linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [Review][PATCH][0/29] sysctl rewrite for speed and clarity
@ 2012-01-27  4:45 Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
                   ` (58 more replies)
  0 siblings, 59 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:45 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu, Damien Millescamps


The problem:

   The current implementation of sysctl has data structures and semantics
   that are hard to understand, and causes sysctl to run slowly.
   Sysctl running slowly is a scaling problem in general and the network
   stack in particular.

After some heavy testing and way way too much patch rewriting I believe I
have a patchset that seems sane and reasonable and gets to what looks to
my weary eyes a sane result.

My patchset was heavily inspired by the previous work done by Lucian
Grinjincu.  Lucian kept the need for this work alive, and did a much
better job than I ever though he did.  Unfortunately I missed a lot
of the good things Lucian did my rushed attempts at code review.

This code is also available at:
kernel.org/pub/scm/linux/kernel/git/ebiederm/sysctl.git

Eric W. Biederman (28):
      sysctl: Consolidate !CONFIG_SYSCTL handling
      sysctl: Register the base sysctl table like any other sysctl table.
      sysctl: Move the implementation into fs/proc/proc_sysctl.c
      sysctl: Make the directories have nlink == 1
      sysctl: Implement retire_sysctl_set
      sysctl: Remove the unnecessary sysctl_set parent concept.
      sysctl: Create local copies of directory names used in paths
      sysctl: Add support for register sysctl tables with a normal cstring path.
      sysctl: Add ctl_table chains into cstring paths
      sysctl: register only tables of sysctl files
      sysctl: Improve the sysctl sanity checks
      sysctl: Remove the now unused ctl_table parent field.
      sysctl: A more obvious version of grab_header.
      sysctl: Initial support for auto-unregistering sysctl tables.
      sysctl: Factor out init_header from __register_sysctl_paths
      sysctl: Factor out insert_header and erase_header
      sysctl: Normalize the root_table data structure.
      sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry.
      sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry
      sysctl: Add a root pointer to ctl_table_set
      sysctl: Stop requiring explicit management of sysctl directories
      sysctl: Add sysctl_print_dir and use it in get_subdir
      sysctl: Replace root_list with links between sysctl_table_sets.
      sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy
      sysctl: Move sysctl_check_dups into insert_header
      sysctl: Make the header lists per directory.
      sysctl: Index sysctl directories with rbtrees.
      sysctl: Add register_sysctl for normal sysctl users

Lucas De Marchi (1):
      sysctl: remove impossible condition check

 fs/proc/internal.h     |    3 +
 fs/proc/proc_sysctl.c  | 1240 +++++++++++++++++++++++++++++++++++++++++++++---
 include/linux/sysctl.h |  106 +++--
 kernel/Makefile        |    1 -
 kernel/sysctl.c        |  501 +-------------------
 kernel/sysctl_check.c  |  160 -------
 lib/Kconfig.debug      |    8 -
 net/sysctl_net.c       |   24 +-
 8 files changed, 1247 insertions(+), 796 deletions(-)


Eric


^ permalink raw reply	[flat|nested] 76+ messages in thread

* [PATCH 01/29] sysctl: remove impossible condition check
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling Eric W. Biederman
                   ` (57 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Damien Millescamps, Lucas De Marchi,
	Eric W. Biederman

From: Lucas De Marchi <lucas.demarchi@profusion.mobi>

Remove checks for conditions that will never happen. If procname is NULL
the loop would already had bailed out, so there's no need to check it
again.

At the same time this also compacts the function find_in_table() by
refactoring it to be easier to read.

Signed-off-by: Lucas De Marchi <lucas.demarchi@profusion.mobi>
Reviewed-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   14 ++------------
 1 files changed, 2 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a6b6217..d82f4a8 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -59,17 +59,11 @@ out:
 
 static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
 {
-	int len;
 	for ( ; p->procname; p++) {
-
-		if (!p->procname)
-			continue;
-
-		len = strlen(p->procname);
-		if (len != name->len)
+		if (strlen(p->procname) != name->len)
 			continue;
 
-		if (memcmp(p->procname, name->name, len) != 0)
+		if (memcmp(p->procname, name->name, name->len) != 0)
 			continue;
 
 		/* I have a match */
@@ -266,10 +260,6 @@ static int scan(struct ctl_table_header *head, ctl_table *table,
 	for (; table->procname; table++, (*pos)++) {
 		int res;
 
-		/* Can't do anything without a proc name */
-		if (!table->procname)
-			continue;
-
 		if (*pos < file->f_pos)
 			continue;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table Eric W. Biederman
                   ` (56 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

- In sysctl.h move functions only available if CONFIG_SYSCL
  is defined inside of #ifdef CONFIG_SYSCTL

- Move the stub function definitions for !CONFIG_SYSCTL
  into sysctl.h and make them static inlines.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 include/linux/sysctl.h |   95 +++++++++++++++++++++++++++++++-----------------
 kernel/sysctl.c        |   26 -------------
 2 files changed, 62 insertions(+), 59 deletions(-)

diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index bb9127d..cf3ee7f 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -937,30 +937,8 @@ enum
 struct ctl_table;
 struct nsproxy;
 struct ctl_table_root;
-
-struct ctl_table_set {
-	struct list_head list;
-	struct ctl_table_set *parent;
-	int (*is_seen)(struct ctl_table_set *);
-};
-
-extern void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
-	int (*is_seen)(struct ctl_table_set *));
-
 struct ctl_table_header;
 
-extern void sysctl_head_get(struct ctl_table_header *);
-extern void sysctl_head_put(struct ctl_table_header *);
-extern int sysctl_is_seen(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev);
-extern struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-						struct ctl_table_header *prev);
-extern void sysctl_head_finish(struct ctl_table_header *prev);
-extern int sysctl_perm(struct ctl_table_root *root,
-		struct ctl_table *table, int op);
-
 typedef struct ctl_table ctl_table;
 
 typedef int proc_handler (struct ctl_table *ctl, int write,
@@ -1023,8 +1001,6 @@ static inline void *proc_sys_poll_event(struct ctl_table_poll *poll)
 	return (void *)(unsigned long)atomic_read(&poll->event);
 }
 
-void proc_sys_poll_notify(struct ctl_table_poll *poll);
-
 #define __CTL_TABLE_POLL_INITIALIZER(name) {				\
 	.event = ATOMIC_INIT(0),					\
 	.wait = __WAIT_QUEUE_HEAD_INITIALIZER(name.wait) }
@@ -1047,15 +1023,6 @@ struct ctl_table
 	void *extra2;
 };
 
-struct ctl_table_root {
-	struct list_head root_list;
-	struct ctl_table_set default_set;
-	struct ctl_table_set *(*lookup)(struct ctl_table_root *root,
-					   struct nsproxy *namespaces);
-	int (*permissions)(struct ctl_table_root *root,
-			struct nsproxy *namespaces, struct ctl_table *table);
-};
-
 /* struct ctl_table_header is used to maintain dynamic lists of
    struct ctl_table trees. */
 struct ctl_table_header
@@ -1078,11 +1045,45 @@ struct ctl_table_header
 	struct ctl_table_header *parent;
 };
 
+struct ctl_table_set {
+	struct list_head list;
+	struct ctl_table_set *parent;
+	int (*is_seen)(struct ctl_table_set *);
+};
+
+struct ctl_table_root {
+	struct list_head root_list;
+	struct ctl_table_set default_set;
+	struct ctl_table_set *(*lookup)(struct ctl_table_root *root,
+					   struct nsproxy *namespaces);
+	int (*permissions)(struct ctl_table_root *root,
+			struct nsproxy *namespaces, struct ctl_table *table);
+};
+
 /* struct ctl_path describes where in the hierarchy a table is added */
 struct ctl_path {
 	const char *procname;
 };
 
+#ifdef CONFIG_SYSCTL
+
+void proc_sys_poll_notify(struct ctl_table_poll *poll);
+
+extern void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_set *parent,
+	int (*is_seen)(struct ctl_table_set *));
+
+extern void sysctl_head_get(struct ctl_table_header *);
+extern void sysctl_head_put(struct ctl_table_header *);
+extern int sysctl_is_seen(struct ctl_table_header *);
+extern struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *);
+extern struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev);
+extern struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
+						struct ctl_table_header *prev);
+extern void sysctl_head_finish(struct ctl_table_header *prev);
+extern int sysctl_perm(struct ctl_table_root *root,
+		struct ctl_table *table, int op);
+
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_root *root, struct nsproxy *namespaces,
@@ -1094,6 +1095,34 @@ struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 void unregister_sysctl_table(struct ctl_table_header * table);
 int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table);
 
+#else /* CONFIG_SYSCTL */
+static inline struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
+{
+	return NULL;
+}
+
+static inline struct ctl_table_header *register_sysctl_paths(
+			const struct ctl_path *path, struct ctl_table *table)
+{
+	return NULL;
+}
+
+static inline void unregister_sysctl_table(struct ctl_table_header * table)
+{
+}
+
+static inline void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_set *parent,
+	int (*is_seen)(struct ctl_table_set *))
+{
+}
+
+static inline void sysctl_head_put(struct ctl_table_header *head)
+{
+}
+
+#endif /* CONFIG_SYSCTL */
+
 #endif /* __KERNEL__ */
 
 #endif /* _LINUX_SYSCTL_H */
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index f487f25..d5bbddd 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2017,32 +2017,6 @@ void setup_sysctl_set(struct ctl_table_set *p,
 	p->is_seen = is_seen;
 }
 
-#else /* !CONFIG_SYSCTL */
-struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
-{
-	return NULL;
-}
-
-struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
-						    struct ctl_table *table)
-{
-	return NULL;
-}
-
-void unregister_sysctl_table(struct ctl_table_header * table)
-{
-}
-
-void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
-	int (*is_seen)(struct ctl_table_set *))
-{
-}
-
-void sysctl_head_put(struct ctl_table_header *head)
-{
-}
-
 #endif /* CONFIG_SYSCTL */
 
 /*
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c Eric W. Biederman
                   ` (55 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Simplify the code by treating the base sysctl table like any other
sysctl table and register it with register_sysctl_table.

To ensure this table is registered early enough to avoid problems
call sysctl_init from proc_sys_init.

Rename sysctl_net.c:sysctl_init() to net_sysctl_init() to avoid
name conflicts now that kernel/sysctl.c:sysctl_init() is no longer
static.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |    3 ++-
 include/linux/sysctl.h |    1 +
 kernel/sysctl.c        |   13 ++++---------
 net/sysctl_net.c       |    4 ++--
 4 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index d82f4a8..9d29d28 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -468,5 +468,6 @@ int __init proc_sys_init(void)
 	proc_sys_root->proc_iops = &proc_sys_dir_operations;
 	proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
 	proc_sys_root->nlink = 0;
-	return 0;
+
+	return sysctl_init();
 }
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index cf3ee7f..5e3532e 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1095,6 +1095,7 @@ struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 void unregister_sysctl_table(struct ctl_table_header * table);
 int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table);
 
+extern int sysctl_init(void);
 #else /* CONFIG_SYSCTL */
 static inline struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
 {
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index d5bbddd..ad46024 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -192,7 +192,7 @@ static int sysrq_sysctl_handler(ctl_table *table, int write,
 
 #endif
 
-static struct ctl_table root_table[];
+static struct ctl_table root_table[1];
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
 	{{.count = 1,
@@ -222,7 +222,7 @@ int sysctl_legacy_va_layout;
 
 /* The default sysctl tables: */
 
-static struct ctl_table root_table[] = {
+static struct ctl_table sysctl_base_table[] = {
 	{
 		.procname	= "kernel",
 		.mode		= 0555,
@@ -1747,17 +1747,12 @@ static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
 	}
 }
 
-static __init int sysctl_init(void)
+int __init sysctl_init(void)
 {
-	sysctl_set_parent(NULL, root_table);
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-	sysctl_check_table(current->nsproxy, root_table);
-#endif
+	register_sysctl_table(sysctl_base_table);
 	return 0;
 }
 
-core_initcall(sysctl_init);
-
 static struct ctl_table *is_branch_in(struct ctl_table *branch,
 				      struct ctl_table *table)
 {
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index e758139..a6bbee2 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -90,7 +90,7 @@ static struct pernet_operations sysctl_pernet_ops = {
 	.exit = sysctl_net_exit,
 };
 
-static __init int sysctl_init(void)
+static __init int net_sysctl_init(void)
 {
 	int ret;
 	ret = register_pernet_subsys(&sysctl_pernet_ops);
@@ -102,7 +102,7 @@ static __init int sysctl_init(void)
 out:
 	return ret;
 }
-subsys_initcall(sysctl_init);
+subsys_initcall(net_sysctl_init);
 
 struct ctl_table_header *register_net_sysctl_table(struct net *net,
 	const struct ctl_path *path, struct ctl_table *table)
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (2 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 05/29] sysctl: Make the directories have nlink == 1 Eric W. Biederman
                   ` (54 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Move the core sysctl code from kernel/sysctl.c and kernel/sysctl_check.c
into fs/proc/proc_sysctl.c.

Currently sysctl maintenance is hampered by the sysctl implementation
being split across 3 files with artificial layering between them.
Consolidate the entire sysctl implementation into 1 file so that
it is easier to see what is going on and hopefully allowing for
simpler maintenance.

For functions that are now only used in fs/proc/proc_sysctl.c remove
their declarations from sysctl.h and make them static in fs/proc/proc_sysctl.c

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/internal.h     |    3 +
 fs/proc/proc_sysctl.c  |  622 ++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/sysctl.h |   16 --
 kernel/Makefile        |    1 -
 kernel/sysctl.c        |  464 ------------------------------------
 kernel/sysctl_check.c  |  160 -------------
 6 files changed, 625 insertions(+), 641 deletions(-)
 delete mode 100644 kernel/sysctl_check.c

diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 2925775..3b5ecd9 100644
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -10,12 +10,15 @@
  */
 
 #include <linux/proc_fs.h>
+struct  ctl_table_header;
 
 extern struct proc_dir_entry proc_root;
 #ifdef CONFIG_PROC_SYSCTL
 extern int proc_sys_init(void);
+extern void sysctl_head_put(struct ctl_table_header *head);
 #else
 static inline void proc_sys_init(void) { }
+static inline void sysctl_head_put(struct ctl_table_header *head) { }
 #endif
 #ifdef CONFIG_NET
 extern int proc_net_init(void);
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 9d29d28..06e6f10 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -7,6 +7,7 @@
 #include <linux/proc_fs.h>
 #include <linux/security.h>
 #include <linux/namei.h>
+#include <linux/module.h>
 #include "internal.h"
 
 static const struct dentry_operations proc_sys_dentry_operations;
@@ -24,6 +25,209 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
 	wake_up_interruptible(&poll->wait);
 }
 
+static struct ctl_table root_table[1];
+static struct ctl_table_root sysctl_table_root;
+static struct ctl_table_header root_table_header = {
+	{{.count = 1,
+	.ctl_table = root_table,
+	.ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+	.root = &sysctl_table_root,
+	.set = &sysctl_table_root.default_set,
+};
+static struct ctl_table_root sysctl_table_root = {
+	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
+	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+};
+
+static DEFINE_SPINLOCK(sysctl_lock);
+
+/* called under sysctl_lock */
+static int use_table(struct ctl_table_header *p)
+{
+	if (unlikely(p->unregistering))
+		return 0;
+	p->used++;
+	return 1;
+}
+
+/* called under sysctl_lock */
+static void unuse_table(struct ctl_table_header *p)
+{
+	if (!--p->used)
+		if (unlikely(p->unregistering))
+			complete(p->unregistering);
+}
+
+/* called under sysctl_lock, will reacquire if has to wait */
+static void start_unregistering(struct ctl_table_header *p)
+{
+	/*
+	 * if p->used is 0, nobody will ever touch that entry again;
+	 * we'll eliminate all paths to it before dropping sysctl_lock
+	 */
+	if (unlikely(p->used)) {
+		struct completion wait;
+		init_completion(&wait);
+		p->unregistering = &wait;
+		spin_unlock(&sysctl_lock);
+		wait_for_completion(&wait);
+		spin_lock(&sysctl_lock);
+	} else {
+		/* anything non-NULL; we'll never dereference it */
+		p->unregistering = ERR_PTR(-EINVAL);
+	}
+	/*
+	 * do not remove from the list until nobody holds it; walking the
+	 * list in do_sysctl() relies on that.
+	 */
+	list_del_init(&p->ctl_entry);
+}
+
+static void sysctl_head_get(struct ctl_table_header *head)
+{
+	spin_lock(&sysctl_lock);
+	head->count++;
+	spin_unlock(&sysctl_lock);
+}
+
+void sysctl_head_put(struct ctl_table_header *head)
+{
+	spin_lock(&sysctl_lock);
+	if (!--head->count)
+		kfree_rcu(head, rcu);
+	spin_unlock(&sysctl_lock);
+}
+
+static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
+{
+	if (!head)
+		BUG();
+	spin_lock(&sysctl_lock);
+	if (!use_table(head))
+		head = ERR_PTR(-ENOENT);
+	spin_unlock(&sysctl_lock);
+	return head;
+}
+
+static void sysctl_head_finish(struct ctl_table_header *head)
+{
+	if (!head)
+		return;
+	spin_lock(&sysctl_lock);
+	unuse_table(head);
+	spin_unlock(&sysctl_lock);
+}
+
+static struct ctl_table_set *
+lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
+{
+	struct ctl_table_set *set = &root->default_set;
+	if (root->lookup)
+		set = root->lookup(root, namespaces);
+	return set;
+}
+
+static struct list_head *
+lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
+{
+	struct ctl_table_set *set = lookup_header_set(root, namespaces);
+	return &set->list;
+}
+
+static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
+						struct ctl_table_header *prev)
+{
+	struct ctl_table_root *root;
+	struct list_head *header_list;
+	struct ctl_table_header *head;
+	struct list_head *tmp;
+
+	spin_lock(&sysctl_lock);
+	if (prev) {
+		head = prev;
+		tmp = &prev->ctl_entry;
+		unuse_table(prev);
+		goto next;
+	}
+	tmp = &root_table_header.ctl_entry;
+	for (;;) {
+		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
+
+		if (!use_table(head))
+			goto next;
+		spin_unlock(&sysctl_lock);
+		return head;
+	next:
+		root = head->root;
+		tmp = tmp->next;
+		header_list = lookup_header_list(root, namespaces);
+		if (tmp != header_list)
+			continue;
+
+		do {
+			root = list_entry(root->root_list.next,
+					struct ctl_table_root, root_list);
+			if (root == &sysctl_table_root)
+				goto out;
+			header_list = lookup_header_list(root, namespaces);
+		} while (list_empty(header_list));
+		tmp = header_list->next;
+	}
+out:
+	spin_unlock(&sysctl_lock);
+	return NULL;
+}
+
+static struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
+{
+	return __sysctl_head_next(current->nsproxy, prev);
+}
+
+void register_sysctl_root(struct ctl_table_root *root)
+{
+	spin_lock(&sysctl_lock);
+	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
+	spin_unlock(&sysctl_lock);
+}
+
+/*
+ * sysctl_perm does NOT grant the superuser all rights automatically, because
+ * some sysctl variables are readonly even to root.
+ */
+
+static int test_perm(int mode, int op)
+{
+	if (!current_euid())
+		mode >>= 6;
+	else if (in_egroup_p(0))
+		mode >>= 3;
+	if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
+		return 0;
+	return -EACCES;
+}
+
+static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
+{
+	int mode;
+
+	if (root->permissions)
+		mode = root->permissions(root, current->nsproxy, table);
+	else
+		mode = table->mode;
+
+	return test_perm(mode, op);
+}
+
+static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
+{
+	for (; table->procname; table++) {
+		table->parent = parent;
+		if (table->child)
+			sysctl_set_parent(table, table->child);
+	}
+}
+
+
 static struct inode *proc_sys_make_inode(struct super_block *sb,
 		struct ctl_table_header *head, struct ctl_table *table)
 {
@@ -435,6 +639,21 @@ static int proc_sys_delete(const struct dentry *dentry)
 	return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
 }
 
+static int sysctl_is_seen(struct ctl_table_header *p)
+{
+	struct ctl_table_set *set = p->set;
+	int res;
+	spin_lock(&sysctl_lock);
+	if (p->unregistering)
+		res = 0;
+	else if (!set->is_seen)
+		res = 1;
+	else
+		res = set->is_seen(set);
+	spin_unlock(&sysctl_lock);
+	return res;
+}
+
 static int proc_sys_compare(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
@@ -460,6 +679,409 @@ static const struct dentry_operations proc_sys_dentry_operations = {
 	.d_compare	= proc_sys_compare,
 };
 
