b/drivers/base/node.c
index ed391cb09999..80b1e91b9608 100644
--- a/drivers/base/node.c
+++ b/drivers/base/node.c
@@ -20,6 +20,7 @@
#include <linux/pm_runtime.h>
#include <linux/swap.h>
#include <linux/slab.h>
+#include <linux/hugetlb.h>
static struct bus_type node_subsys = {
.name = "node",
@@ -608,10 +609,12 @@ static int register_node(struct node *node, int num)
node->dev.groups = node_dev_groups;
error = device_register(&node->dev);
- if (error)
+ if (error) {
put_device(&node->dev);
- else
+ } else {
+ hugetlb_register_node(node);
compaction_register_node(node);
+ }
return error;
}
@@ -625,6 +628,7 @@ static int register_node(struct node *node, int num)
*/
void unregister_node(struct node *node)
{
+ hugetlb_unregister_node(node);
compaction_unregister_node(node);
node_remove_accesses(node);
node_remove_caches(node);
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 852f911d676e..2d98d0ac4a48 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -16,6 +16,7 @@
struct ctl_table;
struct user_struct;
struct mmu_gather;
+struct node;
#ifndef is_hugepd
typedef struct { unsigned long pd; } hugepd_t;
@@ -967,6 +968,11 @@ static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
}
#endif
+#ifdef CONFIG_NUMA
+void hugetlb_register_node(struct node *node);
+void hugetlb_unregister_node(struct node *node);
+#endif
+
#else /* CONFIG_HUGETLB_PAGE */
struct hstate {};
@@ -1141,6 +1147,14 @@ static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte)
{
}
+
+static inline void hugetlb_register_node(struct node *node)
+{
+}
+
+static inline void hugetlb_unregister_node(struct node *node)
+{
+}
#endif /* CONFIG_HUGETLB_PAGE */
static inline spinlock_t *huge_pte_lock(struct hstate *h,
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index d0617d64d718..06ac469582da 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3880,24 +3880,14 @@ static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
return 0;
}
-static void __init hugetlb_sysfs_init(void)
-{
- struct hstate *h;
- int err;
-
- hugepages_kobj = kobject_create_and_add("hugepages", mm_kobj);
- if (!hugepages_kobj)
- return;
+#ifdef CONFIG_NUMA
+static bool hugetlb_sysfs_initialized __ro_after_init;
- for_each_hstate(h) {
- err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
- hstate_kobjs, &hstate_attr_group);
- if (err)
- pr_err("HugeTLB: Unable to add hstate %s", h->name);
- }
+static inline void hugetlb_mark_sysfs_initialized(void)
+{
+ hugetlb_sysfs_initialized = true;
}
-#ifdef CONFIG_NUMA
/*
* node_hstate/s - associate per node hstate attributes, via their kobjects,
@@ -3953,7 +3943,7 @@ static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp)
* Unregister hstate attributes from a single node device.
* No-op if no hstate attributes attached.
*/
-static void hugetlb_unregister_node(struct node *node)
+void hugetlb_unregister_node(struct node *node)
{
struct hstate *h;
struct node_hstate *nhs = &node_hstates[node->dev.id];
@@ -3983,19 +3973,22 @@ static void hugetlb_unregister_node(struct node *node)
* Register hstate attributes for a single node device.
* No-op if attributes already registered.
