Re: [PATCH v4 1/2] fs/proc: optimize exactly register one ctl_table

[Xiaoming Ni]

On 2022/3/3 15:08, Meng Tang wrote:
> Sysctls are being moved out of kernel/sysctl.c and out to
> their own respective subsystems / users to help with easier
> maintance and avoid merge conflicts. But when we move just
> one entry and to its own new file the last entry for this
> new file must be empty, so we are essentialy bloating the
> kernel one extra empty entry per each newly moved sysctl.
>
> To help with this, this adds support for registering just
> one ctl_table, therefore not bloating the kernel when we
> move a single ctl_table to its own file.
>
> Since the process of registering just one single table is the
> same as that of registering an array table, so the code is
> similar to registering an array table. The difference between
> registering just one table and registering an array table is
> that we no longer traversal through pointers when registering
> a single table. These lead to that we have to add a complete
> implementation process for register just one ctl_table, so we
> have to add so much code.
>
> Suggested-by: Matthew Wilcox <willy@xxxxxxxxxxxxx>
> Signed-off-by: Meng Tang <tangmeng@xxxxxxxxxxxxx>
> ---
>   fs/proc/proc_sysctl.c  | 159 +++++++++++++++++++++++++++++------------
>   include/linux/sysctl.h |   9 ++-
>   2 files changed, 121 insertions(+), 47 deletions(-)
>
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index 6c87c99f0856..e06d2094457a 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -19,6 +19,8 @@
>   #include <linux/kmemleak.h>
>   #include "internal.h"
>   
> +#define REGISTER_SINGLE_ONE (register_single_one ? true : false)
> +
>   static const struct dentry_operations proc_sys_dentry_operations;
>   static const struct file_operations proc_sys_file_operations;
>   static const struct inode_operations proc_sys_inode_operations;
> @@ -100,8 +102,8 @@ static DEFINE_SPINLOCK(sysctl_lock);
>   static void drop_sysctl_table(struct ctl_table_header *header);
>   static int sysctl_follow_link(struct ctl_table_header **phead,
>   	struct ctl_table **pentry);
> -static int insert_links(struct ctl_table_header *head);
> -static void put_links(struct ctl_table_header *header);
> +static int insert_links(struct ctl_table_header *head, bool register_single_one);
> +static void put_links(struct ctl_table_header *header, bool register_single_one);
>   
>   static void sysctl_print_dir(struct ctl_dir *dir)
>   {
> @@ -200,7 +202,7 @@ static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry)
>   
>   static void init_header(struct ctl_table_header *head,
>   	struct ctl_table_root *root, struct ctl_table_set *set,
> -	struct ctl_node *node, struct ctl_table *table)
> +	struct ctl_node *node, struct ctl_table *table, bool register_single_one)
>   {
>   	head->ctl_table = table;
>   	head->ctl_table_arg = table;
> @@ -215,19 +217,26 @@ static void init_header(struct ctl_table_header *head,
>   	INIT_HLIST_HEAD(&head->inodes);
>   	if (node) {
>   		struct ctl_table *entry;
> -		for (entry = table; entry->procname; entry++, node++)
> +		for (entry = table; entry->procname; entry++, node++) {
>   			node->header = head;
> +			if (register_single_one)
The scalability is reduced.
If you add a file interface in the future, you need to make at least two 
code changes.

Instead of having each consumer keep the current table size in mind, you 
can obtain the table size by ARRAY_SIZE() in the API interface.

For example,

+ #define register_sysctl_init(path, table) 
__register_sysctl_init(path, table, ARRAY_SIZE(table))
...
-		for (entry = table; entry->procname; entry++, node++)
+		for (entry = table; entry->procname && num > 0; entry++, node++, num--) {


Xiaoming Ni
thanks