+static struct ctl_table *is_branch_in(struct ctl_table *branch,
+				      struct ctl_table *table)
+{
+	struct ctl_table *p;
+	const char *s = branch->procname;
+
+	/* branch should have named subdirectory as its first element */
+	if (!s || !branch->child)
+		return NULL;
+
+	/* ... and nothing else */
+	if (branch[1].procname)
+		return NULL;
+
+	/* table should contain subdirectory with the same name */
+	for (p = table; p->procname; p++) {
+		if (!p->child)
+			continue;
+		if (p->procname && strcmp(p->procname, s) == 0)
+			return p;
+	}
+	return NULL;
+}
+
+/* see if attaching q to p would be an improvement */
+static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
+{
+	struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
+	struct ctl_table *next;
+	int is_better = 0;
+	int not_in_parent = !p->attached_by;
+
+	while ((next = is_branch_in(by, to)) != NULL) {
+		if (by == q->attached_by)
+			is_better = 1;
+		if (to == p->attached_by)
+			not_in_parent = 1;
+		by = by->child;
+		to = next->child;
+	}
+
+	if (is_better && not_in_parent) {
+		q->attached_by = by;
+		q->attached_to = to;
+		q->parent = p;
+	}
+}
+
+#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
+static int sysctl_depth(struct ctl_table *table)
+{
+	struct ctl_table *tmp;
+	int depth;
+
+	depth = 0;
+	for (tmp = table; tmp->parent; tmp = tmp->parent)
+		depth++;
+
+	return depth;
+}
+
+static struct ctl_table *sysctl_parent(struct ctl_table *table, int n)
+{
+	int i;
+
+	for (i = 0; table && i < n; i++)
+		table = table->parent;
+
+	return table;
+}
+
+
+static void sysctl_print_path(struct ctl_table *table)
+{
+	struct ctl_table *tmp;
+	int depth, i;
+	depth = sysctl_depth(table);
+	if (table->procname) {
+		for (i = depth; i >= 0; i--) {
+			tmp = sysctl_parent(table, i);
+			printk("/%s", tmp->procname?tmp->procname:"");
+		}
+	}
+	printk(" ");
+}
+
+static struct ctl_table *sysctl_check_lookup(struct nsproxy *namespaces,
+						struct ctl_table *table)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *ref, *test;
+	int depth, cur_depth;
+
+	depth = sysctl_depth(table);
+
+	for (head = __sysctl_head_next(namespaces, NULL); head;
+	     head = __sysctl_head_next(namespaces, head)) {
+		cur_depth = depth;
+		ref = head->ctl_table;
+repeat:
+		test = sysctl_parent(table, cur_depth);
+		for (; ref->procname; ref++) {
+			int match = 0;
+			if (cur_depth && !ref->child)
+				continue;
+
+			if (test->procname && ref->procname &&
+			    (strcmp(test->procname, ref->procname) == 0))
+					match++;
+
+			if (match) {
+				if (cur_depth != 0) {
+					cur_depth--;
+					ref = ref->child;
+					goto repeat;
+				}
+				goto out;
+			}
+		}
+	}
+	ref = NULL;
+out:
+	sysctl_head_finish(head);
+	return ref;
+}
+
+static void set_fail(const char **fail, struct ctl_table *table, const char *str)
+{
+	if (*fail) {
+		printk(KERN_ERR "sysctl table check failed: ");
+		sysctl_print_path(table);
+		printk(" %s\n", *fail);
+		dump_stack();
+	}
+	*fail = str;
+}
+
+static void sysctl_check_leaf(struct nsproxy *namespaces,
+				struct ctl_table *table, const char **fail)
+{
+	struct ctl_table *ref;
+
+	ref = sysctl_check_lookup(namespaces, table);
+	if (ref && (ref != table))
+		set_fail(fail, table, "Sysctl already exists");
+}
+
+static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
+{
+	int error = 0;
+	for (; table->procname; table++) {
+		const char *fail = NULL;
+
+		if (table->parent) {
+			if (!table->parent->procname)
+				set_fail(&fail, table, "Parent without procname");
+		}
+		if (table->child) {
+			if (table->data)
+				set_fail(&fail, table, "Directory with data?");
+			if (table->maxlen)
+				set_fail(&fail, table, "Directory with maxlen?");
+			if ((table->mode & (S_IRUGO|S_IXUGO)) != table->mode)
+				set_fail(&fail, table, "Writable sysctl directory");
+			if (table->proc_handler)
+				set_fail(&fail, table, "Directory with proc_handler");
+			if (table->extra1)
+				set_fail(&fail, table, "Directory with extra1");
+			if (table->extra2)
+				set_fail(&fail, table, "Directory with extra2");
+		} else {
+			if ((table->proc_handler == proc_dostring) ||
+			    (table->proc_handler == proc_dointvec) ||
+			    (table->proc_handler == proc_dointvec_minmax) ||
+			    (table->proc_handler == proc_dointvec_jiffies) ||
+			    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
+			    (table->proc_handler == proc_dointvec_ms_jiffies) ||
+			    (table->proc_handler == proc_doulongvec_minmax) ||
+			    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
+				if (!table->data)
+					set_fail(&fail, table, "No data");
+				if (!table->maxlen)
+					set_fail(&fail, table, "No maxlen");
+			}
+#ifdef CONFIG_PROC_SYSCTL
+			if (!table->proc_handler)
+				set_fail(&fail, table, "No proc_handler");
+#endif
+			sysctl_check_leaf(namespaces, table, &fail);
+		}
+		if (table->mode > 0777)
+			set_fail(&fail, table, "bogus .mode");
+		if (fail) {
+			set_fail(&fail, table, NULL);
+			error = -EINVAL;
+		}
+		if (table->child)
+			error |= sysctl_check_table(namespaces, table->child);
+	}
+	return error;
+}
+#endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
+
+/**
+ * __register_sysctl_paths - register a sysctl hierarchy
+ * @root: List of sysctl headers to register on
+ * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * The members of the &struct ctl_table structure are used as follows:
+ *
+ * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
+ *            enter a sysctl file
+ *
+ * data - a pointer to data for use by proc_handler
+ *
+ * maxlen - the maximum size in bytes of the data
+ *
+ * mode - the file permissions for the /proc/sys file, and for sysctl(2)
+ *
+ * child - a pointer to the child sysctl table if this entry is a directory, or
+ *         %NULL.
+ *
+ * proc_handler - the text handler routine (described below)
+ *
+ * de - for internal use by the sysctl routines
+ *
+ * extra1, extra2 - extra pointers usable by the proc handler routines
+ *
+ * Leaf nodes in the sysctl tree will be represented by a single file
+ * under /proc; non-leaf nodes will be represented by directories.
+ *
+ * sysctl(2) can automatically manage read and write requests through
+ * the sysctl table.  The data and maxlen fields of the ctl_table
+ * struct enable minimal validation of the values being written to be
+ * performed, and the mode field allows minimal authentication.
+ *
+ * There must be a proc_handler routine for any terminal nodes
+ * mirrored under /proc/sys (non-terminals are handled by a built-in
+ * directory handler).  Several default handlers are available to
+ * cover common cases -
+ *
+ * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
+ * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
+ * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
+ *
+ * It is the handler's job to read the input buffer from user memory
+ * and process it. The handler should return 0 on success.
+ *
+ * This routine returns %NULL on a failure to register, and a pointer
+ * to the table header on success.
+ */
+struct ctl_table_header *__register_sysctl_paths(
+	struct ctl_table_root *root,
+	struct nsproxy *namespaces,
+	const struct ctl_path *path, struct ctl_table *table)
+{
+	struct ctl_table_header *header;
+	struct ctl_table *new, **prevp;
+	unsigned int n, npath;
+	struct ctl_table_set *set;
+
+	/* Count the path components */
+	for (npath = 0; path[npath].procname; ++npath)
+		;
+
+	/*
+	 * For each path component, allocate a 2-element ctl_table array.
+	 * The first array element will be filled with the sysctl entry
+	 * for this, the second will be the sentinel (procname == 0).
+	 *
+	 * We allocate everything in one go so that we don't have to
+	 * worry about freeing additional memory in unregister_sysctl_table.
+	 */
+	header = kzalloc(sizeof(struct ctl_table_header) +
+			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
+	if (!header)
+		return NULL;
+
+	new = (struct ctl_table *) (header + 1);
+
+	/* Now connect the dots */
+	prevp = &header->ctl_table;
+	for (n = 0; n < npath; ++n, ++path) {
+		/* Copy the procname */
+		new->procname = path->procname;
+		new->mode     = 0555;
+
+		*prevp = new;
+		prevp = &new->child;
+
+		new += 2;
+	}
+	*prevp = table;
+	header->ctl_table_arg = table;
+
+	INIT_LIST_HEAD(&header->ctl_entry);
+	header->used = 0;
+	header->unregistering = NULL;
+	header->root = root;
+	sysctl_set_parent(NULL, header->ctl_table);
+	header->count = 1;
+#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
+	if (sysctl_check_table(namespaces, header->ctl_table)) {
+		kfree(header);
+		return NULL;
+	}
+#endif
+	spin_lock(&sysctl_lock);
+	header->set = lookup_header_set(root, namespaces);
+	header->attached_by = header->ctl_table;
+	header->attached_to = root_table;
+	header->parent = &root_table_header;
+	for (set = header->set; set; set = set->parent) {
+		struct ctl_table_header *p;
+		list_for_each_entry(p, &set->list, ctl_entry) {
+			if (p->unregistering)
+				continue;
+			try_attach(p, header);
+		}
+	}
+	header->parent->count++;
+	list_add_tail(&header->ctl_entry, &header->set->list);
+	spin_unlock(&sysctl_lock);
+
+	return header;
+}
+
+/**
+ * register_sysctl_table_path - register a sysctl table hierarchy
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See __register_sysctl_paths for more details.
+ */
+struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
+						struct ctl_table *table)
+{
+	return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
+					path, table);
+}
+EXPORT_SYMBOL(register_sysctl_paths);
+
+/**
+ * register_sysctl_table - register a sysctl table hierarchy
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See register_sysctl_paths for more details.
+ */
+struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
+{
+	static const struct ctl_path null_path[] = { {} };
+
+	return register_sysctl_paths(null_path, table);
+}
+EXPORT_SYMBOL(register_sysctl_table);
+
+/**
+ * unregister_sysctl_table - unregister a sysctl table hierarchy
+ * @header: the header returned from register_sysctl_table
+ *
+ * Unregisters the sysctl table and all children. proc entries may not
+ * actually be removed until they are no longer used by anyone.
+ */
+void unregister_sysctl_table(struct ctl_table_header * header)
+{
+	might_sleep();
+
+	if (header == NULL)
+		return;
+
+	spin_lock(&sysctl_lock);
+	start_unregistering(header);
+	if (!--header->parent->count) {
+		WARN_ON(1);
+		kfree_rcu(header->parent, rcu);
+	}
+	if (!--header->count)
+		kfree_rcu(header, rcu);
+	spin_unlock(&sysctl_lock);
+}
+EXPORT_SYMBOL(unregister_sysctl_table);
+
+void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_set *parent,
+	int (*is_seen)(struct ctl_table_set *))
+{
+	INIT_LIST_HEAD(&p->list);
+	p->parent = parent ? parent : &sysctl_table_root.default_set;
+	p->is_seen = is_seen;
+}
+
+
 int __init proc_sys_init(void)
 {
 	struct proc_dir_entry *proc_sys_root;
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 5e3532e..08cabbf 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1073,17 +1073,6 @@ extern void setup_sysctl_set(struct ctl_table_set *p,
 	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *));
 
-extern void sysctl_head_get(struct ctl_table_header *);
-extern void sysctl_head_put(struct ctl_table_header *);
-extern int sysctl_is_seen(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev);
-extern struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-						struct ctl_table_header *prev);
-extern void sysctl_head_finish(struct ctl_table_header *prev);
-extern int sysctl_perm(struct ctl_table_root *root,
-		struct ctl_table *table, int op);
-
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_root *root, struct nsproxy *namespaces,
@@ -1093,7 +1082,6 @@ struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 						struct ctl_table *table);
 
 void unregister_sysctl_table(struct ctl_table_header * table);
-int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table);
 
 extern int sysctl_init(void);
 #else /* CONFIG_SYSCTL */
@@ -1118,10 +1106,6 @@ static inline void setup_sysctl_set(struct ctl_table_set *p,
 {
 }
 
-static inline void sysctl_head_put(struct ctl_table_header *head)
-{
-}
-
 #endif /* CONFIG_SYSCTL */
 
 #endif /* __KERNEL__ */
diff --git a/kernel/Makefile b/kernel/Makefile
index 2d9de86..cb41b95 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -27,7 +27,6 @@ obj-y += power/
 
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
-obj-$(CONFIG_SYSCTL_SYSCALL_CHECK) += sysctl_check.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o
 obj-y += time/
 obj-$(CONFIG_DEBUG_MUTEXES) += mutex-debug.o
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index ad46024..b774909 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -192,20 +192,6 @@ static int sysrq_sysctl_handler(ctl_table *table, int write,
 
 #endif
 
-static struct ctl_table root_table[1];
-static struct ctl_table_root sysctl_table_root;
-static struct ctl_table_header root_table_header = {
-	{{.count = 1,
-	.ctl_table = root_table,
-	.ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
-	.root = &sysctl_table_root,
-	.set = &sysctl_table_root.default_set,
-};
-static struct ctl_table_root sysctl_table_root = {
-	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
-};
-
 static struct ctl_table kern_table[];
 static struct ctl_table vm_table[];
 static struct ctl_table fs_table[];
@@ -1559,459 +1545,12 @@ static struct ctl_table dev_table[] = {
 	{ }
 };
 
-static DEFINE_SPINLOCK(sysctl_lock);
-
-/* called under sysctl_lock */
-static int use_table(struct ctl_table_header *p)
-{
-	if (unlikely(p->unregistering))
-		return 0;
-	p->used++;
-	return 1;
-}
-
-/* called under sysctl_lock */
-static void unuse_table(struct ctl_table_header *p)
-{
-	if (!--p->used)
-		if (unlikely(p->unregistering))
-			complete(p->unregistering);
-}
-
-/* called under sysctl_lock, will reacquire if has to wait */
-static void start_unregistering(struct ctl_table_header *p)
-{
-	/*
-	 * if p->used is 0, nobody will ever touch that entry again;
-	 * we'll eliminate all paths to it before dropping sysctl_lock
-	 */
-	if (unlikely(p->used)) {
-		struct completion wait;
-		init_completion(&wait);
-		p->unregistering = &wait;
-		spin_unlock(&sysctl_lock);
-		wait_for_completion(&wait);
-		spin_lock(&sysctl_lock);
-	} else {
-		/* anything non-NULL; we'll never dereference it */
-		p->unregistering = ERR_PTR(-EINVAL);
-	}
-	/*
-	 * do not remove from the list until nobody holds it; walking the
-	 * list in do_sysctl() relies on that.
-	 */
-	list_del_init(&p->ctl_entry);
-}
-
-void sysctl_head_get(struct ctl_table_header *head)
-{
-	spin_lock(&sysctl_lock);
-	head->count++;
-	spin_unlock(&sysctl_lock);
-}
-
-void sysctl_head_put(struct ctl_table_header *head)
-{
-	spin_lock(&sysctl_lock);
-	if (!--head->count)
-		kfree_rcu(head, rcu);
-	spin_unlock(&sysctl_lock);
-}
-
-struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
-{
-	if (!head)
-		BUG();
-	spin_lock(&sysctl_lock);
-	if (!use_table(head))
-		head = ERR_PTR(-ENOENT);
-	spin_unlock(&sysctl_lock);
-	return head;
-}
-
-void sysctl_head_finish(struct ctl_table_header *head)
-{
-	if (!head)
-		return;
-	spin_lock(&sysctl_lock);
-	unuse_table(head);
-	spin_unlock(&sysctl_lock);
-}
-
-static struct ctl_table_set *
-lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
-{
-	struct ctl_table_set *set = &root->default_set;
-	if (root->lookup)
-		set = root->lookup(root, namespaces);
-	return set;
-}
-
-static struct list_head *
-lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
-{
-	struct ctl_table_set *set = lookup_header_set(root, namespaces);
-	return &set->list;
-}
-
-struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-					    struct ctl_table_header *prev)
-{
-	struct ctl_table_root *root;
-	struct list_head *header_list;
-	struct ctl_table_header *head;
-	struct list_head *tmp;
-
-	spin_lock(&sysctl_lock);
-	if (prev) {
-		head = prev;
-		tmp = &prev->ctl_entry;
-		unuse_table(prev);
-		goto next;
-	}
-	tmp = &root_table_header.ctl_entry;
-	for (;;) {
-		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
-
-		if (!use_table(head))
-			goto next;
-		spin_unlock(&sysctl_lock);
-		return head;
-	next:
-		root = head->root;
-		tmp = tmp->next;
-		header_list = lookup_header_list(root, namespaces);
-		if (tmp != header_list)
-			continue;
-
-		do {
-			root = list_entry(root->root_list.next,
-					struct ctl_table_root, root_list);
-			if (root == &sysctl_table_root)
-				goto out;
-			header_list = lookup_header_list(root, namespaces);
-		} while (list_empty(header_list));
-		tmp = header_list->next;
-	}
-out:
-	spin_unlock(&sysctl_lock);
-	return NULL;
-}
-
-struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
-{
-	return __sysctl_head_next(current->nsproxy, prev);
-}
-
-void register_sysctl_root(struct ctl_table_root *root)
-{
-	spin_lock(&sysctl_lock);
-	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
-	spin_unlock(&sysctl_lock);
-}
-
-/*
- * sysctl_perm does NOT grant the superuser all rights automatically, because
- * some sysctl variables are readonly even to root.
- */
-
-static int test_perm(int mode, int op)
-{
-	if (!current_euid())
-		mode >>= 6;
-	else if (in_egroup_p(0))
-		mode >>= 3;
-	if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
-		return 0;
-	return -EACCES;
-}
-
-int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
-{
-	int mode;
-
-	if (root->permissions)
-		mode = root->permissions(root, current->nsproxy, table);
-	else
-		mode = table->mode;
-
-	return test_perm(mode, op);
-}
-
-static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
-{
-	for (; table->procname; table++) {
-		table->parent = parent;
-		if (table->child)
-			sysctl_set_parent(table, table->child);
-	}
-}
-
 int __init sysctl_init(void)
 {
 	register_sysctl_table(sysctl_base_table);
 	return 0;
 }
 
-static struct ctl_table *is_branch_in(struct ctl_table *branch,
-				      struct ctl_table *table)
-{
-	struct ctl_table *p;
-	const char *s = branch->procname;
-
-	/* branch should have named subdirectory as its first element */
-	if (!s || !branch->child)
-		return NULL;
-
-	/* ... and nothing else */
-	if (branch[1].procname)
-		return NULL;
-
-	/* table should contain subdirectory with the same name */
-	for (p = table; p->procname; p++) {
-		if (!p->child)
-			continue;
-		if (p->procname && strcmp(p->procname, s) == 0)
-			return p;
-	}
-	return NULL;
-}
-
-/* see if attaching q to p would be an improvement */
-static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
-{
-	struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
-	struct ctl_table *next;
-	int is_better = 0;
-	int not_in_parent = !p->attached_by;
-
-	while ((next = is_branch_in(by, to)) != NULL) {
-		if (by == q->attached_by)
-			is_better = 1;
-		if (to == p->attached_by)
-			not_in_parent = 1;
-		by = by->child;
-		to = next->child;
-	}
-
-	if (is_better && not_in_parent) {
-		q->attached_by = by;
-		q->attached_to = to;
-		q->parent = p;
-	}
-}
-
-/**
- * __register_sysctl_paths - register a sysctl hierarchy
- * @root: List of sysctl headers to register on
- * @namespaces: Data to compute which lists of sysctl entries are visible
- * @path: The path to the directory the sysctl table is in.
- * @table: the top-level table structure
- *
- * Register a sysctl table hierarchy. @table should be a filled in ctl_table
- * array. A completely 0 filled entry terminates the table.
- *
- * The members of the &struct ctl_table structure are used as follows:
- *
- * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
- *            enter a sysctl file
- *
- * data - a pointer to data for use by proc_handler
- *
- * maxlen - the maximum size in bytes of the data
- *
- * mode - the file permissions for the /proc/sys file, and for sysctl(2)
- *
- * child - a pointer to the child sysctl table if this entry is a directory, or
- *         %NULL.
- *
- * proc_handler - the text handler routine (described below)
- *
- * de - for internal use by the sysctl routines
- *
- * extra1, extra2 - extra pointers usable by the proc handler routines
- *
- * Leaf nodes in the sysctl tree will be represented by a single file
- * under /proc; non-leaf nodes will be represented by directories.
- *
- * sysctl(2) can automatically manage read and write requests through
- * the sysctl table.  The data and maxlen fields of the ctl_table
- * struct enable minimal validation of the values being written to be
- * performed, and the mode field allows minimal authentication.
- *
- * There must be a proc_handler routine for any terminal nodes
- * mirrored under /proc/sys (non-terminals are handled by a built-in
- * directory handler).  Several default handlers are available to
- * cover common cases -
- *
- * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
- * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
- * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
- *
- * It is the handler's job to read the input buffer from user memory
- * and process it. The handler should return 0 on success.
- *
- * This routine returns %NULL on a failure to register, and a pointer
- * to the table header on success.
- */
-struct ctl_table_header *__register_sysctl_paths(
-	struct ctl_table_root *root,
-	struct nsproxy *namespaces,
-	const struct ctl_path *path, struct ctl_table *table)
-{
-	struct ctl_table_header *header;
-	struct ctl_table *new, **prevp;
-	unsigned int n, npath;
-	struct ctl_table_set *set;
-
-	/* Count the path components */
-	for (npath = 0; path[npath].procname; ++npath)
-		;
-
-	/*
-	 * For each path component, allocate a 2-element ctl_table array.
-	 * The first array element will be filled with the sysctl entry
-	 * for this, the second will be the sentinel (procname == 0).
-	 *
-	 * We allocate everything in one go so that we don't have to
-	 * worry about freeing additional memory in unregister_sysctl_table.
-	 */
-	header = kzalloc(sizeof(struct ctl_table_header) +
-			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
-	if (!header)
-		return NULL;
-
-	new = (struct ctl_table *) (header + 1);
-
-	/* Now connect the dots */
-	prevp = &header->ctl_table;
-	for (n = 0; n < npath; ++n, ++path) {
-		/* Copy the procname */
-		new->procname = path->procname;
-		new->mode     = 0555;
-
-		*prevp = new;
-		prevp = &new->child;
-
-		new += 2;
-	}
-	*prevp = table;
-	header->ctl_table_arg = table;
-
-	INIT_LIST_HEAD(&header->ctl_entry);
-	header->used = 0;
-	header->unregistering = NULL;
-	header->root = root;
-	sysctl_set_parent(NULL, header->ctl_table);
-	header->count = 1;
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-	if (sysctl_check_table(namespaces, header->ctl_table)) {
-		kfree(header);
-		return NULL;
-	}
-#endif
-	spin_lock(&sysctl_lock);
-	header->set = lookup_header_set(root, namespaces);
-	header->attached_by = header->ctl_table;
-	header->attached_to = root_table;
-	header->parent = &root_table_header;
-	for (set = header->set; set; set = set->parent) {
-		struct ctl_table_header *p;
-		list_for_each_entry(p, &set->list, ctl_entry) {
-			if (p->unregistering)
-				continue;
-			try_attach(p, header);
-		}
-	}
-	header->parent->count++;
-	list_add_tail(&header->ctl_entry, &header->set->list);
-	spin_unlock(&sysctl_lock);
-
-	return header;
-}
-
-/**
- * register_sysctl_table_path - register a sysctl table hierarchy
- * @path: The path to the directory the sysctl table is in.
- * @table: the top-level table structure
- *
- * Register a sysctl table hierarchy. @table should be a filled in ctl_table
- * array. A completely 0 filled entry terminates the table.
- *
- * See __register_sysctl_paths for more details.
- */
-struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
-						struct ctl_table *table)
-{
-	return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
-					path, table);
-}
-
-/**
- * register_sysctl_table - register a sysctl table hierarchy
- * @table: the top-level table structure
- *
- * Register a sysctl table hierarchy. @table should be a filled in ctl_table
- * array. A completely 0 filled entry terminates the table.
- *
- * See register_sysctl_paths for more details.
- */
-struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
-{
-	static const struct ctl_path null_path[] = { {} };
-
-	return register_sysctl_paths(null_path, table);
-}
-
-/**
- * unregister_sysctl_table - unregister a sysctl table hierarchy
- * @header: the header returned from register_sysctl_table
- *
- * Unregisters the sysctl table and all children. proc entries may not
- * actually be removed until they are no longer used by anyone.
- */
-void unregister_sysctl_table(struct ctl_table_header * header)
-{
-	might_sleep();
-
-	if (header == NULL)
-		return;
-
-	spin_lock(&sysctl_lock);
-	start_unregistering(header);
-	if (!--header->parent->count) {
-		WARN_ON(1);
-		kfree_rcu(header->parent, rcu);
-	}
-	if (!--header->count)
-		kfree_rcu(header, rcu);
-	spin_unlock(&sysctl_lock);
-}
-
-int sysctl_is_seen(struct ctl_table_header *p)
-{
-	struct ctl_table_set *set = p->set;
-	int res;
-	spin_lock(&sysctl_lock);
-	if (p->unregistering)
-		res = 0;
-	else if (!set->is_seen)
-		res = 1;
-	else
-		res = set->is_seen(set);
-	spin_unlock(&sysctl_lock);
-	return res;
-}
-
-void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
-	int (*is_seen)(struct ctl_table_set *))
-{
-	INIT_LIST_HEAD(&p->list);
-	p->parent = parent ? parent : &sysctl_table_root.default_set;
-	p->is_seen = is_seen;
-}
-
 #endif /* CONFIG_SYSCTL */
 