*/
-static int hugetlb_register_node(struct node *node)
+void hugetlb_register_node(struct node *node)
{
struct hstate *h;
struct node_hstate *nhs = &node_hstates[node->dev.id];
int err;
+ if (!hugetlb_sysfs_initialized)
+ return;
+
if (nhs->hugepages_kobj)
- return 0; /* already allocated */
+ return; /* already allocated */
nhs->hugepages_kobj = kobject_create_and_add("hugepages",
&node->dev.kobj);
if (!nhs->hugepages_kobj)
- return -ENOMEM;
+ return;
for_each_hstate(h) {
err = hugetlb_sysfs_add_hstate(h, nhs->hugepages_kobj,
@@ -4005,28 +3998,9 @@ static int hugetlb_register_node(struct node *node)
pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
h->name, node->dev.id);
hugetlb_unregister_node(node);
- return -ENOMEM;
+ break;
}
}
- return 0;
-}
-
-static int __meminit hugetlb_memory_callback(struct notifier_block *self,
- unsigned long action, void *arg)
-{
- int ret = 0;
- struct memory_notify *mnb = arg;
- int nid = mnb->status_change_nid;
-
- if (nid == NUMA_NO_NODE)
- return NOTIFY_DONE;
-
- if (action == MEM_GOING_ONLINE)
- ret = hugetlb_register_node(node_devices[nid]);
- else if (action == MEM_CANCEL_ONLINE || action == MEM_OFFLINE)
- hugetlb_unregister_node(node_devices[nid]);
-
- return notifier_from_errno(ret);
}
/*
@@ -4038,11 +4012,8 @@ static void __init hugetlb_register_all_nodes(void)
{
int nid;
- get_online_mems();
- hotplug_memory_notifier(hugetlb_memory_callback, 0);
- for_each_node_state(nid, N_MEMORY)
+ for_each_online_node(nid)
hugetlb_register_node(node_devices[nid]);
- put_online_mems();
}
#else /* !CONFIG_NUMA */
@@ -4056,8 +4027,31 @@ static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp)
static void hugetlb_register_all_nodes(void) { }
+static inline void hugetlb_mark_sysfs_initialized(void)
+{
+}
#endif
+static void __init hugetlb_sysfs_init(void)
+{
+ struct hstate *h;
+ int err;
+
+ hugepages_kobj = kobject_create_and_add("hugepages", mm_kobj);
+ if (!hugepages_kobj)
+ return;
+
+ for_each_hstate(h) {
+ err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
+ hstate_kobjs, &hstate_attr_group);
+ if (err)
+ pr_err("HugeTLB: Unable to add hstate %s", h->name);
+ }
+
+ hugetlb_mark_sysfs_initialized();
+ hugetlb_register_all_nodes();
+}
+
static int __init hugetlb_init(void)
{
int i;
@@ -4112,7 +4106,6 @@ static int __init hugetlb_init(void)
report_hugepages();
hugetlb_sysfs_init();
- hugetlb_register_all_nodes();
hugetlb_cgroup_file_init();
#ifdef CONFIG_SMP
--
2.11.0
On 08.09.22 09:58, Muchun Song wrote:
> The memory-notify-based approach aims to handle meory-less nodes, however, it just adds
> the complexity of code as pointed by David in thread [1]. The handling of memory-less
> nodes is introduced by commit 4faf8d950ec4 ("hugetlb: handle memory hot-plug events").
> From its commit message, we cannot find any necessity of handling this case. So, we can
> simply register/unregister sysfs entries in register_node/unregister_node to simlify the
> code.
>
> BTW, hotplug callback added because in hugetlb_register_all_nodes() we register sysfs
> nodes only for N_MEMORY nodes, seeing commit 9b5e5d0fdc91, which said it was a preparation
> for handling memory-less nodes via memory hotplug. Since we want to remove memory hotplug,
> so make sure we only register per-node sysfs for online (N_ONLINE) nodes in
> hugetlb_register_all_nodes().
>
> https://lore.kernel.org/linux-mm/[email protected]/ [1]
> Suggested-by: David Hildenbrand <[email protected]>
> Signed-off-by: Muchun Song <[email protected]>
> ---
> v3:
> - Fix 'struct node' is not declared reported by LTP.
>
> v2:
> - Move declaration of function related to hugetlb to hugetlb.h (David).
> - Introduce hugetlb_sysfs_initialized() and call it from hugetlb_sysfs_init() (David).
> - Move hugetlb_register_all_nodes() into hugetlb_sysfs_init() (David).
> - Fix implicit-function-declaration reported by LKP.
> - Register per-node sysfs for online (N_ONLINE) nodes instead of N_MEMORY (Aneesh).