 /*
@@ -2977,6 +2516,3 @@ EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
 EXPORT_SYMBOL(proc_dostring);
 EXPORT_SYMBOL(proc_doulongvec_minmax);
 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
-EXPORT_SYMBOL(register_sysctl_table);
-EXPORT_SYMBOL(register_sysctl_paths);
-EXPORT_SYMBOL(unregister_sysctl_table);
diff --git a/kernel/sysctl_check.c b/kernel/sysctl_check.c
deleted file mode 100644
index 362da65..0000000
--- a/kernel/sysctl_check.c
+++ /dev/null
@@ -1,160 +0,0 @@
-#include <linux/stat.h>
-#include <linux/sysctl.h>
-#include "../fs/xfs/xfs_sysctl.h"
-#include <linux/sunrpc/debug.h>
-#include <linux/string.h>
-#include <net/ip_vs.h>
-
-
-static int sysctl_depth(struct ctl_table *table)
-{
-	struct ctl_table *tmp;
-	int depth;
-
-	depth = 0;
-	for (tmp = table; tmp->parent; tmp = tmp->parent)
-		depth++;
-
-	return depth;
-}
-
-static struct ctl_table *sysctl_parent(struct ctl_table *table, int n)
-{
-	int i;
-
-	for (i = 0; table && i < n; i++)
-		table = table->parent;
-
-	return table;
-}
-
-
-static void sysctl_print_path(struct ctl_table *table)
-{
-	struct ctl_table *tmp;
-	int depth, i;
-	depth = sysctl_depth(table);
-	if (table->procname) {
-		for (i = depth; i >= 0; i--) {
-			tmp = sysctl_parent(table, i);
-			printk("/%s", tmp->procname?tmp->procname:"");
-		}
-	}
-	printk(" ");
-}
-
-static struct ctl_table *sysctl_check_lookup(struct nsproxy *namespaces,
-						struct ctl_table *table)
-{
-	struct ctl_table_header *head;
-	struct ctl_table *ref, *test;
-	int depth, cur_depth;
-
-	depth = sysctl_depth(table);
-
-	for (head = __sysctl_head_next(namespaces, NULL); head;
-	     head = __sysctl_head_next(namespaces, head)) {
-		cur_depth = depth;
-		ref = head->ctl_table;
-repeat:
-		test = sysctl_parent(table, cur_depth);
-		for (; ref->procname; ref++) {
-			int match = 0;
-			if (cur_depth && !ref->child)
-				continue;
-
-			if (test->procname && ref->procname &&
-			    (strcmp(test->procname, ref->procname) == 0))
-					match++;
-
-			if (match) {
-				if (cur_depth != 0) {
-					cur_depth--;
-					ref = ref->child;
-					goto repeat;
-				}
-				goto out;
-			}
-		}
-	}
-	ref = NULL;
-out:
-	sysctl_head_finish(head);
-	return ref;
-}
-
-static void set_fail(const char **fail, struct ctl_table *table, const char *str)
-{
-	if (*fail) {
-		printk(KERN_ERR "sysctl table check failed: ");
-		sysctl_print_path(table);
-		printk(" %s\n", *fail);
-		dump_stack();
-	}
-	*fail = str;
-}
-
-static void sysctl_check_leaf(struct nsproxy *namespaces,
-				struct ctl_table *table, const char **fail)
-{
-	struct ctl_table *ref;
-
-	ref = sysctl_check_lookup(namespaces, table);
-	if (ref && (ref != table))
-		set_fail(fail, table, "Sysctl already exists");
-}
-
-int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
-{
-	int error = 0;
-	for (; table->procname; table++) {
-		const char *fail = NULL;
-
-		if (table->parent) {
-			if (!table->parent->procname)
-				set_fail(&fail, table, "Parent without procname");
-		}
-		if (table->child) {
-			if (table->data)
-				set_fail(&fail, table, "Directory with data?");
-			if (table->maxlen)
-				set_fail(&fail, table, "Directory with maxlen?");
-			if ((table->mode & (S_IRUGO|S_IXUGO)) != table->mode)
-				set_fail(&fail, table, "Writable sysctl directory");
-			if (table->proc_handler)
-				set_fail(&fail, table, "Directory with proc_handler");
-			if (table->extra1)
-				set_fail(&fail, table, "Directory with extra1");
-			if (table->extra2)
-				set_fail(&fail, table, "Directory with extra2");
-		} else {
-			if ((table->proc_handler == proc_dostring) ||
-			    (table->proc_handler == proc_dointvec) ||
-			    (table->proc_handler == proc_dointvec_minmax) ||
-			    (table->proc_handler == proc_dointvec_jiffies) ||
-			    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
-			    (table->proc_handler == proc_dointvec_ms_jiffies) ||
-			    (table->proc_handler == proc_doulongvec_minmax) ||
-			    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
-				if (!table->data)
-					set_fail(&fail, table, "No data");
-				if (!table->maxlen)
-					set_fail(&fail, table, "No maxlen");
-			}
-#ifdef CONFIG_PROC_SYSCTL
-			if (!table->proc_handler)
-				set_fail(&fail, table, "No proc_handler");
-#endif
-			sysctl_check_leaf(namespaces, table, &fail);
-		}
-		if (table->mode > 0777)
-			set_fail(&fail, table, "bogus .mode");
-		if (fail) {
-			set_fail(&fail, table, NULL);
-			error = -EINVAL;
-		}
-		if (table->child)
-			error |= sysctl_check_table(namespaces, table->child);
-	}
-	return error;
-}
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 05/29] sysctl: Make the directories have nlink == 1
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (3 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 06/29] sysctl: Implement retire_sysctl_set Eric W. Biederman
                   ` (53 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

I goofed when I made sysctl directories have nlink == 0.
nlink == 0 means the directory has been deleted.
nlink == 1 meands a directory does not count subdirectories.

Use the default nlink == 1 for sysctl directories.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 06e6f10..f6aa751 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -253,7 +253,6 @@ static struct inode *proc_sys_make_inode(struct super_block *sb,
 		inode->i_fop = &proc_sys_file_operations;
 	} else {
 		inode->i_mode |= S_IFDIR;
-		clear_nlink(inode);
 		inode->i_op = &proc_sys_dir_operations;
 		inode->i_fop = &proc_sys_dir_file_operations;
 	}
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 06/29] sysctl: Implement retire_sysctl_set
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (4 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 05/29] sysctl: Make the directories have nlink == 1 Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept Eric W. Biederman
                   ` (52 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

This adds a small helper retire_sysctl_set to remove the intimate knowledge about
the how a sysctl_set is implemented from net/sysct_net.c

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |    4 ++++
 include/linux/sysctl.h |    1 +
 net/sysctl_net.c       |    2 +-
 3 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index f6aa751..9d8223c 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1080,6 +1080,10 @@ void setup_sysctl_set(struct ctl_table_set *p,
 	p->is_seen = is_seen;
 }
 
+void retire_sysctl_set(struct ctl_table_set *set)
+{
+	WARN_ON(!list_empty(&set->list));
+}
 
 int __init proc_sys_init(void)
 {
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 08cabbf..475ff0e 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1072,6 +1072,7 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll);
 extern void setup_sysctl_set(struct ctl_table_set *p,
 	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *));
+extern void retire_sysctl_set(struct ctl_table_set *set);
 
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_paths(
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index a6bbee2..ffd67a6 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -82,7 +82,7 @@ static int __net_init sysctl_net_init(struct net *net)
 
 static void __net_exit sysctl_net_exit(struct net *net)
 {
-	WARN_ON(!list_empty(&net->sysctls.list));
+	retire_sysctl_set(&net->sysctls);
 }
 
 static struct pernet_operations sysctl_pernet_ops = {
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (5 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 06/29] sysctl: Implement retire_sysctl_set Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 08/29] sysctl: Create local copies of directory names used in paths Eric W. Biederman
                   ` (51 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

In sysctl_net register the two networking roots in the proper order.

In register_sysctl walk the sysctl sets in the reverse order of the
sysctl roots.

Remove parent from ctl_table_set and setup_sysctl_set as it is no
longer needed.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   11 ++++++++---
 include/linux/sysctl.h |    3 ---
 net/sysctl_net.c       |    5 ++---
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 9d8223c..86d32a3 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -995,13 +995,20 @@ struct ctl_table_header *__register_sysctl_paths(
 	header->attached_by = header->ctl_table;
 	header->attached_to = root_table;
 	header->parent = &root_table_header;
-	for (set = header->set; set; set = set->parent) {
+	set = header->set;
+	root = header->root;
+	for (;;) {
 		struct ctl_table_header *p;
 		list_for_each_entry(p, &set->list, ctl_entry) {
 			if (p->unregistering)
 				continue;
 			try_attach(p, header);
 		}
+		if (root == &sysctl_table_root)
+			break;
+		root = list_entry(root->root_list.prev,
+				  struct ctl_table_root, root_list);
+		set = lookup_header_set(root, namespaces);
 	}
 	header->parent->count++;
 	list_add_tail(&header->ctl_entry, &header->set->list);
@@ -1072,11 +1079,9 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 EXPORT_SYMBOL(unregister_sysctl_table);
 
 void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *))
 {
 	INIT_LIST_HEAD(&p->list);
-	p->parent = parent ? parent : &sysctl_table_root.default_set;
 	p->is_seen = is_seen;
 }
 
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 475ff0e..43c36ac 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1047,7 +1047,6 @@ struct ctl_table_header
 
 struct ctl_table_set {
 	struct list_head list;
-	struct ctl_table_set *parent;
 	int (*is_seen)(struct ctl_table_set *);
 };
 
@@ -1070,7 +1069,6 @@ struct ctl_path {
 void proc_sys_poll_notify(struct ctl_table_poll *poll);
 
 extern void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *));
 extern void retire_sysctl_set(struct ctl_table_set *set);
 
@@ -1102,7 +1100,6 @@ static inline void unregister_sysctl_table(struct ctl_table_header * table)
 }
 
 static inline void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *))
 {
 }
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index ffd67a6..07c6b87 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -75,7 +75,6 @@ static struct ctl_table_root net_sysctl_ro_root = {
 static int __net_init sysctl_net_init(struct net *net)
 {
 	setup_sysctl_set(&net->sysctls,
-			 &net_sysctl_ro_root.default_set,
 			 is_seen);
 	return 0;
 }
@@ -96,9 +95,9 @@ static __init int net_sysctl_init(void)
 	ret = register_pernet_subsys(&sysctl_pernet_ops);
 	if (ret)
 		goto out;
-	register_sysctl_root(&net_sysctl_root);
-	setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL, NULL);
+	setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL);
 	register_sysctl_root(&net_sysctl_ro_root);
+	register_sysctl_root(&net_sysctl_root);
 out:
 	return ret;
 }
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 08/29] sysctl: Create local copies of directory names used in paths
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (6 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path Eric W. Biederman
                   ` (50 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Creating local copies of directory names is a good idea for
two reasons.
- The dynamic names used by callers must be copied into new
  strings by the callers today to ensure the strings do not
  change between register and unregister of the sysctl table.

- Sysctl directories have a potentially different lifetime
  than the time between register and unregister of any
  particular sysctl table.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   11 ++++++++---
 1 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 86d32a3..bcf60fb 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -943,10 +943,12 @@ struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table *new, **prevp;
 	unsigned int n, npath;
 	struct ctl_table_set *set;
+	size_t path_bytes = 0;
+	char *new_name;
 
 	/* Count the path components */
 	for (npath = 0; path[npath].procname; ++npath)
-		;
+		path_bytes += strlen(path[npath].procname) + 1;
 
 	/*
 	 * For each path component, allocate a 2-element ctl_table array.
@@ -956,24 +958,27 @@ struct ctl_table_header *__register_sysctl_paths(
 	 * We allocate everything in one go so that we don't have to
 	 * worry about freeing additional memory in unregister_sysctl_table.
 	 */
-	header = kzalloc(sizeof(struct ctl_table_header) +
+	header = kzalloc(sizeof(struct ctl_table_header) + path_bytes +
 			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
 	if (!header)
 		return NULL;
 
 	new = (struct ctl_table *) (header + 1);
+	new_name = (char *)(new + (2 * npath));
 
 	/* Now connect the dots */
 	prevp = &header->ctl_table;
 	for (n = 0; n < npath; ++n, ++path) {
 		/* Copy the procname */
-		new->procname = path->procname;
+		strcpy(new_name, path->procname);
+		new->procname = new_name;
 		new->mode     = 0555;
 
 		*prevp = new;
 		prevp = &new->child;
 
 		new += 2;
+		new_name += strlen(new_name) + 1;
 	}
 	*prevp = table;
 	header->ctl_table_arg = table;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (7 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 08/29] sysctl: Create local copies of directory names used in paths Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths Eric W. Biederman
                   ` (49 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Make __register_sysctl_table the core sysctl registration operation and
make it take a char * string as path.

Now that binary paths have been banished into the real of backwards
compatibility in kernel/binary_sysctl.c where they can be safely
ignored there is no longer a need to use struct ctl_path to represent
path names when registering ctl_tables.

Start the transition to using normal char * strings to represent
pathnames when registering sysctl tables.  Normal strings are easier
to deal with both in the internal sysctl implementation and for
programmers registering sysctl tables.

__register_sysctl_paths is turned into a backwards compatibility wrapper
that converts a ctl_path array into a normal char * string.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   94 ++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/sysctl.h |    3 ++
 2 files changed, 87 insertions(+), 10 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index bcf60fb..5704ff0 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -882,7 +882,7 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
 #endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
 
 /**
- * __register_sysctl_paths - register a sysctl hierarchy
+ * __register_sysctl_table - register a sysctl table
  * @root: List of sysctl headers to register on
  * @namespaces: Data to compute which lists of sysctl entries are visible
  * @path: The path to the directory the sysctl table is in.
@@ -934,21 +934,34 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
  * This routine returns %NULL on a failure to register, and a pointer
  * to the table header on success.
  */
-struct ctl_table_header *__register_sysctl_paths(
+struct ctl_table_header *__register_sysctl_table(
 	struct ctl_table_root *root,
 	struct nsproxy *namespaces,
-	const struct ctl_path *path, struct ctl_table *table)
+	const char *path, struct ctl_table *table)
 {
 	struct ctl_table_header *header;
 	struct ctl_table *new, **prevp;
-	unsigned int n, npath;
+	const char *name, *nextname;
+	unsigned int npath = 0;
 	struct ctl_table_set *set;
 	size_t path_bytes = 0;
 	char *new_name;
 
 	/* Count the path components */
-	for (npath = 0; path[npath].procname; ++npath)
-		path_bytes += strlen(path[npath].procname) + 1;
+	for (name = path; name; name = nextname) {
+		int namelen;
+		nextname = strchr(name, '/');
+		if (nextname) {
+			namelen = nextname - name;
+			nextname++;
+		} else {
+			namelen = strlen(name);
+		}
+		if (namelen == 0)
+			continue;
+		path_bytes += namelen + 1;
+		npath++;
+	}
 
 	/*
 	 * For each path component, allocate a 2-element ctl_table array.
@@ -968,9 +981,20 @@ struct ctl_table_header *__register_sysctl_paths(
 
 	/* Now connect the dots */
 	prevp = &header->ctl_table;
-	for (n = 0; n < npath; ++n, ++path) {
-		/* Copy the procname */
-		strcpy(new_name, path->procname);
+	for (name = path; name; name = nextname) {
+		int namelen;
+		nextname = strchr(name, '/');
+		if (nextname) {
+			namelen = nextname - name;
+			nextname++;
+		} else {
+			namelen = strlen(name);
+		}
+		if (namelen == 0)
+			continue;
+		memcpy(new_name, name, namelen);
+		new_name[namelen] = '\0';
+
 		new->procname = new_name;
 		new->mode     = 0555;
 
@@ -978,7 +1002,7 @@ struct ctl_table_header *__register_sysctl_paths(
 		prevp = &new->child;
 
 		new += 2;
-		new_name += strlen(new_name) + 1;
+		new_name += namelen + 1;
 	}
 	*prevp = table;
 	header->ctl_table_arg = table;
@@ -1022,6 +1046,56 @@ struct ctl_table_header *__register_sysctl_paths(
 	return header;
 }
 
+static char *append_path(const char *path, char *pos, const char *name)
+{
+	int namelen;
+	namelen = strlen(name);
+	if (((pos - path) + namelen + 2) >= PATH_MAX)
+		return NULL;
+	memcpy(pos, name, namelen);
+	pos[namelen] = '/';
+	pos[namelen + 1] = '\0';
+	pos += namelen + 1;
+	return pos;
+}
+
+/**
+ * __register_sysctl_paths - register a sysctl table hierarchy
+ * @root: List of sysctl headers to register on
+ * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See __register_sysctl_table for more details.
+ */
+struct ctl_table_header *__register_sysctl_paths(
+	struct ctl_table_root *root,
+	struct nsproxy *namespaces,
+	const struct ctl_path *path, struct ctl_table *table)
+{
+	struct ctl_table_header *header = NULL;
+	const struct ctl_path *component;
+	char *new_path, *pos;
+
+	pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL);
+	if (!new_path)
+		return NULL;
+
+	pos[0] = '\0';
+	for (component = path; component->procname; component++) {
+		pos = append_path(new_path, pos, component->procname);
+		if (!pos)
+			goto out;
+	}
+	header = __register_sysctl_table(root, namespaces, new_path, table);
+out:
+	kfree(new_path);
+	return header;
+}
+
 /**
  * register_sysctl_table_path - register a sysctl table hierarchy
  * @path: The path to the directory the sysctl table is in.
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 43c36ac..a514e0f 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1073,6 +1073,9 @@ extern void setup_sysctl_set(struct ctl_table_set *p,
 extern void retire_sysctl_set(struct ctl_table_set *set);
 
 void register_sysctl_root(struct ctl_table_root *root);
+struct ctl_table_header *__register_sysctl_table(
+	struct ctl_table_root *root, struct nsproxy *namespaces,
+	const char *path, struct ctl_table *table);
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_root *root, struct nsproxy *namespaces,
 	const struct ctl_path *path, struct ctl_table *table);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (8 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 11/29] sysctl: register only tables of sysctl files Eric W. Biederman
                   ` (48 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

For any component of table passed to __register_sysctl_paths
that actually serves as a path, add that to the cstring path
that is passed to __register_sysctl_table.

The result is that for most calls to __register_sysctl_paths
we only pass a table to __register_sysctl_table that contains
no child directories.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 5704ff0..9b91dee 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1076,6 +1076,7 @@ struct ctl_table_header *__register_sysctl_paths(
 	struct nsproxy *namespaces,
 	const struct ctl_path *path, struct ctl_table *table)
 {
+	struct ctl_table *ctl_table_arg = table;
 	struct ctl_table_header *header = NULL;
 	const struct ctl_path *component;
 	char *new_path, *pos;
@@ -1090,7 +1091,15 @@ struct ctl_table_header *__register_sysctl_paths(
 		if (!pos)
 			goto out;
 	}
+	while (table->procname && table->child && !table[1].procname) {
+		pos = append_path(new_path, pos, table->procname);
+		if (!pos)
+			goto out;
+		table = table->child;
+	}
 	header = __register_sysctl_table(root, namespaces, new_path, table);
+	if (header)
+		header->ctl_table_arg = ctl_table_arg;
 out:
 	kfree(new_path);
 	return header;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 11/29] sysctl: register only tables of sysctl files
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (9 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 12/29] sysctl: Improve the sysctl sanity checks Eric W. Biederman
                   ` (47 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Split the registration of a complex ctl_table array which may have
arbitrary numbers of directories (->child != NULL) and tables of files
into a series of simpler registrations that only register tables of files.

Graphically:

   register('dir', { + file-a
                     + file-b
                     + subdir1
                       + file-c
                     + subdir2
                       + file-d
                       + file-e })

is transformed into:
   wrapper->subheaders[0] = register('dir', {file1-a, file1-b})
   wrapper->subheaders[1] = register('dir/subdir1', {file-c})
   wrapper->subheaders[2] = register('dir/subdir2', {file-d, file-e})
   return wrapper

This guarantees that __register_sysctl_table will only see a simple
ctl_table array with all entries having (->child == NULL).

Care was taken to pass the original simple ctl_table arrays to
__register_sysctl_table whenever possible.

This change is derived from a similar patch written
by Lucrian Grijincu.

Inspired-by: Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  165 ++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/sysctl.h |    2 +-
 2 files changed, 148 insertions(+), 19 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 9b91dee..6bab2ae 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -882,7 +882,7 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
 #endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
 
 /**
- * __register_sysctl_table - register a sysctl table
+ * __register_sysctl_table - register a leaf sysctl table
  * @root: List of sysctl headers to register on
  * @namespaces: Data to compute which lists of sysctl entries are visible
  * @path: The path to the directory the sysctl table is in.
@@ -900,29 +900,19 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
  *
  * maxlen - the maximum size in bytes of the data
  *
- * mode - the file permissions for the /proc/sys file, and for sysctl(2)
+ * mode - the file permissions for the /proc/sys file
  *
- * child - a pointer to the child sysctl table if this entry is a directory, or
- *         %NULL.
+ * child - must be %NULL.
  *
  * proc_handler - the text handler routine (described below)
  *
- * de - for internal use by the sysctl routines
- *
  * extra1, extra2 - extra pointers usable by the proc handler routines
  *
  * Leaf nodes in the sysctl tree will be represented by a single file
  * under /proc; non-leaf nodes will be represented by directories.
  *
- * sysctl(2) can automatically manage read and write requests through
- * the sysctl table.  The data and maxlen fields of the ctl_table
- * struct enable minimal validation of the values being written to be
- * performed, and the mode field allows minimal authentication.
- *
- * There must be a proc_handler routine for any terminal nodes
- * mirrored under /proc/sys (non-terminals are handled by a built-in
- * directory handler).  Several default handlers are available to
- * cover common cases -
+ * There must be a proc_handler routine for any terminal nodes.
+ * Several default handlers are available to cover common cases -
  *
  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
@@ -1059,6 +1049,100 @@ static char *append_path(const char *path, char *pos, const char *name)
 	return pos;
 }
 
+static int count_subheaders(struct ctl_table *table)
+{
+	int has_files = 0;
+	int nr_subheaders = 0;
+	struct ctl_table *entry;
+
+	/* special case: no directory and empty directory */
+	if (!table || !table->procname)
+		return 1;
+
+	for (entry = table; entry->procname; entry++) {
+		if (entry->child)
+			nr_subheaders += count_subheaders(entry->child);
+		else
+			has_files = 1;
+	}
+	return nr_subheaders + has_files;
+}
+
+static int register_leaf_sysctl_tables(const char *path, char *pos,
+	struct ctl_table_header ***subheader,
+	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table *table)
+{
+	struct ctl_table *ctl_table_arg = NULL;
+	struct ctl_table *entry, *files;
+	int nr_files = 0;
+	int nr_dirs = 0;
+	int err = -ENOMEM;
+
+	for (entry = table; entry->procname; entry++) {
+		if (entry->child)
+			nr_dirs++;
+		else
+			nr_files++;
+	}
+
+	files = table;
+	/* If there are mixed files and directories we need a new table */
+	if (nr_dirs && nr_files) {
+		struct ctl_table *new;
+		files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1),
+				GFP_KERNEL);
+		if (!files)
+			goto out;
+
+		ctl_table_arg = files;
+		for (new = files, entry = table; entry->procname; entry++) {
+			if (entry->child)
+				continue;
+			*new = *entry;
+			new++;
+		}
+	}
+
+	/* Register everything except a directory full of subdirectories */
+	if (nr_files || !nr_dirs) {
+		struct ctl_table_header *header;
+		header = __register_sysctl_table(root, namespaces, path, files);
+		if (!header) {
+			kfree(ctl_table_arg);
+			goto out;
+		}
+
+		/* Remember if we need to free the file table */
+		header->ctl_table_arg = ctl_table_arg;
+		**subheader = header;
+		(*subheader)++;
+	}
+
+	/* Recurse into the subdirectories. */
+	for (entry = table; entry->procname; entry++) {
+		char *child_pos;
+
+		if (!entry->child)
+			continue;
+
+		err = -ENAMETOOLONG;
+		child_pos = append_path(path, pos, entry->procname);
+		if (!child_pos)
+			goto out;
+
+		err = register_leaf_sysctl_tables(path, child_pos, subheader,
+						  root, namespaces, entry->child);
+		pos[0] = '\0';
+		if (err)
+			goto out;
+	}
+	err = 0;
+out:
+	/* On failure our caller will unregister all registered subheaders */
+	return err;
+}
+
 /**
  * __register_sysctl_paths - register a sysctl table hierarchy
  * @root: List of sysctl headers to register on
@@ -1077,7 +1161,8 @@ struct ctl_table_header *__register_sysctl_paths(
 	const struct ctl_path *path, struct ctl_table *table)
 {
 	struct ctl_table *ctl_table_arg = table;
-	struct ctl_table_header *header = NULL;
+	int nr_subheaders = count_subheaders(table);
+	struct ctl_table_header *header = NULL, **subheaders, **subheader;
 	const struct ctl_path *component;
 	char *new_path, *pos;
 
@@ -1097,12 +1182,39 @@ struct ctl_table_header *__register_sysctl_paths(
 			goto out;
 		table = table->child;
 	}
-	header = __register_sysctl_table(root, namespaces, new_path, table);
-	if (header)
+	if (nr_subheaders == 1) {
+		header = __register_sysctl_table(root, namespaces, new_path, table);
+		if (header)
+			header->ctl_table_arg = ctl_table_arg;
+	} else {
+		header = kzalloc(sizeof(*header) +
+				 sizeof(*subheaders)*nr_subheaders, GFP_KERNEL);
+		if (!header)
+			goto out;
+
+		subheaders = (struct ctl_table_header **) (header + 1);
+		subheader = subheaders;
 		header->ctl_table_arg = ctl_table_arg;
+
+		if (register_leaf_sysctl_tables(new_path, pos, &subheader,
+						root, namespaces, table))
+			goto err_register_leaves;
+	}
+
 out:
 	kfree(new_path);
 	return header;
+
+err_register_leaves:
+	while (subheader > subheaders) {
+		struct ctl_table_header *subh = *(--subheader);
+		struct ctl_table *table = subh->ctl_table_arg;
+		unregister_sysctl_table(subh);
+		kfree(table);
+	}
+	kfree(header);
+	header = NULL;
+	goto out;
 }
 
 /**
@@ -1149,11 +1261,28 @@ EXPORT_SYMBOL(register_sysctl_table);
  */
 void unregister_sysctl_table(struct ctl_table_header * header)
 {
+	int nr_subheaders;
 	might_sleep();
 
 	if (header == NULL)
 		return;
 
+	nr_subheaders = count_subheaders(header->ctl_table_arg);
+	if (unlikely(nr_subheaders > 1)) {
+		struct ctl_table_header **subheaders;
+		int i;
+
+		subheaders = (struct ctl_table_header **)(header + 1);
+		for (i = nr_subheaders -1; i >= 0; i--) {
+			struct ctl_table_header *subh = subheaders[i];
+			struct ctl_table *table = subh->ctl_table_arg;
+			unregister_sysctl_table(subh);
+			kfree(table);
+		}
+		kfree(header);
+		return;
+	}
+
 	spin_lock(&sysctl_lock);
 	start_unregistering(header);
 	if (!--header->parent->count) {
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index a514e0f..25c7dfe 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1015,7 +1015,7 @@ struct ctl_table
 	void *data;
 	int maxlen;
 	umode_t mode;
-	struct ctl_table *child;
+	struct ctl_table *child;	/* Deprecated */
 	struct ctl_table *parent;	/* Automatically set */
 	proc_handler *proc_handler;	/* Callback for text formatting */
 	struct ctl_table_poll *poll;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 12/29] sysctl: Improve the sysctl sanity checks
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (10 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 11/29] sysctl: register only tables of sysctl files Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field Eric W. Biederman
                   ` (46 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

- Stop validating subdirectories now that we only register leaf tables

- Cleanup and improve the duplicate filename check.
  * Run the duplicate filename check under the sysctl_lock to guarantee
    we never add duplicate names.
  * Reduce the duplicate filename check to nearly O(M*N) where M is the
    number of entries in tthe table we are registering and N is the
    number of entries in the directory before we got there.

- Move the duplicate filename check into it's own function and call
  it directtly from __register_sysctl_table

- Kill the config option as the sanity checks are now cheap enough
  the config option is unnecessary. The original reason for the config
  option was because we had a huge table used to verify the proc filename
  to binary sysctl mapping.  That table has now evolved into the binary_sysctl
  translation layer and is no longer part of the sysctl_check code.

- Tighten up the permission checks.  Guarnateeing that files only have read
  or write permissions.

- Removed redudant check for parents having a procname as now everything has
  a procname.

- Generalize the backtrace logic so that we print a backtrace from
  any failure of __register_sysctl_table that was not caused by
  a memmory allocation failure.  The backtrace allows us to track
  down who erroneously registered a sysctl table.

Bechmark before (CONFIG_SYSCTL_CHECK=y):
    make-dummies 0 999 -> 12s
    rmmod dummy        -> 0.08s

Bechmark before (CONFIG_SYSCTL_CHECK=n):
    make-dummies 0 999 -> 0.7s
    rmmod dummy        -> 0.06s
    make-dummies 0 99999 -> 1m13s
    rmmod dummy          -> 0.38s

Benchmark after:
    make-dummies 0 999 -> 0.65s
    rmmod dummy        -> 0.055s
    make-dummies 0 9999 -> 1m10s
    rmmod dummy         -> 0.39s

The sysctl sanity checks now impose no measurable cost.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |  222 +++++++++++++++++++------------------------------
 lib/Kconfig.debug     |    8 --
 2 files changed, 86 insertions(+), 144 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 6bab2ae..a492ff6 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -726,160 +726,106 @@ static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
 	}
 }
 
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-static int sysctl_depth(struct ctl_table *table)
+static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
+	struct ctl_table *table)
 {
-	struct ctl_table *tmp;
-	int depth;
-
-	depth = 0;
-	for (tmp = table; tmp->parent; tmp = tmp->parent)
-		depth++;
+	struct ctl_table *entry, *test;
+	int error = 0;
 
-	return depth;
+	for (entry = old; entry->procname; entry++) {
+		for (test = table; test->procname; test++) {
+			if (strcmp(entry->procname, test->procname) == 0) {
+				printk(KERN_ERR "sysctl duplicate entry: %s/%s\n",
+					path, test->procname);
+				error = -EEXIST;
+			}
+		}
+	}
+	return error;
 }
 
-static struct ctl_table *sysctl_parent(struct ctl_table *table, int n)
+static int sysctl_check_dups(struct nsproxy *namespaces,
+	struct ctl_table_header *header,
+	const char *path, struct ctl_table *table)
 {
-	int i;
+	struct ctl_table_root *root;
+	struct ctl_table_set *set;
+	struct ctl_table_header *dir_head, *head;
+	struct ctl_table *dir_table;
+	int error = 0;
 
-	for (i = 0; table && i < n; i++)
-		table = table->parent;
+	/* No dups if we are the only member of our directory */
+	if (header->attached_by != table)
+		return 0;
 
-	return table;
-}
+	dir_head = header->parent;
+	dir_table = header->attached_to;
 
+	error = sysctl_check_table_dups(path, dir_table, table);
 
-static void sysctl_print_path(struct ctl_table *table)
-{
-	struct ctl_table *tmp;
-	int depth, i;
-	depth = sysctl_depth(table);
-	if (table->procname) {
-		for (i = depth; i >= 0; i--) {
-			tmp = sysctl_parent(table, i);
-			printk("/%s", tmp->procname?tmp->procname:"");
-		}
-	}
-	printk(" ");
-}
+	root = &sysctl_table_root;
+	do {
+		set = lookup_header_set(root, namespaces);
 
-static struct ctl_table *sysctl_check_lookup(struct nsproxy *namespaces,
-						struct ctl_table *table)
-{
-	struct ctl_table_header *head;
-	struct ctl_table *ref, *test;
-	int depth, cur_depth;
-
-	depth = sysctl_depth(table);
-
-	for (head = __sysctl_head_next(namespaces, NULL); head;
-	     head = __sysctl_head_next(namespaces, head)) {
-		cur_depth = depth;
-		ref = head->ctl_table;
-repeat:
-		test = sysctl_parent(table, cur_depth);
-		for (; ref->procname; ref++) {
-			int match = 0;
-			if (cur_depth && !ref->child)
+		list_for_each_entry(head, &set->list, ctl_entry) {
+			if (head->unregistering)
 				continue;
-
-			if (test->procname && ref->procname &&
-			    (strcmp(test->procname, ref->procname) == 0))
-					match++;
-
-			if (match) {
-				if (cur_depth != 0) {
-					cur_depth--;
-					ref = ref->child;
-					goto repeat;
-				}
-				goto out;
-			}
+			if (head->attached_to != dir_table)
+				continue;
+			error = sysctl_check_table_dups(path, head->attached_by,
+							table);
 		}
-	}
-	ref = NULL;
-out:
-	sysctl_head_finish(head);
-	return ref;
+		root = list_entry(root->root_list.next,
+				  struct ctl_table_root, root_list);
+	} while (root != &sysctl_table_root);
+	return error;
 }
 
-static void set_fail(const char **fail, struct ctl_table *table, const char *str)
+static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
 {
-	if (*fail) {
-		printk(KERN_ERR "sysctl table check failed: ");
-		sysctl_print_path(table);
-		printk(" %s\n", *fail);
-		dump_stack();
-	}
-	*fail = str;
-}
+	struct va_format vaf;
+	va_list args;
 
-static void sysctl_check_leaf(struct nsproxy *namespaces,
-				struct ctl_table *table, const char **fail)
-{
-	struct ctl_table *ref;
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n",
+		path, table->procname, &vaf);
 
-	ref = sysctl_check_lookup(namespaces, table);
-	if (ref && (ref != table))
-		set_fail(fail, table, "Sysctl already exists");
+	va_end(args);
+	return -EINVAL;
 }
 
-static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
+static int sysctl_check_table(const char *path, struct ctl_table *table)
 {
-	int error = 0;
+	int err = 0;
 	for (; table->procname; table++) {
-		const char *fail = NULL;
-
-		if (table->parent) {
-			if (!table->parent->procname)
-				set_fail(&fail, table, "Parent without procname");
-		}
-		if (table->child) {
-			if (table->data)
-				set_fail(&fail, table, "Directory with data?");
-			if (table->maxlen)
-				set_fail(&fail, table, "Directory with maxlen?");
-			if ((table->mode & (S_IRUGO|S_IXUGO)) != table->mode)
-				set_fail(&fail, table, "Writable sysctl directory");
-			if (table->proc_handler)
-				set_fail(&fail, table, "Directory with proc_handler");
-			if (table->extra1)
-				set_fail(&fail, table, "Directory with extra1");
-			if (table->extra2)
-				set_fail(&fail, table, "Directory with extra2");
-		} else {
-			if ((table->proc_handler == proc_dostring) ||
-			    (table->proc_handler == proc_dointvec) ||
-			    (table->proc_handler == proc_dointvec_minmax) ||
-			    (table->proc_handler == proc_dointvec_jiffies) ||
-			    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
-			    (table->proc_handler == proc_dointvec_ms_jiffies) ||
-			    (table->proc_handler == proc_doulongvec_minmax) ||
-			    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
-				if (!table->data)
-					set_fail(&fail, table, "No data");
-				if (!table->maxlen)
-					set_fail(&fail, table, "No maxlen");
-			}
-#ifdef CONFIG_PROC_SYSCTL
-			if (!table->proc_handler)
-				set_fail(&fail, table, "No proc_handler");
-#endif
-			sysctl_check_leaf(namespaces, table, &fail);
-		}
-		if (table->mode > 0777)
-			set_fail(&fail, table, "bogus .mode");
-		if (fail) {
-			set_fail(&fail, table, NULL);
-			error = -EINVAL;
-		}
 		if (table->child)
-			error |= sysctl_check_table(namespaces, table->child);
+			err = sysctl_err(path, table, "Not a file");
+
+		if ((table->proc_handler == proc_dostring) ||
+		    (table->proc_handler == proc_dointvec) ||
+		    (table->proc_handler == proc_dointvec_minmax) ||
+		    (table->proc_handler == proc_dointvec_jiffies) ||
+		    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
+		    (table->proc_handler == proc_dointvec_ms_jiffies) ||
+		    (table->proc_handler == proc_doulongvec_minmax) ||
+		    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
+			if (!table->data)
+				err = sysctl_err(path, table, "No data");
+			if (!table->maxlen)
+				err = sysctl_err(path, table, "No maxlen");
+		}
+		if (!table->proc_handler)
+			err = sysctl_err(path, table, "No proc_handler");
+
+		if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode)
+			err = sysctl_err(path, table, "bogus .mode 0%o",
+				table->mode);
 	}
-	return error;
+	return err;
 }
-#endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
 
 /**
  * __register_sysctl_table - register a leaf sysctl table
@@ -1003,12 +949,8 @@ struct ctl_table_header *__register_sysctl_table(
 	header->root = root;
 	sysctl_set_parent(NULL, header->ctl_table);
 	header->count = 1;
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-	if (sysctl_check_table(namespaces, header->ctl_table)) {
-		kfree(header);
-		return NULL;
-	}
-#endif
+	if (sysctl_check_table(path, table))
+		goto fail;
 	spin_lock(&sysctl_lock);
 	header->set = lookup_header_set(root, namespaces);
 	header->attached_by = header->ctl_table;
@@ -1029,11 +971,19 @@ struct ctl_table_header *__register_sysctl_table(
 				  struct ctl_table_root, root_list);
 		set = lookup_header_set(root, namespaces);
 	}
+	if (sysctl_check_dups(namespaces, header, path, table))
+		goto fail_locked;
 	header->parent->count++;
 	list_add_tail(&header->ctl_entry, &header->set->list);
 	spin_unlock(&sysctl_lock);
 
 	return header;
+fail_locked:
+	spin_unlock(&sysctl_lock);
+fail:
+	kfree(header);
+	dump_stack();
+	return NULL;
 }
 
 static char *append_path(const char *path, char *pos, const char *name)
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8745ac7..943a618 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1113,14 +1113,6 @@ config LATENCYTOP
 	  Enable this option if you want to use the LatencyTOP tool
 	  to find out which userspace is blocking on what kernel operations.
 
-config SYSCTL_SYSCALL_CHECK
-	bool "Sysctl checks"
-	depends on SYSCTL
-	---help---
-	  sys_sysctl uses binary paths that have been found challenging
-	  to properly maintain and use. This enables checks that help
-	  you to keep things correct.
-
 source mm/Kconfig.debug
 source kernel/trace/Kconfig
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (11 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 12/29] sysctl: Improve the sysctl sanity checks Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 14/29] sysctl: A more obvious version of grab_header Eric W. Biederman
                   ` (45 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

While useful at one time for selinux and the sysctl sanity
checks those users no longer use the parent field and we can
safely remove it.

Inspired-by: Lucian Adrian Grijincu <lucian.grijincu@gmil.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   12 +-----------
 include/linux/sysctl.h |    1 -
 2 files changed, 1 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a492ff6..e573f9b 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -218,16 +218,6 @@ static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int
 	return test_perm(mode, op);
 }
 
-static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
-{
-	for (; table->procname; table++) {
-		table->parent = parent;
-		if (table->child)
-			sysctl_set_parent(table, table->child);
-	}
-}
-
-
 static struct inode *proc_sys_make_inode(struct super_block *sb,
 		struct ctl_table_header *head, struct ctl_table *table)
 {
@@ -947,10 +937,10 @@ struct ctl_table_header *__register_sysctl_table(
 	header->used = 0;
 	header->unregistering = NULL;
 	header->root = root;
-	sysctl_set_parent(NULL, header->ctl_table);
 	header->count = 1;
 	if (sysctl_check_table(path, table))
 		goto fail;
+
 	spin_lock(&sysctl_lock);
 	header->set = lookup_header_set(root, namespaces);
 	header->attached_by = header->ctl_table;
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 25c7dfe..094bc5c 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1016,7 +1016,6 @@ struct ctl_table
 	int maxlen;
 	umode_t mode;
 	struct ctl_table *child;	/* Deprecated */
-	struct ctl_table *parent;	/* Automatically set */
 	proc_handler *proc_handler;	/* Callback for text formatting */
 	struct ctl_table_poll *poll;
 	void *extra1;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 14/29] sysctl: A more obvious version of grab_header.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (12 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables Eric W. Biederman
                   ` (44 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Instead of relying on sysct_head_next(NULL) to magically
return the right header for the root directory instead
explicitly transform NULL into the root directories header.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e573f9b..1544485 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -267,10 +267,10 @@ static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
 
 static struct ctl_table_header *grab_header(struct inode *inode)
 {
-	if (PROC_I(inode)->sysctl)
-		return sysctl_head_grab(PROC_I(inode)->sysctl);
-	else
-		return sysctl_head_next(NULL);
+	struct ctl_table_header *head = PROC_I(inode)->sysctl;
+	if (!head)
+		head = &root_table_header;
+	return sysctl_head_grab(head);
 }
 
 static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (13 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 14/29] sysctl: A more obvious version of grab_header Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths Eric W. Biederman
                   ` (43 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Add nreg to ctl_table_header.  When nreg drops to 0 the ctl_table_header
will be unregistered.