>
> drivers/base/node.c | 8 +++--
> include/linux/hugetlb.h | 14 +++++++++
> mm/hugetlb.c | 81 ++++++++++++++++++++++---------------------------
> 3 files changed, 57 insertions(+), 46 deletions(-)
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -3880,24 +3880,14 @@ static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
> return 0;
> }
>
> -static void __init hugetlb_sysfs_init(void)
> -{
> - struct hstate *h;
> - int err;
> -
> - hugepages_kobj = kobject_create_and_add("hugepages", mm_kobj);
> - if (!hugepages_kobj)
> - return;
> +#ifdef CONFIG_NUMA
> +static bool hugetlb_sysfs_initialized __ro_after_init;
>
> - for_each_hstate(h) {
> - err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
> - hstate_kobjs, &hstate_attr_group);
> - if (err)
> - pr_err("HugeTLB: Unable to add hstate %s", h->name);
> - }
> +static inline void hugetlb_mark_sysfs_initialized(void)
> +{
> + hugetlb_sysfs_initialized = true;
> }
Do we really need a separate function for this? Why not simply always
set that from hugetlb_sysfs_init() ?
I'm also not sure if we really want to optimize out one variable for
!CONFIG_NUMA.
Anyhow, in general
Acked-by: David Hildenbrand <[email protected]>
--
Thanks,
David / dhildenb
> On Sep 8, 2022, at 20:21, David Hildenbrand <[email protected]> wrote:
>
> On 08.09.22 09:58, Muchun Song wrote:
>> The memory-notify-based approach aims to handle meory-less nodes, however, it just adds
>> the complexity of code as pointed by David in thread [1]. The handling of memory-less
>> nodes is introduced by commit 4faf8d950ec4 ("hugetlb: handle memory hot-plug events").
>> From its commit message, we cannot find any necessity of handling this case. So, we can
>> simply register/unregister sysfs entries in register_node/unregister_node to simlify the
>> code.
>> BTW, hotplug callback added because in hugetlb_register_all_nodes() we register sysfs
>> nodes only for N_MEMORY nodes, seeing commit 9b5e5d0fdc91, which said it was a preparation
>> for handling memory-less nodes via memory hotplug. Since we want to remove memory hotplug,
>> so make sure we only register per-node sysfs for online (N_ONLINE) nodes in
>> hugetlb_register_all_nodes().
>> https://lore.kernel.org/linux-mm/[email protected]/ [1]
>> Suggested-by: David Hildenbrand <[email protected]>
>> Signed-off-by: Muchun Song <[email protected]>
>> ---
>> v3:
>> - Fix 'struct node' is not declared reported by LTP.
>> v2:
>> - Move declaration of function related to hugetlb to hugetlb.h (David).
>> - Introduce hugetlb_sysfs_initialized() and call it from hugetlb_sysfs_init() (David).
>> - Move hugetlb_register_all_nodes() into hugetlb_sysfs_init() (David).
>> - Fix implicit-function-declaration reported by LKP.
>> - Register per-node sysfs for online (N_ONLINE) nodes instead of N_MEMORY (Aneesh).
>> drivers/base/node.c | 8 +++--
>> include/linux/hugetlb.h | 14 +++++++++
>> mm/hugetlb.c | 81 ++++++++++++++++++++++---------------------------
>> 3 files changed, 57 insertions(+), 46 deletions(-)
>
>
>
>> --- a/mm/hugetlb.c
>> +++ b/mm/hugetlb.c
>> @@ -3880,24 +3880,14 @@ static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
>> return 0;
>> }
>> -static void __init hugetlb_sysfs_init(void)
>> -{
>> - struct hstate *h;
>> - int err;
>> -
>> - hugepages_kobj = kobject_create_and_add("hugepages", mm_kobj);
>> - if (!hugepages_kobj)
>> - return;
>> +#ifdef CONFIG_NUMA
>> +static bool hugetlb_sysfs_initialized __ro_after_init;
>> - for_each_hstate(h) {
>> - err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
>> - hstate_kobjs, &hstate_attr_group);
>> - if (err)
>> - pr_err("HugeTLB: Unable to add hstate %s", h->name);
>> - }
>> +static inline void hugetlb_mark_sysfs_initialized(void)
>> +{
>> + hugetlb_sysfs_initialized = true;
>> }
>
> Do we really need a separate function for this? Why not simply always set that from hugetlb_sysfs_init() ?
I can remove this helper.
>
> I'm also not sure if we really want to optimize out one variable for !CONFIG_NUMA.
Either is fine to me. I think the optimization does not bring any complexity.
So I’ll keep it the same in the next version unless anyone is against this.
>
> Anyhow, in general
>
> Acked-by: David Hildenbrand <[email protected]>
Thanks David.
>
>
> --
> Thanks,
>
> David / dhildenb
>