Factor out drop_sysctl_table from unregister_sysctl_table, and add
the logic for decrementing nreg.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   28 +++++++++++++++++++---------
 include/linux/sysctl.h |    1 +
 2 files changed, 20 insertions(+), 9 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 1544485..13faa48 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -29,8 +29,9 @@ static struct ctl_table root_table[1];
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
 	{{.count = 1,
-	.ctl_table = root_table,
-	.ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+	  .nreg = 1,
+	  .ctl_table = root_table,
+	  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
 	.root = &sysctl_table_root,
 	.set = &sysctl_table_root.default_set,
 };
@@ -938,6 +939,7 @@ struct ctl_table_header *__register_sysctl_table(
 	header->unregistering = NULL;
 	header->root = root;
 	header->count = 1;
+	header->nreg = 1;
 	if (sysctl_check_table(path, table))
 		goto fail;
 
@@ -1192,6 +1194,20 @@ struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
 }
 EXPORT_SYMBOL(register_sysctl_table);
 
+static void drop_sysctl_table(struct ctl_table_header *header)
+{
+	if (--header->nreg)
+		return;
+
+	start_unregistering(header);
+	if (!--header->parent->count) {
+		WARN_ON(1);
+		kfree_rcu(header->parent, rcu);
+	}
+	if (!--header->count)
+		kfree_rcu(header, rcu);
+}
+
 /**
  * unregister_sysctl_table - unregister a sysctl table hierarchy
  * @header: the header returned from register_sysctl_table
@@ -1224,13 +1240,7 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 	}
 
 	spin_lock(&sysctl_lock);
-	start_unregistering(header);
-	if (!--header->parent->count) {
-		WARN_ON(1);
-		kfree_rcu(header->parent, rcu);
-	}
-	if (!--header->count)
-		kfree_rcu(header, rcu);
+	drop_sysctl_table(header);
 	spin_unlock(&sysctl_lock);
 }
 EXPORT_SYMBOL(unregister_sysctl_table);
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 094bc5c..e40b8f6 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1032,6 +1032,7 @@ struct ctl_table_header
 			struct list_head ctl_entry;
 			int used;
 			int count;
+			int nreg;
 		};
 		struct rcu_head rcu;
 	};
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (14 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 17/29] sysctl: Factor out insert_header and erase_header Eric W. Biederman
                   ` (42 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Factor out a routing to initialize the sysctl_table_header.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   25 +++++++++++++++++--------
 1 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 13faa48..4979925 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -42,6 +42,21 @@ static struct ctl_table_root sysctl_table_root = {
 
 static DEFINE_SPINLOCK(sysctl_lock);
 
+static void init_header(struct ctl_table_header *head,
+	struct ctl_table_root *root, struct ctl_table_set *set,
+	struct ctl_table *table)
+{
+	head->ctl_table_arg = table;
+	INIT_LIST_HEAD(&head->ctl_entry);
+	head->used = 0;
+	head->count = 1;
+	head->nreg = 1;
+	head->unregistering = NULL;
+	head->root = root;
+	head->set = set;
+	head->parent = NULL;
+}
+
 /* called under sysctl_lock */
 static int use_table(struct ctl_table_header *p)
 {
@@ -932,14 +947,8 @@ struct ctl_table_header *__register_sysctl_table(
 		new_name += namelen + 1;
 	}
 	*prevp = table;
-	header->ctl_table_arg = table;
-
-	INIT_LIST_HEAD(&header->ctl_entry);
-	header->used = 0;
-	header->unregistering = NULL;
-	header->root = root;
-	header->count = 1;
-	header->nreg = 1;
+
+	init_header(header, root, NULL, table);
 	if (sysctl_check_table(path, table))
 		goto fail;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 17/29] sysctl: Factor out insert_header and erase_header
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (15 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:49 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
                   ` (41 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   16 +++++++++++++---
 1 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 4979925..7e96a26 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -57,6 +57,17 @@ static void init_header(struct ctl_table_header *head,
 	head->parent = NULL;
 }
 
+static void erase_header(struct ctl_table_header *head)
+{
+	list_del_init(&head->ctl_entry);
+}
+
+static void insert_header(struct ctl_table_header *header)
+{
+	header->parent->count++;
+	list_add_tail(&header->ctl_entry, &header->set->list);
+}
+
 /* called under sysctl_lock */
 static int use_table(struct ctl_table_header *p)
 {
@@ -96,7 +107,7 @@ static void start_unregistering(struct ctl_table_header *p)
 	 * do not remove from the list until nobody holds it; walking the
 	 * list in do_sysctl() relies on that.
 	 */
-	list_del_init(&p->ctl_entry);
+	erase_header(p);
 }
 
 static void sysctl_head_get(struct ctl_table_header *head)
@@ -974,8 +985,7 @@ struct ctl_table_header *__register_sysctl_table(
 	}
 	if (sysctl_check_dups(namespaces, header, path, table))
 		goto fail_locked;
-	header->parent->count++;
-	list_add_tail(&header->ctl_entry, &header->set->list);
+	insert_header(header);
 	spin_unlock(&sysctl_lock);
 
 	return header;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 18/29] sysctl: Normalize the root_table data structure.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (16 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 17/29] sysctl: Factor out insert_header and erase_header Eric W. Biederman
@ 2012-01-27  4:49 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
                   ` (40 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:49 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Every other directory has a .child member and we look at the .child
for our entries.  Do the same for the root_table.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   15 +++++++++++----
 1 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 7e96a26..88d1b06 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -25,7 +25,14 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
 	wake_up_interruptible(&poll->wait);
 }
 
-static struct ctl_table root_table[1];
+static struct ctl_table root_table[] = {
+	{
+		.procname = "",
+		.mode = S_IRUGO|S_IXUGO,
+		.child = &root_table[1],
+	},
+	{ }
+};
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
 	{{.count = 1,
@@ -319,7 +326,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 		goto out;
 	}
 
-	table = table ? table->child : head->ctl_table;
+	table = table ? table->child : &head->ctl_table[1];
 
 	p = find_in_table(table, name);
 	if (!p) {
@@ -510,7 +517,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 		goto out;
 	}
 
-	table = table ? table->child : head->ctl_table;
+	table = table ? table->child : &head->ctl_table[1];
 
 	ret = 0;
 	/* Avoid a switch here: arm builds fail with missing __cmpdi2 */
@@ -966,7 +973,7 @@ struct ctl_table_header *__register_sysctl_table(
 	spin_lock(&sysctl_lock);
 	header->set = lookup_header_set(root, namespaces);
 	header->attached_by = header->ctl_table;
-	header->attached_to = root_table;
+	header->attached_to = &root_table[1];
 	header->parent = &root_table_header;
 	set = header->set;
 	root = header->root;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (17 preceding siblings ...)
  2012-01-27  4:49 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry Eric W. Biederman
                   ` (39 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Replace the helpers that proc_sys_lookup uses with helpers that work
in terms of an entire sysctl directory.  This is worse for sysctl_lock
hold times but it is much better for code clarity and the code cleanups
to come.

find_in_table is no longer needed so it is removed.

find_entry a general helper to find entries in a directory is added.

lookup_entry is a simple wrapper around find_entry that takes the
sysctl_lock increases the use count if an entry is found and drops
the sysctl_lock.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |  102 ++++++++++++++++++++++++++++++++++++------------
 1 files changed, 76 insertions(+), 26 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 88d1b06..3b63f29 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -49,6 +49,55 @@ static struct ctl_table_root sysctl_table_root = {
 
 static DEFINE_SPINLOCK(sysctl_lock);
 
+static int namecmp(const char *name1, int len1, const char *name2, int len2)
+{
+	int minlen;
+	int cmp;
+
+	minlen = len1;
+	if (minlen > len2)
+		minlen = len2;
+
+	cmp = memcmp(name1, name2, minlen);
+	if (cmp == 0)
+		cmp = len1 - len2;
+	return cmp;
+}
+
+static struct ctl_table *find_entry(struct ctl_table_header **phead,
+	struct ctl_table_set *set,
+	struct ctl_table_header *dir_head, struct ctl_table *dir,
+	const char *name, int namelen)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *entry;
+
+	if (dir_head->set == set) {
+		for (entry = dir; entry->procname; entry++) {
+			const char *procname = entry->procname;
+			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
+				*phead = dir_head;
+				return entry;
+			}
+		}
+	}
+
+	list_for_each_entry(head, &set->list, ctl_entry) {
+		if (head->unregistering)
+			continue;
+		if (head->attached_to != dir)
+			continue;
+		for (entry = head->attached_by; entry->procname; entry++) {
+			const char *procname = entry->procname;
+			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
+				*phead = head;
+				return entry;
+			}
+		}
+	}
+	return NULL;
+}
+
 static void init_header(struct ctl_table_header *head,
 	struct ctl_table_root *root, struct ctl_table_set *set,
 	struct ctl_table *table)
@@ -168,6 +217,32 @@ lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
 	return &set->list;
 }
 
+static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
+				      struct ctl_table_header *dir_head,
+				      struct ctl_table *dir,
+				      const char *name, int namelen)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *entry;
+	struct ctl_table_root *root;
+	struct ctl_table_set *set;
+
+	spin_lock(&sysctl_lock);
+	root = &sysctl_table_root;
+	do {
+		set = lookup_header_set(root, current->nsproxy);
+		entry = find_entry(&head, set, dir_head, dir, name, namelen);
+		if (entry && use_table(head))
+			*phead = head;
+		else
+			entry = NULL;
+		root = list_entry(root->root_list.next,
+				  struct ctl_table_root, root_list);
+	} while (!entry && root != &sysctl_table_root);
+	spin_unlock(&sysctl_lock);
+	return entry;
+}
+
 static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
 						struct ctl_table_header *prev)
 {
@@ -284,21 +359,6 @@ out:
 	return inode;
 }
 
-static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
-{
-	for ( ; p->procname; p++) {
-		if (strlen(p->procname) != name->len)
-			continue;
-
-		if (memcmp(p->procname, name->name, name->len) != 0)
-			continue;
-
-		/* I have a match */
-		return p;
-	}
-	return NULL;
-}
-
 static struct ctl_table_header *grab_header(struct inode *inode)
 {
 	struct ctl_table_header *head = PROC_I(inode)->sysctl;
@@ -328,17 +388,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 
 	table = table ? table->child : &head->ctl_table[1];
 
-	p = find_in_table(table, name);
-	if (!p) {
-		for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
-			if (h->attached_to != table)
-				continue;
-			p = find_in_table(h->attached_by, name);
-			if (p)
-				break;
-		}
-	}
-
+	p = lookup_entry(&h, head, table, name->name, name->len);
 	if (!p)
 		goto out;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (18 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
                   ` (38 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Replace sysctl_head_next with first_entry and next_entry.  These new
iterators operate at the level of sysctl table entries and filter
out any sysctl tables that should not be shown.

Utilizing two specialized functions instead of a single function removes
conditionals for handling awkward special cases that only come up
at the beginning of iteration, making the iterators easier to read
and understand.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   98 +++++++++++++++++++++++++++++++------------------
 1 files changed, 62 insertions(+), 36 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 3b63f29..d9c3ae6 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -243,31 +243,25 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	return entry;
 }
 
-static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-						struct ctl_table_header *prev)
+static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
+	struct ctl_table_root *root, struct list_head *tmp)
 {
-	struct ctl_table_root *root;
+	struct nsproxy *namespaces = current->nsproxy;
 	struct list_head *header_list;
 	struct ctl_table_header *head;
-	struct list_head *tmp;
 
-	spin_lock(&sysctl_lock);
-	if (prev) {
-		head = prev;
-		tmp = &prev->ctl_entry;
-		unuse_table(prev);
-		goto next;
-	}
-	tmp = &root_table_header.ctl_entry;
+	goto next;
 	for (;;) {
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
+		root = head->root;
 
-		if (!use_table(head))
+		if (head->attached_to != dir ||
+		    !head->attached_by->procname ||
+		    !use_table(head))
 			goto next;
-		spin_unlock(&sysctl_lock);
+
 		return head;
 	next:
-		root = head->root;
 		tmp = tmp->next;
 		header_list = lookup_header_list(root, namespaces);
 		if (tmp != header_list)
@@ -283,13 +277,53 @@ static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
 		tmp = header_list->next;
 	}
 out:
-	spin_unlock(&sysctl_lock);
 	return NULL;
 }
 
-static struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
+static void first_entry(
+	struct ctl_table_header *dir_head, struct ctl_table *dir,
+	struct ctl_table_header **phead, struct ctl_table **pentry)
 {
-	return __sysctl_head_next(current->nsproxy, prev);
+	struct ctl_table_header *head = dir_head;
+	struct ctl_table *entry = dir;
+
+	spin_lock(&sysctl_lock);
+	if (entry->procname) {
+		use_table(head);
+	} else {
+		head = next_usable_entry(dir, &sysctl_table_root,
+					 &sysctl_table_root.default_set.list);
+		if (head)
+			entry = head->attached_by;
+	}
+	spin_unlock(&sysctl_lock);
+	*phead = head;
+	*pentry = entry;
+}
+
+static void next_entry(struct ctl_table *dir,
+	struct ctl_table_header **phead, struct ctl_table **pentry)
+{
+	struct ctl_table_header *head = *phead;
+	struct ctl_table *entry = *pentry;
+
+	entry++;
+	if (!entry->procname) {
+		struct ctl_table_root *root = head->root;
+		struct list_head *tmp = &head->ctl_entry;
+		if (head->attached_to != dir) {
+			root = &sysctl_table_root;
+			tmp = &sysctl_table_root.default_set.list;
+		}
+		spin_lock(&sysctl_lock);
+		unuse_table(head);
+		head = next_usable_entry(dir, root, tmp);
+		spin_unlock(&sysctl_lock);
+		if (head)
+			entry = head->attached_by;
+	}
+	*phead = head;
+	*pentry = entry;
 }
 
 void register_sysctl_root(struct ctl_table_root *root)
@@ -533,20 +567,17 @@ static int scan(struct ctl_table_header *head, ctl_table *table,
 		unsigned long *pos, struct file *file,
 		void *dirent, filldir_t filldir)
 {
+	int res;
 
-	for (; table->procname; table++, (*pos)++) {
-		int res;
+	if ((*pos)++ < file->f_pos)
+		return 0;
 
-		if (*pos < file->f_pos)
-			continue;
+	res = proc_sys_fill_cache(file, dirent, filldir, head, table);
 
-		res = proc_sys_fill_cache(file, dirent, filldir, head, table);
-		if (res)
-			return res;
+	if (res == 0)
+		file->f_pos = *pos;
 
-		file->f_pos = *pos + 1;
-	}
-	return 0;
+	return res;
 }
 
 static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
@@ -556,6 +587,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct ctl_table_header *head = grab_header(inode);
 	struct ctl_table *table = PROC_I(inode)->sysctl_entry;
 	struct ctl_table_header *h = NULL;
+	struct ctl_table *entry;
 	unsigned long pos;
 	int ret = -EINVAL;
 
@@ -585,14 +617,8 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	}
 	pos = 2;
 
-	ret = scan(head, table, &pos, filp, dirent, filldir);
-	if (ret)
-		goto out;
-
-	for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
-		if (h->attached_to != table)
-			continue;
-		ret = scan(h, h->attached_by, &pos, filp, dirent, filldir);
+	for (first_entry(head, table, &h, &entry); h; next_entry(table, &h, &entry)) {
+		ret = scan(h, entry, &pos, filp, dirent, filldir);
 		if (ret) {
 			sysctl_head_finish(h);
 			break;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (19 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
                   ` (37 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Add a ctl_table_root pointer to ctl_table set so it is easy to
go from a ctl_table_set to a ctl_table_root.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |    3 +++
 include/linux/sysctl.h |    3 +++
 net/sysctl_net.c       |    5 ++---
 3 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index d9c3ae6..65c13dd 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -45,6 +45,7 @@ static struct ctl_table_header root_table_header = {
 static struct ctl_table_root sysctl_table_root = {
 	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+	.default_set.root = &sysctl_table_root,
 };
 
 static DEFINE_SPINLOCK(sysctl_lock);
@@ -1348,9 +1349,11 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 EXPORT_SYMBOL(unregister_sysctl_table);
 
 void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *))
 {
 	INIT_LIST_HEAD(&p->list);
+	p->root = root;
 	p->is_seen = is_seen;
 }
 
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index e40b8f6..e73ba33 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1047,6 +1047,7 @@ struct ctl_table_header
 
 struct ctl_table_set {
 	struct list_head list;
+	struct ctl_table_root *root;
 	int (*is_seen)(struct ctl_table_set *);
 };
 
@@ -1069,6 +1070,7 @@ struct ctl_path {
 void proc_sys_poll_notify(struct ctl_table_poll *poll);
 
 extern void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *));
 extern void retire_sysctl_set(struct ctl_table_set *set);
 
@@ -1103,6 +1105,7 @@ static inline void unregister_sysctl_table(struct ctl_table_header * table)
 }
 
 static inline void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *))
 {
 }
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index 07c6b87..e998c64 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -74,8 +74,7 @@ static struct ctl_table_root net_sysctl_ro_root = {
 
 static int __net_init sysctl_net_init(struct net *net)
 {
-	setup_sysctl_set(&net->sysctls,
-			 is_seen);
+	setup_sysctl_set(&net->sysctls, &net_sysctl_root, is_seen);
 	return 0;
 }
 
@@ -95,7 +94,7 @@ static __init int net_sysctl_init(void)
 	ret = register_pernet_subsys(&sysctl_pernet_ops);
 	if (ret)
 		goto out;
-	setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL);
+	setup_sysctl_set(&net_sysctl_ro_root.default_set, &net_sysctl_ro_root, NULL);
 	register_sysctl_root(&net_sysctl_ro_root);
 	register_sysctl_root(&net_sysctl_root);
 out:
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (20 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir Eric W. Biederman
                   ` (36 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Simplify the code and the sysctl semantics by autogenerating
sysctl directories when a sysctl table is registered that needs
the directories and autodeleting the directories when there are
no more sysctl tables registered that need them.

Autogenerating directories keeps sysctl tables from depending
on each other, removing all of the arcane register/unregister
ordering constraints and makes it impossible to get the order
wrong when reigsering and unregistering sysctl tables.

Autogenerating directories yields one unique entity that dentries
can point to, retaining the current effective use of the dcache.

Add struct ctl_dir as the type of these new autogenerated
directories.

The attached_by and attached_to fields in ctl_table_header are
removed as they are no longer needed.

The child field in ctl_table is no longer needed by the core of
the sysctl code.  ctl_table.child can be removed once all of the
existing users have been updated.

Benchmark before:
    make-dummies 0 999 -> 0.7s
    rmmod dummy        -> 0.07s
    make-dummies 0 9999 -> 1m10s
    rmmod dummy         -> 0.4s

Benchmark after:
    make-dummies 0 999 -> 0.44s
    rmmod dummy        -> 0.065s
    make-dummies 0 9999 -> 1m36s
    rmmod dummy         -> 0.4s

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  342 ++++++++++++++++++++----------------------------
 include/linux/sysctl.h |   10 +-
 2 files changed, 150 insertions(+), 202 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 65c13dd..3c0767d 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -28,28 +28,31 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
 static struct ctl_table root_table[] = {
 	{
 		.procname = "",
-		.mode = S_IRUGO|S_IXUGO,
-		.child = &root_table[1],
+		.mode = S_IFDIR|S_IRUGO|S_IXUGO,
 	},
 	{ }
 };
 static struct ctl_table_root sysctl_table_root;
-static struct ctl_table_header root_table_header = {
-	{{.count = 1,
-	  .nreg = 1,
-	  .ctl_table = root_table,
-	  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
-	.root = &sysctl_table_root,
-	.set = &sysctl_table_root.default_set,
+static struct ctl_dir sysctl_root_dir = {
+	.header = {
+		{{.count = 1,
+		  .nreg = 1,
+		  .ctl_table = root_table,
+		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+		.root = &sysctl_table_root,
+		.set = &sysctl_table_root.default_set,
+	},
 };
 static struct ctl_table_root sysctl_table_root = {
 	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+	.default_set.list = LIST_HEAD_INIT(sysctl_root_dir.header.ctl_entry),
 	.default_set.root = &sysctl_table_root,
 };
 
 static DEFINE_SPINLOCK(sysctl_lock);
 
+static void drop_sysctl_table(struct ctl_table_header *header);
+
 static int namecmp(const char *name1, int len1, const char *name2, int len2)
 {
 	int minlen;
@@ -66,29 +69,18 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 }
 
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
-	struct ctl_table_set *set,
-	struct ctl_table_header *dir_head, struct ctl_table *dir,
+	struct ctl_table_set *set, struct ctl_dir *dir,
 	const char *name, int namelen)
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
-	if (dir_head->set == set) {
-		for (entry = dir; entry->procname; entry++) {
-			const char *procname = entry->procname;
-			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
-				*phead = dir_head;
-				return entry;
-			}
-		}
-	}
-
 	list_for_each_entry(head, &set->list, ctl_entry) {
 		if (head->unregistering)
 			continue;
-		if (head->attached_to != dir)
+		if (head->parent != dir)
 			continue;
-		for (entry = head->attached_by; entry->procname; entry++) {
+		for (entry = head->ctl_table; entry->procname; entry++) {
 			const char *procname = entry->procname;
 			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
 				*phead = head;
@@ -103,6 +95,7 @@ static void init_header(struct ctl_table_header *head,
 	struct ctl_table_root *root, struct ctl_table_set *set,
 	struct ctl_table *table)
 {
+	head->ctl_table = table;
 	head->ctl_table_arg = table;
 	INIT_LIST_HEAD(&head->ctl_entry);
 	head->used = 0;
@@ -119,9 +112,10 @@ static void erase_header(struct ctl_table_header *head)
 	list_del_init(&head->ctl_entry);
 }
 
-static void insert_header(struct ctl_table_header *header)
+static void insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
-	header->parent->count++;
+	header->parent = dir;
+	header->parent->header.nreg++;
 	list_add_tail(&header->ctl_entry, &header->set->list);
 }
 
@@ -219,8 +213,7 @@ lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
 }
 
 static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
-				      struct ctl_table_header *dir_head,
-				      struct ctl_table *dir,
+				      struct ctl_dir *dir,
 				      const char *name, int namelen)
 {
 	struct ctl_table_header *head;
@@ -232,7 +225,7 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	root = &sysctl_table_root;
 	do {
 		set = lookup_header_set(root, current->nsproxy);
-		entry = find_entry(&head, set, dir_head, dir, name, namelen);
+		entry = find_entry(&head, set, dir, name, namelen);
 		if (entry && use_table(head))
 			*phead = head;
 		else
@@ -244,7 +237,7 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	return entry;
 }
 
-static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
+static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
 	struct ctl_table_root *root, struct list_head *tmp)
 {
 	struct nsproxy *namespaces = current->nsproxy;
@@ -256,8 +249,8 @@ static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
 		root = head->root;
 
-		if (head->attached_to != dir ||
-		    !head->attached_by->procname ||
+		if (head->parent != dir ||
+		    !head->ctl_table->procname ||
 		    !use_table(head))
 			goto next;
 
@@ -281,47 +274,35 @@ out:
 	return NULL;
 }
 
-static void first_entry(
-	struct ctl_table_header *dir_head, struct ctl_table *dir,
+static void first_entry(struct ctl_dir *dir,
 	struct ctl_table_header **phead, struct ctl_table **pentry)
 {
-	struct ctl_table_header *head = dir_head;
-	struct ctl_table *entry = dir;
+	struct ctl_table_header *head;
+	struct ctl_table *entry = NULL;
 
 	spin_lock(&sysctl_lock);
-	if (entry->procname) {
-		use_table(head);
-	} else {
-		head = next_usable_entry(dir, &sysctl_table_root,
-					 &sysctl_table_root.default_set.list);
-		if (head)
-			entry = head->attached_by;
-	}
+	head = next_usable_entry(dir, &sysctl_table_root,
+				 &sysctl_table_root.default_set.list);
 	spin_unlock(&sysctl_lock);
+	if (head)
+		entry = head->ctl_table;
 	*phead = head;
 	*pentry = entry;
 }
 
-static void next_entry(struct ctl_table *dir,
-	struct ctl_table_header **phead, struct ctl_table **pentry)
+static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry)
 {
 	struct ctl_table_header *head = *phead;
 	struct ctl_table *entry = *pentry;
 
 	entry++;
 	if (!entry->procname) {
-		struct ctl_table_root *root = head->root;
-		struct list_head *tmp = &head->ctl_entry;
-		if (head->attached_to != dir) {
-			root = &sysctl_table_root;
-			tmp = &sysctl_table_root.default_set.list;
-		}
 		spin_lock(&sysctl_lock);
 		unuse_table(head);
-		head = next_usable_entry(dir, root, tmp);
+		head = next_usable_entry(head->parent, head->root, &head->ctl_entry);
 		spin_unlock(&sysctl_lock);
 		if (head)
-			entry = head->attached_by;
+			entry = head->ctl_table;
 	}
 	*phead = head;
 	*pentry = entry;
@@ -381,7 +362,7 @@ static struct inode *proc_sys_make_inode(struct super_block *sb,
 
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 	inode->i_mode = table->mode;
-	if (!table->child) {
+	if (!S_ISDIR(table->mode)) {
 		inode->i_mode |= S_IFREG;
 		inode->i_op = &proc_sys_inode_operations;
 		inode->i_fop = &proc_sys_file_operations;
@@ -398,7 +379,7 @@ static struct ctl_table_header *grab_header(struct inode *inode)
 {
 	struct ctl_table_header *head = PROC_I(inode)->sysctl;
 	if (!head)
-		head = &root_table_header;
+		head = &sysctl_root_dir.header;
 	return sysctl_head_grab(head);
 }
 
@@ -406,24 +387,19 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 					struct nameidata *nd)
 {
 	struct ctl_table_header *head = grab_header(dir);
-	struct ctl_table *table = PROC_I(dir)->sysctl_entry;
 	struct ctl_table_header *h = NULL;
 	struct qstr *name = &dentry->d_name;
 	struct ctl_table *p;
 	struct inode *inode;
 	struct dentry *err = ERR_PTR(-ENOENT);
+	struct ctl_dir *ctl_dir;
 
 	if (IS_ERR(head))
 		return ERR_CAST(head);
 
-	if (table && !table->child) {
-		WARN_ON(1);
-		goto out;
-	}
+	ctl_dir = container_of(head, struct ctl_dir, header);
 
-	table = table ? table->child : &head->ctl_table[1];
-
-	p = lookup_entry(&h, head, table, name->name, name->len);
+	p = lookup_entry(&h, ctl_dir, name->name, name->len);
 	if (!p)
 		goto out;
 
@@ -586,21 +562,16 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct dentry *dentry = filp->f_path.dentry;
 	struct inode *inode = dentry->d_inode;
 	struct ctl_table_header *head = grab_header(inode);
-	struct ctl_table *table = PROC_I(inode)->sysctl_entry;
 	struct ctl_table_header *h = NULL;
 	struct ctl_table *entry;
+	struct ctl_dir *ctl_dir;
 	unsigned long pos;
 	int ret = -EINVAL;
 
 	if (IS_ERR(head))
 		return PTR_ERR(head);
 
-	if (table && !table->child) {
-		WARN_ON(1);
-		goto out;
-	}
-
-	table = table ? table->child : &head->ctl_table[1];
+	ctl_dir = container_of(head, struct ctl_dir, header);
 
 	ret = 0;
 	/* Avoid a switch here: arm builds fail with missing __cmpdi2 */
@@ -618,7 +589,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	}
 	pos = 2;
 
-	for (first_entry(head, table, &h, &entry); h; next_entry(table, &h, &entry)) {
+	for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) {
 		ret = scan(h, entry, &pos, filp, dirent, filldir);
 		if (ret) {
 			sysctl_head_finish(h);
@@ -779,52 +750,86 @@ static const struct dentry_operations proc_sys_dentry_operations = {
 	.d_compare	= proc_sys_compare,
 };
 
-static struct ctl_table *is_branch_in(struct ctl_table *branch,
-				      struct ctl_table *table)
+static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *dir,
+	const char *name, int namelen)
 {
-	struct ctl_table *p;
-	const char *s = branch->procname;
+	struct ctl_table_header *head;
+	struct ctl_table *entry;
 
-	/* branch should have named subdirectory as its first element */
-	if (!s || !branch->child)
-		return NULL;
+	entry = find_entry(&head, set, dir, name, namelen);
+	if (!entry)
+		return ERR_PTR(-ENOENT);
+	if (S_ISDIR(entry->mode))
+		return container_of(head, struct ctl_dir, header);
+	return ERR_PTR(-ENOTDIR);
+}
+
+static struct ctl_dir *new_dir(struct ctl_table_set *set,
+	const char *name, int namelen)
+{
+	struct ctl_table *table;
+	struct ctl_dir *new;
+	char *new_name;
 
-	/* ... and nothing else */
-	if (branch[1].procname)
+	new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 +
+		      namelen + 1, GFP_KERNEL);
+	if (!new)
 		return NULL;
 
-	/* table should contain subdirectory with the same name */
-	for (p = table; p->procname; p++) {
-		if (!p->child)
-			continue;
-		if (p->procname && strcmp(p->procname, s) == 0)
-			return p;
-	}
-	return NULL;
+	table = (struct ctl_table *)(new + 1);
+	new_name = (char *)(table + 2);
+	memcpy(new_name, name, namelen);
+	new_name[namelen] = '\0';
+	table[0].procname = new_name;
+	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
+	init_header(&new->header, set->root, set, table);
+
+	return new;
 }
 
-/* see if attaching q to p would be an improvement */
-static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
+static struct ctl_dir *get_subdir(struct ctl_table_set *set,
+	struct ctl_dir *dir, const char *name, int namelen)
 {
-	struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
-	struct ctl_table *next;
-	int is_better = 0;
-	int not_in_parent = !p->attached_by;
-
-	while ((next = is_branch_in(by, to)) != NULL) {
-		if (by == q->attached_by)
-			is_better = 1;
-		if (to == p->attached_by)
-			not_in_parent = 1;
-		by = by->child;
-		to = next->child;
-	}
+	struct ctl_dir *subdir, *new = NULL;
 
-	if (is_better && not_in_parent) {
-		q->attached_by = by;
-		q->attached_to = to;
-		q->parent = p;
+	spin_lock(&sysctl_lock);
+	subdir = find_subdir(dir->header.set, dir, name, namelen);
+	if (!IS_ERR(subdir))
+		goto found;
+	if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)
+		subdir = find_subdir(set, dir, name, namelen);
+	if (!IS_ERR(subdir))
+		goto found;
+	if (PTR_ERR(subdir) != -ENOENT)
+		goto failed;
+
+	spin_unlock(&sysctl_lock);
+	new = new_dir(set, name, namelen);
+	spin_lock(&sysctl_lock);
+	subdir = ERR_PTR(-ENOMEM);
+	if (!new)
+		goto failed;
+
+	subdir = find_subdir(set, dir, name, namelen);
+	if (!IS_ERR(subdir))
+		goto found;
+	if (PTR_ERR(subdir) != -ENOENT)
+		goto failed;
+
+	insert_header(dir, &new->header);
+	subdir = new;
+found:
+	subdir->header.nreg++;
+failed:
+	if (unlikely(IS_ERR(subdir))) {
+		printk(KERN_ERR "sysctl could not get directory: %*.*s %ld\n",
+			namelen, namelen, name, PTR_ERR(subdir));
 	}
+	drop_sysctl_table(&dir->header);
+	if (new)
+		drop_sysctl_table(&new->header);
+	spin_unlock(&sysctl_lock);
+	return subdir;
 }
 
 static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
@@ -846,24 +851,14 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 }
 
 static int sysctl_check_dups(struct nsproxy *namespaces,
-	struct ctl_table_header *header,
+	struct ctl_dir *dir,
 	const char *path, struct ctl_table *table)
 {
 	struct ctl_table_root *root;
 	struct ctl_table_set *set;
-	struct ctl_table_header *dir_head, *head;
-	struct ctl_table *dir_table;
+	struct ctl_table_header *head;
 	int error = 0;
 
-	/* No dups if we are the only member of our directory */
-	if (header->attached_by != table)
-		return 0;
-
-	dir_head = header->parent;
-	dir_table = header->attached_to;
-
-	error = sysctl_check_table_dups(path, dir_table, table);
-
 	root = &sysctl_table_root;
 	do {
 		set = lookup_header_set(root, namespaces);
@@ -871,9 +866,9 @@ static int sysctl_check_dups(struct nsproxy *namespaces,
 		list_for_each_entry(head, &set->list, ctl_entry) {
 			if (head->unregistering)
 				continue;
-			if (head->attached_to != dir_table)
+			if (head->parent != dir)
 				continue;
-			error = sysctl_check_table_dups(path, head->attached_by,
+			error = sysctl_check_table_dups(path, head->ctl_table,
 							table);
 		}
 		root = list_entry(root->root_list.next,
@@ -977,47 +972,25 @@ struct ctl_table_header *__register_sysctl_table(
 	const char *path, struct ctl_table *table)
 {
 	struct ctl_table_header *header;
-	struct ctl_table *new, **prevp;
 	const char *name, *nextname;
-	unsigned int npath = 0;
 	struct ctl_table_set *set;
-	size_t path_bytes = 0;
-	char *new_name;
-
-	/* Count the path components */
-	for (name = path; name; name = nextname) {
-		int namelen;
-		nextname = strchr(name, '/');
-		if (nextname) {
-			namelen = nextname - name;
-			nextname++;
-		} else {
-			namelen = strlen(name);
-		}
-		if (namelen == 0)
-			continue;
-		path_bytes += namelen + 1;
-		npath++;
-	}
+	struct ctl_dir *dir;
 
-	/*
-	 * For each path component, allocate a 2-element ctl_table array.
-	 * The first array element will be filled with the sysctl entry
-	 * for this, the second will be the sentinel (procname == 0).
-	 *
-	 * We allocate everything in one go so that we don't have to
-	 * worry about freeing additional memory in unregister_sysctl_table.
-	 */
-	header = kzalloc(sizeof(struct ctl_table_header) + path_bytes +
-			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
+	header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
 	if (!header)
 		return NULL;
 
-	new = (struct ctl_table *) (header + 1);
-	new_name = (char *)(new + (2 * npath));
+	init_header(header, root, NULL, table);
+	if (sysctl_check_table(path, table))
+		goto fail;
+
+	spin_lock(&sysctl_lock);
+	header->set = set = lookup_header_set(root, namespaces);
+	dir = &sysctl_root_dir;
+	dir->header.nreg++;
+	spin_unlock(&sysctl_lock);
 
-	/* Now connect the dots */
-	prevp = &header->ctl_table;
+	/* Find the directory for the ctl_table */
 	for (name = path; name; name = nextname) {
 		int namelen;
 		nextname = strchr(name, '/');
@@ -1029,51 +1002,21 @@ struct ctl_table_header *__register_sysctl_table(
 		}
 		if (namelen == 0)
 			continue;
-		memcpy(new_name, name, namelen);
-		new_name[namelen] = '\0';
-
-		new->procname = new_name;
-		new->mode     = 0555;
-
-		*prevp = new;
-		prevp = &new->child;
 
-		new += 2;
-		new_name += namelen + 1;
+		dir = get_subdir(set, dir, name, namelen);
+		if (IS_ERR(dir))
+			goto fail;
 	}
-	*prevp = table;
-
-	init_header(header, root, NULL, table);
-	if (sysctl_check_table(path, table))
-		goto fail;
-
 	spin_lock(&sysctl_lock);
-	header->set = lookup_header_set(root, namespaces);
-	header->attached_by = header->ctl_table;
-	header->attached_to = &root_table[1];
-	header->parent = &root_table_header;
-	set = header->set;
-	root = header->root;
-	for (;;) {
-		struct ctl_table_header *p;
-		list_for_each_entry(p, &set->list, ctl_entry) {
-			if (p->unregistering)
-				continue;
-			try_attach(p, header);
-		}
-		if (root == &sysctl_table_root)
-			break;
-		root = list_entry(root->root_list.prev,
-				  struct ctl_table_root, root_list);
-		set = lookup_header_set(root, namespaces);
-	}
-	if (sysctl_check_dups(namespaces, header, path, table))
-		goto fail_locked;
-	insert_header(header);
+	if (sysctl_check_dups(namespaces, dir, path, table))
+		goto fail_put_dir_locked;
+	insert_header(dir, header);
+	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 
 	return header;
-fail_locked:
+fail_put_dir_locked:
+	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 fail:
 	kfree(header);
@@ -1299,16 +1242,17 @@ EXPORT_SYMBOL(register_sysctl_table);
 
 static void drop_sysctl_table(struct ctl_table_header *header)
 {
+	struct ctl_dir *parent = header->parent;
+
 	if (--header->nreg)
 		return;
 
 	start_unregistering(header);
-	if (!--header->parent->count) {
-		WARN_ON(1);
-		kfree_rcu(header->parent, rcu);
-	}
 	if (!--header->count)
 		kfree_rcu(header, rcu);
+
+	if (parent)
+		drop_sysctl_table(&parent->header);
 }
 
 /**
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index e73ba33..3084b62 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -938,6 +938,7 @@ struct ctl_table;
 struct nsproxy;
 struct ctl_table_root;
 struct ctl_table_header;
+struct ctl_dir;
 
 typedef struct ctl_table ctl_table;
 
@@ -1040,9 +1041,12 @@ struct ctl_table_header
 	struct ctl_table *ctl_table_arg;
 	struct ctl_table_root *root;
 	struct ctl_table_set *set;
-	struct ctl_table *attached_by;
-	struct ctl_table *attached_to;
-	struct ctl_table_header *parent;
+	struct ctl_dir *parent;
+};
+
+struct ctl_dir {
+	/* Header must be at the start of ctl_dir */
+	struct ctl_table_header header;
 };
 
 struct ctl_table_set {
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (21 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
                   ` (35 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

When there are errors it is very nice to know the full sysctl path.
Add a simple function that computes the sysctl path and prints it
out.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 3c0767d..a785565 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -53,6 +53,13 @@ static DEFINE_SPINLOCK(sysctl_lock);
 
 static void drop_sysctl_table(struct ctl_table_header *header);
 
+static void sysctl_print_dir(struct ctl_dir *dir)
+{
+	if (dir->header.parent)
+		sysctl_print_dir(dir->header.parent);
+	printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname);
+}
+
 static int namecmp(const char *name1, int len1, const char *name2, int len2)
 {
 	int minlen;
@@ -822,7 +829,9 @@ found:
 	subdir->header.nreg++;
 failed:
 	if (unlikely(IS_ERR(subdir))) {
-		printk(KERN_ERR "sysctl could not get directory: %*.*s %ld\n",
+		printk(KERN_ERR "sysctl could not get directory: ");
+		sysctl_print_dir(dir);
+		printk(KERN_CONT "/%*.*s %ld\n",
 			namelen, namelen, name, PTR_ERR(subdir));
 	}
 	drop_sysctl_table(&dir->header);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (22 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy Eric W. Biederman
                   ` (34 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Piecing together directories by looking first in one directory
tree, than in another directory tree and finally in a third
directory tree makes it hard to verify that some directory
entries are not multiply defined and makes it hard to create
efficient implementations the sysctl filesystem.

Replace the sysctl wide list of roots with autogenerated
links from the core sysctl directory tree to the other
sysctl directory trees.

This simplifies sysctl directory reading and lookups as now
only entries in a single sysctl directory tree need to be
considered.

Benchmark before:
    make-dummies 0 999 -> 0.44s
    rmmod dummy        -> 0.065s
    make-dummies 0 9999 -> 1m36s
    rmmod dummy         -> 0.4s

Benchmark after:
    make-dummies 0 999 -> 0.63s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 2m35s
    rmmod dummy         -> 18s

The slowdown is caused by the lookups used in insert_headers
and put_links to see if we need to add links or remove links.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  397 +++++++++++++++++++++++++++++++++++------------
 include/linux/sysctl.h |    3 +-
 2 files changed, 296 insertions(+), 104 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a785565..ec54a57 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -32,26 +32,26 @@ static struct ctl_table root_table[] = {
 	},
 	{ }
 };
-static struct ctl_table_root sysctl_table_root;
-static struct ctl_dir sysctl_root_dir = {
-	.header = {
+static struct ctl_table_root sysctl_table_root = {
+	.default_set.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),
+	.default_set.dir.header = {
 		{{.count = 1,
 		  .nreg = 1,
 		  .ctl_table = root_table,
 		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+		.ctl_table_arg = root_table,
 		.root = &sysctl_table_root,
 		.set = &sysctl_table_root.default_set,
 	},
 };
-static struct ctl_table_root sysctl_table_root = {
-	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-	.default_set.list = LIST_HEAD_INIT(sysctl_root_dir.header.ctl_entry),
-	.default_set.root = &sysctl_table_root,
-};
 
 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, struct nsproxy *namespaces);
+static int insert_links(struct ctl_table_header *head);
+static void put_links(struct ctl_table_header *header);
 
 static void sysctl_print_dir(struct ctl_dir *dir)
 {
@@ -76,9 +76,9 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 }
 
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
-	struct ctl_table_set *set, struct ctl_dir *dir,
-	const char *name, int namelen)
+	struct ctl_dir *dir, const char *name, int namelen)
 {
+	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
@@ -119,11 +119,21 @@ static void erase_header(struct ctl_table_header *head)
 	list_del_init(&head->ctl_entry);
 }
 
-static void insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
+static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
+	int err;
+
+	dir->header.nreg++;
 	header->parent = dir;
-	header->parent->header.nreg++;
+	err = insert_links(header);
+	if (err)
+		goto fail_links;
 	list_add_tail(&header->ctl_entry, &header->set->list);
+	return 0;
+fail_links:
+	header->parent = NULL;
+	drop_sysctl_table(&dir->header);
+	return err;
 }
 
 /* called under sysctl_lock */
@@ -212,72 +222,39 @@ lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
 	return set;
 }
 
-static struct list_head *
-lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
-{
-	struct ctl_table_set *set = lookup_header_set(root, namespaces);
-	return &set->list;
-}
-
 static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 				      struct ctl_dir *dir,
 				      const char *name, int namelen)
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
-	struct ctl_table_root *root;
-	struct ctl_table_set *set;
 
 	spin_lock(&sysctl_lock);
-	root = &sysctl_table_root;
-	do {
-		set = lookup_header_set(root, current->nsproxy);
-		entry = find_entry(&head, set, dir, name, namelen);
-		if (entry && use_table(head))
-			*phead = head;
-		else
-			entry = NULL;
-		root = list_entry(root->root_list.next,
-				  struct ctl_table_root, root_list);
-	} while (!entry && root != &sysctl_table_root);
+	entry = find_entry(&head, dir, name, namelen);
+	if (entry && use_table(head))
+		*phead = head;
+	else
+		entry = NULL;
 	spin_unlock(&sysctl_lock);
 	return entry;
 }
 
 static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
-	struct ctl_table_root *root, struct list_head *tmp)
+						  struct list_head *tmp)
 {
-	struct nsproxy *namespaces = current->nsproxy;
-	struct list_head *header_list;
+	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 
-	goto next;
-	for (;;) {
+	for (tmp = tmp->next; tmp != &set->list; tmp = tmp->next) {
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
-		root = head->root;
 
 		if (head->parent != dir ||
 		    !head->ctl_table->procname ||
 		    !use_table(head))
-			goto next;
-
-		return head;
-	next:
-		tmp = tmp->next;
-		header_list = lookup_header_list(root, namespaces);
-		if (tmp != header_list)
 			continue;
 
-		do {
-			root = list_entry(root->root_list.next,
-					struct ctl_table_root, root_list);
-			if (root == &sysctl_table_root)
-				goto out;
-			header_list = lookup_header_list(root, namespaces);
-		} while (list_empty(header_list));
-		tmp = header_list->next;
+		return head;
 	}
-out:
 	return NULL;
 }
 
@@ -288,8 +265,7 @@ static void first_entry(struct ctl_dir *dir,
 	struct ctl_table *entry = NULL;
 
 	spin_lock(&sysctl_lock);
-	head = next_usable_entry(dir, &sysctl_table_root,
-				 &sysctl_table_root.default_set.list);
+	head = next_usable_entry(dir, &dir->header.set->list);
 	spin_unlock(&sysctl_lock);
 	if (head)
 		entry = head->ctl_table;
@@ -306,7 +282,7 @@ static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentr
 	if (!entry->procname) {
 		spin_lock(&sysctl_lock);
 		unuse_table(head);
-		head = next_usable_entry(head->parent, head->root, &head->ctl_entry);
+		head = next_usable_entry(head->parent, &head->ctl_entry);
 		spin_unlock(&sysctl_lock);
 		if (head)
 			entry = head->ctl_table;
@@ -317,9 +293,6 @@ static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentr
 
 void register_sysctl_root(struct ctl_table_root *root)
 {
-	spin_lock(&sysctl_lock);
-	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
-	spin_unlock(&sysctl_lock);
 }
 
 /*
@@ -386,7 +359,7 @@ static struct ctl_table_header *grab_header(struct inode *inode)
 {
 	struct ctl_table_header *head = PROC_I(inode)->sysctl;
 	if (!head)
-		head = &sysctl_root_dir.header;
+		head = &sysctl_table_root.default_set.dir.header;
 	return sysctl_head_grab(head);
 }
 
@@ -400,6 +373,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 	struct inode *inode;
 	struct dentry *err = ERR_PTR(-ENOENT);
 	struct ctl_dir *ctl_dir;
+	int ret;
 
 	if (IS_ERR(head))
 		return ERR_CAST(head);
@@ -410,6 +384,11 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 	if (!p)
 		goto out;
 
+	ret = sysctl_follow_link(&h, &p, current->nsproxy);
+	err = ERR_PTR(ret);
+	if (ret)
+		goto out;
+
 	err = ERR_PTR(-ENOMEM);
 	inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
 	if (h)
@@ -547,6 +526,25 @@ static int proc_sys_fill_cache(struct file *filp, void *dirent,
 	return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
 }
 
+static int proc_sys_link_fill_cache(struct file *filp, void *dirent,
+				    filldir_t filldir,
+				    struct ctl_table_header *head,
+				    struct ctl_table *table)
+{
+	int err, ret = 0;
+	head = sysctl_head_grab(head);
+
+	/* It is not an error if we can not follow the link ignore it */
+	err = sysctl_follow_link(&head, &table, current->nsproxy);
+	if (err)
+		goto out;
+
+	ret = proc_sys_fill_cache(filp, dirent, filldir, head, table);
+out:
+	sysctl_head_finish(head);
+	return ret;
+}
+
 static int scan(struct ctl_table_header *head, ctl_table *table,
 		unsigned long *pos, struct file *file,
 		void *dirent, filldir_t filldir)
@@ -556,7 +554,10 @@ static int scan(struct ctl_table_header *head, ctl_table *table,
 	if ((*pos)++ < file->f_pos)
 		return 0;
 
-	res = proc_sys_fill_cache(file, dirent, filldir, head, table);
+	if (unlikely(S_ISLNK(table->mode)))
+		res = proc_sys_link_fill_cache(file, dirent, filldir, head, table);
+	else
+		res = proc_sys_fill_cache(file, dirent, filldir, head, table);
 
 	if (res == 0)
 		file->f_pos = *pos;
@@ -757,13 +758,13 @@ static const struct dentry_operations proc_sys_dentry_operations = {
 	.d_compare	= proc_sys_compare,
 };
 
-static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *dir,
-	const char *name, int namelen)
+static struct ctl_dir *find_subdir(struct ctl_dir *dir,
+				   const char *name, int namelen)
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
-	entry = find_entry(&head, set, dir, name, namelen);
+	entry = find_entry(&head, dir, name, namelen);
 	if (!entry)
 		return ERR_PTR(-ENOENT);
 	if (S_ISDIR(entry->mode))
@@ -772,7 +773,7 @@ static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *di
 }
 
 static struct ctl_dir *new_dir(struct ctl_table_set *set,
-	const char *name, int namelen)
+			       const char *name, int namelen)
 {
 	struct ctl_table *table;
 	struct ctl_dir *new;
@@ -789,22 +790,19 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 	new_name[namelen] = '\0';
 	table[0].procname = new_name;
 	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
-	init_header(&new->header, set->root, set, table);
+	init_header(&new->header, set->dir.header.root, set, table);
 
 	return new;
 }
 
-static struct ctl_dir *get_subdir(struct ctl_table_set *set,
-	struct ctl_dir *dir, const char *name, int namelen)
+static struct ctl_dir *get_subdir(struct ctl_dir *dir,
+				  const char *name, int namelen)
 {
+	struct ctl_table_set *set = dir->header.set;
 	struct ctl_dir *subdir, *new = NULL;
 
 	spin_lock(&sysctl_lock);
-	subdir = find_subdir(dir->header.set, dir, name, namelen);
-	if (!IS_ERR(subdir))
-		goto found;
-	if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)
-		subdir = find_subdir(set, dir, name, namelen);
+	subdir = find_subdir(dir, name, namelen);
 	if (!IS_ERR(subdir))
 		goto found;
 	if (PTR_ERR(subdir) != -ENOENT)
@@ -817,13 +815,14 @@ static struct ctl_dir *get_subdir(struct ctl_table_set *set,
 	if (!new)
 		goto failed;
 
-	subdir = find_subdir(set, dir, name, namelen);
+	subdir = find_subdir(dir, name, namelen);
 	if (!IS_ERR(subdir))
 		goto found;
 	if (PTR_ERR(subdir) != -ENOENT)
 		goto failed;
 
-	insert_header(dir, &new->header);
+	if (insert_header(dir, &new->header))
+		goto failed;
 	subdir = new;
 found:
 	subdir->header.nreg++;
@@ -841,6 +840,57 @@ failed:
 	return subdir;
 }
 
+static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir)
+{
+	struct ctl_dir *parent;
+	const char *procname;
+	if (!dir->header.parent)
+		return &set->dir;
+	parent = xlate_dir(set, dir->header.parent);
+	if (IS_ERR(parent))
+		return parent;
+	procname = dir->header.ctl_table[0].procname;
+	return find_subdir(parent, procname, strlen(procname));
+}
+
+static int sysctl_follow_link(struct ctl_table_header **phead,
+	struct ctl_table **pentry, struct nsproxy *namespaces)
+{
+	struct ctl_table_header *head;
+	struct ctl_table_root *root;
+	struct ctl_table_set *set;
+	struct ctl_table *entry;
+	struct ctl_dir *dir;
+	int ret;
+
+	/* Get out quickly if not a link */
+	if (!S_ISLNK((*pentry)->mode))
+		return 0;
+
+	ret = 0;
+	spin_lock(&sysctl_lock);
+	root = (*pentry)->data;
+	set = lookup_header_set(root, namespaces);
+	dir = xlate_dir(set, (*phead)->parent);
+	if (IS_ERR(dir))
+		ret = PTR_ERR(dir);
+	else {
+		const char *procname = (*pentry)->procname;
+		head = NULL;
+		entry = find_entry(&head, dir, procname, strlen(procname));
+		ret = -ENOENT;
+		if (entry && use_table(head)) {
+			unuse_table(*phead);
+			*phead = head;
+			*pentry = entry;
+			ret = 0;
+		}
+	}
+
+	spin_unlock(&sysctl_lock);
+	return ret;
+}
+
 static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	struct ctl_table *table)
 {
@@ -859,30 +909,21 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	return error;
 }
 
-static int sysctl_check_dups(struct nsproxy *namespaces,
-	struct ctl_dir *dir,
+static int sysctl_check_dups(struct ctl_dir *dir,
 	const char *path, struct ctl_table *table)
 {
-	struct ctl_table_root *root;
 	struct ctl_table_set *set;
 	struct ctl_table_header *head;
 	int error = 0;
 
-	root = &sysctl_table_root;
-	do {
-		set = lookup_header_set(root, namespaces);
-
-		list_for_each_entry(head, &set->list, ctl_entry) {
-			if (head->unregistering)
-				continue;
-			if (head->parent != dir)
-				continue;
-			error = sysctl_check_table_dups(path, head->ctl_table,
-							table);
-		}
-		root = list_entry(root->root_list.next,
-				  struct ctl_table_root, root_list);
-	} while (root != &sysctl_table_root);
+	set = dir->header.set;
+	list_for_each_entry(head, &set->list, ctl_entry) {
+		if (head->unregistering)
+			continue;
+		if (head->parent != dir)
+			continue;
+		error = sysctl_check_table_dups(path, head->ctl_table, table);
+	}
 	return error;
 }
 
@@ -932,6 +973,115 @@ static int sysctl_check_table(const char *path, struct ctl_table *table)
 	return err;
 }
 
+static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table,
+	struct ctl_table_root *link_root)
+{
+	struct ctl_table *link_table, *entry, *link;
+	struct ctl_table_header *links;
+	char *link_name;
+	int nr_entries, name_bytes;
+
+	name_bytes = 0;
+	nr_entries = 0;
+	for (entry = table; entry->procname; entry++) {
+		nr_entries++;
+		name_bytes += strlen(entry->procname) + 1;
+	}
+
+	links = kzalloc(sizeof(struct ctl_table_header) +
+			sizeof(struct ctl_table)*(nr_entries + 1) +
+			name_bytes,
+			GFP_KERNEL);
+
+	if (!links)
+		return NULL;
+
+	link_table = (struct ctl_table *)(links + 1);
+	link_name = (char *)&link_table[nr_entries + 1];
+
+	for (link = link_table, entry = table; entry->procname; link++, entry++) {
+		int len = strlen(entry->procname) + 1;
+		memcpy(link_name, entry->procname, len);
+		link->procname = link_name;
+		link->mode = S_IFLNK|S_IRWXUGO;
+		link->data = link_root;
+		link_name += len;
+	}
+	init_header(links, dir->header.root, dir->header.set, link_table);
+	links->nreg = nr_entries;
+
+	return links;
+}
+
+static bool get_links(struct ctl_dir *dir,
+	struct ctl_table *table, struct ctl_table_root *link_root)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *entry, *link;
+
+	/* Are there links available for every entry in table? */
+	for (entry = table; entry->procname; entry++) {
+		const char *procname = entry->procname;
+		link = find_entry(&head, dir, procname, strlen(procname));
+		if (!link)
+			return false;
+		if (S_ISDIR(link->mode) && S_ISDIR(entry->mode))
+			continue;
+		if (S_ISLNK(link->mode) && (link->data == link_root))
+			continue;
+		return false;
+	}
+
+	/* The checks passed.  Increase the registration count on the links */
+	for (entry = table; entry->procname; entry++) {
+		const char *procname = entry->procname;
+		link = find_entry(&head, dir, procname, strlen(procname));
+		head->nreg++;
+	}
+	return true;
+}
+
+static int insert_links(struct ctl_table_header *head)
+{
+	struct ctl_table_set *root_set = &sysctl_table_root.default_set;
+	struct ctl_dir *core_parent = NULL;
+	struct ctl_table_header *links;
+	int err;
+
+	if (head->set == root_set)
+		return 0;
+
+	core_parent = xlate_dir(root_set, head->parent);
+	if (IS_ERR(core_parent))
+		return 0;
+
+	if (get_links(core_parent, head->ctl_table, head->root))
+		return 0;
+
+	core_parent->header.nreg++;
+	spin_unlock(&sysctl_lock);
+
+	links = new_links(core_parent, head->ctl_table, head->root);
+
+	spin_lock(&sysctl_lock);
+	err = -ENOMEM;
+	if (!links)
+		goto out;
+
+	err = 0;
+	if (get_links(core_parent, head->ctl_table, head->root)) {
+		kfree(links);
+		goto out;
+	}
+
+	err = insert_header(core_parent, links);
+	if (err)
+		kfree(links);
+out:
+	drop_sysctl_table(&core_parent->header);
+	return err;
+}
+
 /**
  * __register_sysctl_table - register a leaf sysctl table
  * @root: List of sysctl headers to register on
@@ -980,6 +1130,7 @@ struct ctl_table_header *__register_sysctl_table(
 	struct nsproxy *namespaces,
 	const char *path, struct ctl_table *table)
 {
+	struct ctl_table_header *links = NULL;
 	struct ctl_table_header *header;
 	const char *name, *nextname;
 	struct ctl_table_set *set;
@@ -995,7 +1146,7 @@ struct ctl_table_header *__register_sysctl_table(
 
 	spin_lock(&sysctl_lock);
 	header->set = set = lookup_header_set(root, namespaces);
-	dir = &sysctl_root_dir;
+	dir = &set->dir;
 	dir->header.nreg++;
 	spin_unlock(&sysctl_lock);
 
@@ -1012,22 +1163,28 @@ struct ctl_table_header *__register_sysctl_table(
 		if (namelen == 0)
 			continue;
 
-		dir = get_subdir(set, dir, name, namelen);
+		dir = get_subdir(dir, name, namelen);
 		if (IS_ERR(dir))
 			goto fail;
 	}
+
 	spin_lock(&sysctl_lock);
-	if (sysctl_check_dups(namespaces, dir, path, table))
+	if (sysctl_check_dups(dir, path, table))
+		goto fail_put_dir_locked;
+
+	if (insert_header(dir, header))
 		goto fail_put_dir_locked;
-	insert_header(dir, header);
+
 	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 
 	return header;
+
 fail_put_dir_locked:
 	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 fail:
+	kfree(links);
 	kfree(header);
 	dump_stack();
 	return NULL;
@@ -1249,6 +1406,40 @@ struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
 }
 EXPORT_SYMBOL(register_sysctl_table);
 
+static void put_links(struct ctl_table_header *header)
+{
+	struct ctl_table_set *root_set = &sysctl_table_root.default_set;
+	struct ctl_table_root *root = header->root;
+	struct ctl_dir *parent = header->parent;
+	struct ctl_dir *core_parent;
+	struct ctl_table *entry;
+
+	if (header->set == root_set)
+		return;
+
+	core_parent = xlate_dir(root_set, parent);
+	if (IS_ERR(core_parent))
+		return;
+
+	for (entry = header->ctl_table; entry->procname; entry++) {
+		struct ctl_table_header *link_head;
+		struct ctl_table *link;
+		const char *name = entry->procname;
+
+		link = find_entry(&link_head, core_parent, name, strlen(name));
+		if (link &&
+		    ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) ||
+		     (S_ISLNK(link->mode) && (link->data == root)))) {
+			drop_sysctl_table(link_head);
+		}
+		else {
+			printk(KERN_ERR "sysctl link missing during unregister: ");
+			sysctl_print_dir(parent);
+			printk(KERN_CONT "/%s\n", name);
+		}
+	}
+}
+
 static void drop_sysctl_table(struct ctl_table_header *header)
 {
 	struct ctl_dir *parent = header->parent;
@@ -1256,6 +1447,7 @@ static void drop_sysctl_table(struct ctl_table_header *header)
 	if (--header->nreg)
 		return;
 
+	put_links(header);
 	start_unregistering(header);
 	if (!--header->count)
 		kfree_rcu(header, rcu);
@@ -1301,13 +1493,14 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 }
 EXPORT_SYMBOL(unregister_sysctl_table);
 
-void setup_sysctl_set(struct ctl_table_set *p,
+void setup_sysctl_set(struct ctl_table_set *set,
 	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *))
 {
-	INIT_LIST_HEAD(&p->list);
-	p->root = root;
-	p->is_seen = is_seen;
+	memset(set, sizeof(*set), 0);
+	INIT_LIST_HEAD(&set->list);
+	set->is_seen = is_seen;
+	init_header(&set->dir.header, root, set, root_table);
 }
 
 void retire_sysctl_set(struct ctl_table_set *set)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 3084b62..2a1446a 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1051,12 +1051,11 @@ struct ctl_dir {
 
 struct ctl_table_set {
 	struct list_head list;
-	struct ctl_table_root *root;
 	int (*is_seen)(struct ctl_table_set *);
+	struct ctl_dir dir;
 };
 
 struct ctl_table_root {
-	struct list_head root_list;
 	struct ctl_table_set default_set;
 	struct ctl_table_set *(*lookup)(struct ctl_table_root *root,
 					   struct nsproxy *namespaces);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (23 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header Eric W. Biederman
                   ` (33 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

An nsproxy argument here has always been awkard and now the nsproxy argument
is completely unnecessary so remove it, replacing it with the set we want
the registered tables to show up in.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   30 ++++++++++++------------------
 include/linux/sysctl.h |    4 ++--
 net/sysctl_net.c       |   10 +++-------
 3 files changed, 17 insertions(+), 27 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index ec54a57..e0d3e7e 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1084,8 +1084,7 @@ out:
 
 /**
  * __register_sysctl_table - register a leaf sysctl table
- * @root: List of sysctl headers to register on
- * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @set: Sysctl tree to register on
  * @path: The path to the directory the sysctl table is in.
  * @table: the top-level table structure
  *
@@ -1126,26 +1125,24 @@ out:
  * to the table header on success.
  */
 struct ctl_table_header *__register_sysctl_table(
-	struct ctl_table_root *root,
-	struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const char *path, struct ctl_table *table)
 {
+	struct ctl_table_root *root = set->dir.header.root;
 	struct ctl_table_header *links = NULL;
 	struct ctl_table_header *header;
 	const char *name, *nextname;
-	struct ctl_table_set *set;
 	struct ctl_dir *dir;
 
 	header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
 	if (!header)
 		return NULL;
 
-	init_header(header, root, NULL, table);
+	init_header(header, root, set, table);
 	if (sysctl_check_table(path, table))
 		goto fail;
 
 	spin_lock(&sysctl_lock);
-	header->set = set = lookup_header_set(root, namespaces);
 	dir = &set->dir;
 	dir->header.nreg++;
 	spin_unlock(&sysctl_lock);
@@ -1223,8 +1220,7 @@ static int count_subheaders(struct ctl_table *table)
 }
 
 static int register_leaf_sysctl_tables(const char *path, char *pos,
-	struct ctl_table_header ***subheader,
-	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table_header ***subheader, struct ctl_table_set *set,
 	struct ctl_table *table)
 {
 	struct ctl_table *ctl_table_arg = NULL;
@@ -1261,7 +1257,7 @@ static int register_leaf_sysctl_tables(const char *path, char *pos,
 	/* Register everything except a directory full of subdirectories */
 	if (nr_files || !nr_dirs) {
 		struct ctl_table_header *header;
-		header = __register_sysctl_table(root, namespaces, path, files);
+		header = __register_sysctl_table(set, path, files);
 		if (!header) {
 			kfree(ctl_table_arg);
 			goto out;
@@ -1286,7 +1282,7 @@ static int register_leaf_sysctl_tables(const char *path, char *pos,
 			goto out;
 
 		err = register_leaf_sysctl_tables(path, child_pos, subheader,
-						  root, namespaces, entry->child);
+						  set, entry->child);
 		pos[0] = '\0';
 		if (err)
 			goto out;
@@ -1299,8 +1295,7 @@ out:
 
 /**
  * __register_sysctl_paths - register a sysctl table hierarchy
- * @root: List of sysctl headers to register on
- * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @set: Sysctl tree to register on
  * @path: The path to the directory the sysctl table is in.
  * @table: the top-level table structure
  *
@@ -1310,8 +1305,7 @@ out:
  * See __register_sysctl_table for more details.
  */
 struct ctl_table_header *__register_sysctl_paths(
-	struct ctl_table_root *root,
-	struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const struct ctl_path *path, struct ctl_table *table)
 {
 	struct ctl_table *ctl_table_arg = table;
@@ -1337,7 +1331,7 @@ struct ctl_table_header *__register_sysctl_paths(
 		table = table->child;
 	}
 	if (nr_subheaders == 1) {
-		header = __register_sysctl_table(root, namespaces, new_path, table);
+		header = __register_sysctl_table(set, new_path, table);
 		if (header)
 			header->ctl_table_arg = ctl_table_arg;
 	} else {
@@ -1351,7 +1345,7 @@ struct ctl_table_header *__register_sysctl_paths(
 		header->ctl_table_arg = ctl_table_arg;
 
 		if (register_leaf_sysctl_tables(new_path, pos, &subheader,
-						root, namespaces, table))
+						set, table))
 			goto err_register_leaves;
 	}
 
@@ -1384,7 +1378,7 @@ err_register_leaves:
 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 						struct ctl_table *table)
 {
-	return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
+	return __register_sysctl_paths(&sysctl_table_root.default_set,
 					path, table);
 }
 EXPORT_SYMBOL(register_sysctl_paths);
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 2a1446a..cec5941 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1079,10 +1079,10 @@ extern void retire_sysctl_set(struct ctl_table_set *set);
 
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_table(
-	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const char *path, struct ctl_table *table);
 struct ctl_table_header *__register_sysctl_paths(
-	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const struct ctl_path *path, struct ctl_table *table);
 struct ctl_table_header *register_sysctl_table(struct ctl_table * table);
 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index e998c64..c3e65ae 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -105,19 +105,15 @@ subsys_initcall(net_sysctl_init);
 struct ctl_table_header *register_net_sysctl_table(struct net *net,
 	const struct ctl_path *path, struct ctl_table *table)
 {
-	struct nsproxy namespaces;
-	namespaces = *current->nsproxy;
-	namespaces.net_ns = net;
-	return __register_sysctl_paths(&net_sysctl_root,
-					&namespaces, path, table);
+	return __register_sysctl_paths(&net->sysctls, path, table);
 }
 EXPORT_SYMBOL_GPL(register_net_sysctl_table);
 
 struct ctl_table_header *register_net_sysctl_rotable(const
 		struct ctl_path *path, struct ctl_table *table)
 {
-	return __register_sysctl_paths(&net_sysctl_ro_root,
-			&init_nsproxy, path, table);
+	return __register_sysctl_paths(&net_sysctl_ro_root.default_set,
+					path, table);
 }
 EXPORT_SYMBOL_GPL(register_net_sysctl_rotable);
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (24 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 27/29] sysctl: Make the header lists per directory Eric W. Biederman
                   ` (32 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Simplify the callers of insert_header by removing explicit calls to check
for duplicates and instead have insert_header do the work.

This makes the code slightly more maintainable by enabling changes to
data structures where the insertion of new entries without duplicate
suppression is not possible.

There is not always a convenient path string where insert_header
is called so modify sysctl_check_dups to use sysctl_print_dir
when printing the full path when a duplicate is discovered.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   20 +++++++++++---------
 1 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e0d3e7e..160d578 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -52,6 +52,7 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	struct ctl_table **pentry, struct nsproxy *namespaces);
 static int insert_links(struct ctl_table_header *head);
 static void put_links(struct ctl_table_header *header);
+static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table);
 
 static void sysctl_print_dir(struct ctl_dir *dir)
 {
@@ -123,6 +124,10 @@ static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
 	int err;
 
+	err = sysctl_check_dups(dir, header->ctl_table);
+	if (err)
+		return err;
+
 	dir->header.nreg++;
 	header->parent = dir;
 	err = insert_links(header);
@@ -891,7 +896,7 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	return ret;
 }
 
-static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
+static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
 	struct ctl_table *table)
 {
 	struct ctl_table *entry, *test;
@@ -900,8 +905,9 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	for (entry = old; entry->procname; entry++) {
 		for (test = table; test->procname; test++) {
 			if (strcmp(entry->procname, test->procname) == 0) {
-				printk(KERN_ERR "sysctl duplicate entry: %s/%s\n",
-					path, test->procname);
+				printk(KERN_ERR "sysctl duplicate entry: ");
+				sysctl_print_dir(dir);
+				printk(KERN_CONT "/%s\n", test->procname);
 				error = -EEXIST;
 			}
 		}
@@ -909,8 +915,7 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	return error;
 }
 
-static int sysctl_check_dups(struct ctl_dir *dir,
-	const char *path, struct ctl_table *table)
+static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
 {
 	struct ctl_table_set *set;
 	struct ctl_table_header *head;
@@ -922,7 +927,7 @@ static int sysctl_check_dups(struct ctl_dir *dir,
 			continue;
 		if (head->parent != dir)
 			continue;
-		error = sysctl_check_table_dups(path, head->ctl_table, table);
+		error = sysctl_check_table_dups(dir, head->ctl_table, table);
 	}
 	return error;
 }
@@ -1166,9 +1171,6 @@ struct ctl_table_header *__register_sysctl_table(
 	}
 
 	spin_lock(&sysctl_lock);
-	if (sysctl_check_dups(dir, path, table))
-		goto fail_put_dir_locked;
-
 	if (insert_header(dir, header))
 		goto fail_put_dir_locked;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 27/29] sysctl: Make the header lists per directory.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (25 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 28/29] sysctl: Index sysctl directories with rbtrees Eric W. Biederman
                   ` (31 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

Slightly enhance efficiency and clarity of the code by making the
header list per directory instead of per set.

Benchmark before:
    make-dummies 0 999 -> 0.63s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 2m35s
    rmmod dummy         -> 18s

Benchmark after:
    make-dummies 0 999 -> 0.32s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 1m17s
    rmmod dummy         -> 17s

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   28 +++++++++++-----------------
 include/linux/sysctl.h |    2 +-
 2 files changed, 12 insertions(+), 18 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 160d578..e971ccc 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -33,12 +33,12 @@ static struct ctl_table root_table[] = {
 	{ }
 };
 static struct ctl_table_root sysctl_table_root = {
-	.default_set.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),
+	.default_set.dir.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.list),
 	.default_set.dir.header = {
 		{{.count = 1,
 		  .nreg = 1,
 		  .ctl_table = root_table,
-		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),}},
 		.ctl_table_arg = root_table,
 		.root = &sysctl_table_root,
 		.set = &sysctl_table_root.default_set,
@@ -79,15 +79,12 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
 	struct ctl_dir *dir, const char *name, int namelen)
 {
-	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
-	list_for_each_entry(head, &set->list, ctl_entry) {
+	list_for_each_entry(head, &dir->list, ctl_entry) {
 		if (head->unregistering)
 			continue;
-		if (head->parent != dir)
-			continue;
 		for (entry = head->ctl_table; entry->procname; entry++) {
 			const char *procname = entry->procname;
 			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
@@ -133,7 +130,7 @@ static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 	err = insert_links(header);
 	if (err)
 		goto fail_links;
-	list_add_tail(&header->ctl_entry, &header->set->list);
+	list_add_tail(&header->ctl_entry, &header->parent->list);
 	return 0;
 fail_links:
 	header->parent = NULL;
@@ -247,14 +244,12 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
 						  struct list_head *tmp)
 {
-	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 
-	for (tmp = tmp->next; tmp != &set->list; tmp = tmp->next) {
+	for (tmp = tmp->next; tmp != &dir->list; tmp = tmp->next) {
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
 
-		if (head->parent != dir ||
-		    !head->ctl_table->procname ||
+		if (!head->ctl_table->procname ||
 		    !use_table(head))
 			continue;
 
@@ -270,7 +265,7 @@ static void first_entry(struct ctl_dir *dir,
 	struct ctl_table *entry = NULL;
 
 	spin_lock(&sysctl_lock);
-	head = next_usable_entry(dir, &dir->header.set->list);
+	head = next_usable_entry(dir, &dir->list);
 	spin_unlock(&sysctl_lock);
 	if (head)
 		entry = head->ctl_table;
@@ -793,6 +788,7 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 	new_name = (char *)(table + 2);
 	memcpy(new_name, name, namelen);
 	new_name[namelen] = '\0';
+	INIT_LIST_HEAD(&new->list);
 	table[0].procname = new_name;
 	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
 	init_header(&new->header, set->dir.header.root, set, table);
@@ -917,12 +913,10 @@ static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
 
 static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
 {
-	struct ctl_table_set *set;
 	struct ctl_table_header *head;
 	int error = 0;
 
-	set = dir->header.set;
-	list_for_each_entry(head, &set->list, ctl_entry) {
+	list_for_each_entry(head, &dir->list, ctl_entry) {
 		if (head->unregistering)
 			continue;
 		if (head->parent != dir)
@@ -1494,14 +1488,14 @@ void setup_sysctl_set(struct ctl_table_set *set,
 	int (*is_seen)(struct ctl_table_set *))
 {
 	memset(set, sizeof(*set), 0);
-	INIT_LIST_HEAD(&set->list);
 	set->is_seen = is_seen;
+	INIT_LIST_HEAD(&set->dir.list);
 	init_header(&set->dir.header, root, set, root_table);
 }
 
 void retire_sysctl_set(struct ctl_table_set *set)
 {
-	WARN_ON(!list_empty(&set->list));
+	WARN_ON(!list_empty(&set->dir.list));
 }
 
 int __init proc_sys_init(void)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index cec5941..36dec75 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1047,10 +1047,10 @@ struct ctl_table_header
 struct ctl_dir {
 	/* Header must be at the start of ctl_dir */
 	struct ctl_table_header header;
+	struct list_head list;
 };
 
 struct ctl_table_set {
-	struct list_head list;
 	int (*is_seen)(struct ctl_table_set *);
 	struct ctl_dir dir;
 };
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 28/29] sysctl: Index sysctl directories with rbtrees.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (26 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 27/29] sysctl: Make the header lists per directory Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:50 ` [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users Eric W. Biederman
                   ` (30 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

One of the most important jobs of sysctl is to export network stack
tunables.  Several of those tunables are per network device.  In
several instances people are running with 1000+ network devices in
there network stacks, which makes the simple per directory linked list
in sysctl a scaling bottleneck.   Replace O(N^2) sysctl insertion and
lookup times with O(NlogN) by using an rbtree to index the sysctl
directories.

Benchmark before:
    make-dummies 0 999 -> 0.32s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 1m17s
    rmmod dummy         -> 17s

Benchmark after:
    make-dummies 0 999 -> 0.074s
    rmmod dummy        -> 0.070s
    make-dummies 0 9999 -> 3.4s
    rmmod dummy         -> 0.44s

Benchmark after (without dev_snmp6):
    make-dummies 0 9999 -> 0.75s
    rmmod dummy         -> 0.44s
    make-dummies 0 99999 -> 11s
    rmmod dummy          -> 4.3s

At 10,000 dummy devices the bottleneck becomes the time to add and
remove the files under /proc/sys/net/dev_snmp6.  I have commented
out the code that adds and removes files under /proc/sys/net/dev_snmp6
and taken measurments of creating and destroying 100,000 dummies to
verify the sysctl continues to scale.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  224 +++++++++++++++++++++++++++++-------------------
 include/linux/sysctl.h |   10 ++-
 2 files changed, 142 insertions(+), 92 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e971ccc..05c393a 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -33,12 +33,10 @@ static struct ctl_table root_table[] = {
 	{ }
 };
 static struct ctl_table_root sysctl_table_root = {
-	.default_set.dir.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.list),
 	.default_set.dir.header = {
 		{{.count = 1,
 		  .nreg = 1,
-		  .ctl_table = root_table,
-		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),}},
+		  .ctl_table = root_table }},
 		.ctl_table_arg = root_table,
 		.root = &sysctl_table_root,
 		.set = &sysctl_table_root.default_set,
@@ -52,7 +50,6 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	struct ctl_table **pentry, struct nsproxy *namespaces);
 static int insert_links(struct ctl_table_header *head);
 static void put_links(struct ctl_table_header *header);
-static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table);
 
 static void sysctl_print_dir(struct ctl_dir *dir)
 {
@@ -81,28 +78,83 @@ static struct ctl_table *find_entry(struct ctl_table_header **phead,
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
+	struct rb_node *node = dir->root.rb_node;
 
-	list_for_each_entry(head, &dir->list, ctl_entry) {
-		if (head->unregistering)
-			continue;
-		for (entry = head->ctl_table; entry->procname; entry++) {
-			const char *procname = entry->procname;
-			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
-				*phead = head;
-				return entry;
-			}
+	while (node)
+	{
+		struct ctl_node *ctl_node;
+		const char *procname;
+		int cmp;
+
+		ctl_node = rb_entry(node, struct ctl_node, node);
+		head = ctl_node->header;
+		entry = &head->ctl_table[ctl_node - head->node];
+		procname = entry->procname;
+
+		cmp = namecmp(name, namelen, procname, strlen(procname));
+		if (cmp < 0)
+			node = node->rb_left;
+		else if (cmp > 0)
+			node = node->rb_right;
+		else {
+			*phead = head;
+			return entry;
 		}
 	}
 	return NULL;
 }
 
+static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry)
+{
+	struct rb_node *node = &head->node[entry - head->ctl_table].node;
+	struct rb_node **p = &head->parent->root.rb_node;
+	struct rb_node *parent = NULL;
+	const char *name = entry->procname;
+	int namelen = strlen(name);
+
+	while (*p) {
+		struct ctl_table_header *parent_head;
+		struct ctl_table *parent_entry;
+		struct ctl_node *parent_node;
+		const char *parent_name;
+		int cmp;
+
+		parent = *p;
+		parent_node = rb_entry(parent, struct ctl_node, node);
+		parent_head = parent_node->header;
+		parent_entry = &parent_head->ctl_table[parent_node - parent_head->node];
+		parent_name = parent_entry->procname;
+
+		cmp = namecmp(name, namelen, parent_name, strlen(parent_name));
+		if (cmp < 0)
+			p = &(*p)->rb_left;
+		else if (cmp > 0)
+			p = &(*p)->rb_right;
+		else {
+			printk(KERN_ERR "sysctl duplicate entry: ");
+			sysctl_print_dir(head->parent);
+			printk(KERN_CONT "/%s\n", entry->procname);
+			return -EEXIST;
+		}
+	}
+
+	rb_link_node(node, parent, p);
+	return 0;
+}
+
+static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry)
+{
+	struct rb_node *node = &head->node[entry - head->ctl_table].node;
+
+	rb_erase(node, &head->parent->root);
+}
+
 static void init_header(struct ctl_table_header *head,
 	struct ctl_table_root *root, struct ctl_table_set *set,
-	struct ctl_table *table)
+	struct ctl_node *node, struct ctl_table *table)
 {
 	head->ctl_table = table;
 	head->ctl_table_arg = table;
-	INIT_LIST_HEAD(&head->ctl_entry);
 	head->used = 0;
 	head->count = 1;
 	head->nreg = 1;
@@ -110,28 +162,42 @@ static void init_header(struct ctl_table_header *head,
 	head->root = root;
 	head->set = set;
 	head->parent = NULL;
+	head->node = node;
+	if (node) {
+		struct ctl_table *entry;
+		for (entry = table; entry->procname; entry++, node++) {
+			rb_init_node(&node->node);
+			node->header = head;
+		}
+	}
 }
 
 static void erase_header(struct ctl_table_header *head)
 {
-	list_del_init(&head->ctl_entry);
+	struct ctl_table *entry;
+	for (entry = head->ctl_table; entry->procname; entry++)
+		erase_entry(head, entry);
 }
 
 static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
+	struct ctl_table *entry;
 	int err;
 
-	err = sysctl_check_dups(dir, header->ctl_table);
-	if (err)
-		return err;
-
 	dir->header.nreg++;
 	header->parent = dir;
 	err = insert_links(header);
 	if (err)
 		goto fail_links;
-	list_add_tail(&header->ctl_entry, &header->parent->list);
+	for (entry = header->ctl_table; entry->procname; entry++) {
+		err = insert_entry(header, entry);
+		if (err)
+			goto fail;
+	}
 	return 0;
+fail:
+	erase_header(header);
+	put_links(header);
 fail_links:
 	header->parent = NULL;
 	drop_sysctl_table(&dir->header);
@@ -241,19 +307,14 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	return entry;
 }
 
-static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
-						  struct list_head *tmp)
+static struct ctl_node *first_usable_entry(struct rb_node *node)
 {
-	struct ctl_table_header *head;
-
-	for (tmp = tmp->next; tmp != &dir->list; tmp = tmp->next) {
-		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
+	struct ctl_node *ctl_node;
 
-		if (!head->ctl_table->procname ||
-		    !use_table(head))
-			continue;
-
-		return head;
+	for (;node; node = rb_next(node)) {
+		ctl_node = rb_entry(node, struct ctl_node, node);
+		if (use_table(ctl_node->header))
+			return ctl_node;
 	}
 	return NULL;
 }
@@ -261,14 +322,17 @@ static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
 static void first_entry(struct ctl_dir *dir,
 	struct ctl_table_header **phead, struct ctl_table **pentry)
 {
-	struct ctl_table_header *head;
+	struct ctl_table_header *head = NULL;
 	struct ctl_table *entry = NULL;
+	struct ctl_node *ctl_node;
 
 	spin_lock(&sysctl_lock);
-	head = next_usable_entry(dir, &dir->list);
+	ctl_node = first_usable_entry(rb_first(&dir->root));
 	spin_unlock(&sysctl_lock);
-	if (head)
-		entry = head->ctl_table;
+	if (ctl_node) {
+		head = ctl_node->header;
+		entry = &head->ctl_table[ctl_node - head->node];
+	}
 	*phead = head;
 	*pentry = entry;
 }
@@ -277,15 +341,17 @@ static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentr
 {
 	struct ctl_table_header *head = *phead;
 	struct ctl_table *entry = *pentry;
+	struct ctl_node *ctl_node = &head->node[entry - head->ctl_table];
 
-	entry++;
-	if (!entry->procname) {
-		spin_lock(&sysctl_lock);
-		unuse_table(head);
-		head = next_usable_entry(head->parent, &head->ctl_entry);
-		spin_unlock(&sysctl_lock);
-		if (head)
-			entry = head->ctl_table;
+	spin_lock(&sysctl_lock);
+	unuse_table(head);
+
+	ctl_node = first_usable_entry(rb_next(&ctl_node->node));
+	spin_unlock(&sysctl_lock);
+	head = NULL;
+	if (ctl_node) {
+		head = ctl_node->header;
+		entry = &head->ctl_table[ctl_node - head->node];
 	}
 	*phead = head;
 	*pentry = entry;
@@ -777,21 +843,23 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 {
 	struct ctl_table *table;
 	struct ctl_dir *new;
+	struct ctl_node *node;
 	char *new_name;
 
-	new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 +
-		      namelen + 1, GFP_KERNEL);
+	new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) +
+		      sizeof(struct ctl_table)*2 +  namelen + 1,
+		      GFP_KERNEL);
 	if (!new)
 		return NULL;
 
-	table = (struct ctl_table *)(new + 1);
+	node = (struct ctl_node *)(new + 1);
+	table = (struct ctl_table *)(node + 1);
 	new_name = (char *)(table + 2);
 	memcpy(new_name, name, namelen);
 	new_name[namelen] = '\0';
-	INIT_LIST_HEAD(&new->list);
 	table[0].procname = new_name;
 	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
-	init_header(&new->header, set->dir.header.root, set, table);
+	init_header(&new->header, set->dir.header.root, set, node, table);
 
 	return new;
 }
@@ -892,40 +960,6 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	return ret;
 }
 
-static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
-	struct ctl_table *table)
-{
-	struct ctl_table *entry, *test;
-	int error = 0;
-
-	for (entry = old; entry->procname; entry++) {
-		for (test = table; test->procname; test++) {
-			if (strcmp(entry->procname, test->procname) == 0) {
-				printk(KERN_ERR "sysctl duplicate entry: ");
-				sysctl_print_dir(dir);
-				printk(KERN_CONT "/%s\n", test->procname);
-				error = -EEXIST;
-			}
-		}
-	}
-	return error;
-}
-
-static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
-{
-	struct ctl_table_header *head;
-	int error = 0;
-
-	list_for_each_entry(head, &dir->list, ctl_entry) {
-		if (head->unregistering)
-			continue;
-		if (head->parent != dir)
-			continue;
-		error = sysctl_check_table_dups(dir, head->ctl_table, table);
-	}
-	return error;
-}
-
 static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
 {
 	struct va_format vaf;
@@ -977,6 +1011,7 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 {
 	struct ctl_table *link_table, *entry, *link;
 	struct ctl_table_header *links;
+	struct ctl_node *node;
 	char *link_name;
 	int nr_entries, name_bytes;
 
@@ -988,6 +1023,7 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 	}
 
 	links = kzalloc(sizeof(struct ctl_table_header) +
+			sizeof(struct ctl_node)*nr_entries +
 			sizeof(struct ctl_table)*(nr_entries + 1) +
 			name_bytes,
 			GFP_KERNEL);
@@ -995,7 +1031,8 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 	if (!links)
 		return NULL;
 
-	link_table = (struct ctl_table *)(links + 1);
+	node = (struct ctl_node *)(links + 1);
+	link_table = (struct ctl_table *)(node + nr_entries);
 	link_name = (char *)&link_table[nr_entries + 1];
 
 	for (link = link_table, entry = table; entry->procname; link++, entry++) {
@@ -1006,7 +1043,7 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 		link->data = link_root;
 		link_name += len;
 	}
-	init_header(links, dir->header.root, dir->header.set, link_table);
+	init_header(links, dir->header.root, dir->header.set, node, link_table);
 	links->nreg = nr_entries;
 
 	return links;
@@ -1132,12 +1169,20 @@ struct ctl_table_header *__register_sysctl_table(
 	struct ctl_table_header *header;
 	const char *name, *nextname;
 	struct ctl_dir *dir;
+	struct ctl_table *entry;
+	struct ctl_node *node;
+	int nr_entries = 0;
+
+	for (entry = table; entry->procname; entry++)
+		nr_entries++;
 
-	header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
+	header = kzalloc(sizeof(struct ctl_table_header) +
+			 sizeof(struct ctl_node)*nr_entries, GFP_KERNEL);
 	if (!header)
 		return NULL;
 
-	init_header(header, root, set, table);
+	node = (struct ctl_node *)(header + 1);
+	init_header(header, root, set, node, table);
 	if (sysctl_check_table(path, table))
 		goto fail;
 
@@ -1489,13 +1534,12 @@ void setup_sysctl_set(struct ctl_table_set *set,
 {
 	memset(set, sizeof(*set), 0);
 	set->is_seen = is_seen;
-	INIT_LIST_HEAD(&set->dir.list);
-	init_header(&set->dir.header, root, set, root_table);
+	init_header(&set->dir.header, root, set, NULL, root_table);
 }
 
 void retire_sysctl_set(struct ctl_table_set *set)
 {
-	WARN_ON(!list_empty(&set->dir.list));
+	WARN_ON(!RB_EMPTY_ROOT(&set->dir.root));
 }
 
 int __init proc_sys_init(void)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 36dec75..35c50ed 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -932,6 +932,7 @@ enum
 #include <linux/list.h>
 #include <linux/rcupdate.h>
 #include <linux/wait.h>
+#include <linux/rbtree.h>
 
 /* For the /proc/sys support */
 struct ctl_table;
@@ -1023,6 +1024,11 @@ struct ctl_table
 	void *extra2;
 };
 
+struct ctl_node {
+	struct rb_node node;
+	struct ctl_table_header *header;
+};
+
 /* struct ctl_table_header is used to maintain dynamic lists of
    struct ctl_table trees. */
 struct ctl_table_header
@@ -1030,7 +1036,6 @@ struct ctl_table_header
 	union {
 		struct {
 			struct ctl_table *ctl_table;
-			struct list_head ctl_entry;
 			int used;
 			int count;
 			int nreg;
@@ -1042,12 +1047,13 @@ struct ctl_table_header
 	struct ctl_table_root *root;
 	struct ctl_table_set *set;
 	struct ctl_dir *parent;
+	struct ctl_node *node;
 };
 
 struct ctl_dir {
 	/* Header must be at the start of ctl_dir */
 	struct ctl_table_header header;
-	struct list_head list;
+	struct rb_root root;
 };
 
 struct ctl_table_set {
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (27 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 28/29] sysctl: Index sysctl directories with rbtrees Eric W. Biederman
@ 2012-01-27  4:50 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
                   ` (29 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:50 UTC (permalink / raw)
  To: linux-kernel; +Cc: linux-fsdevel, netdev, Damien Millescamps, Eric W. Biederman

The plan is to convert all callers of register_sysctl_table
and register_sysctl_paths to register_sysctl.  The interface
to register_sysctl is enough nicer this should make the callers
a bit more readable.  Additionally after the conversion the
230 lines of backwards compatibility can be removed.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   17 +++++++++++++++++
 include/linux/sysctl.h |    1 +
 2 files changed, 18 insertions(+), 0 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 05c393a..8dc7f0e 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1228,6 +1228,23 @@ fail:
 	return NULL;
 }
 
+/**
+ * register_sysctl - register a sysctl table
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the table structure
+ *
+ * Register a sysctl table. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See __register_sysctl_table for more details.
+ */
+struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table)
+{
+	return __register_sysctl_table(&sysctl_table_root.default_set,
+					path, table);
+}
+EXPORT_SYMBOL(register_sysctl);
+
 static char *append_path(const char *path, char *pos, const char *name)
 {
 	int namelen;
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 35c50ed..c34b4c8 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1090,6 +1090,7 @@ struct ctl_table_header *__register_sysctl_table(
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_set *set,
 	const struct ctl_path *path, struct ctl_table *table);
+struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table);
 struct ctl_table_header *register_sysctl_table(struct ctl_table * table);
 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 						struct ctl_table *table);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 01/29] sysctl: remove impossible condition check
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (28 preceding siblings ...)
  2012-01-27  4:50 ` [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling Eric W. Biederman
                   ` (28 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Lucas De Marchi, Eric W. Biederman

From: Lucas De Marchi <lucas.demarchi@profusion.mobi>

Remove checks for conditions that will never happen. If procname is NULL
the loop would already had bailed out, so there's no need to check it
again.

At the same time this also compacts the function find_in_table() by
refactoring it to be easier to read.

Signed-off-by: Lucas De Marchi <lucas.demarchi@profusion.mobi>
Reviewed-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   14 ++------------
 1 files changed, 2 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a6b6217..d82f4a8 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -59,17 +59,11 @@ out:
 
 static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
 {
-	int len;
 	for ( ; p->procname; p++) {
-
-		if (!p->procname)
-			continue;
-
-		len = strlen(p->procname);
-		if (len != name->len)
+		if (strlen(p->procname) != name->len)
 			continue;
 
-		if (memcmp(p->procname, name->name, len) != 0)
+		if (memcmp(p->procname, name->name, name->len) != 0)
 			continue;
 
 		/* I have a match */
@@ -266,10 +260,6 @@ static int scan(struct ctl_table_header *head, ctl_table *table,
 	for (; table->procname; table++, (*pos)++) {
 		int res;
 
-		/* Can't do anything without a proc name */
-		if (!table->procname)
-			continue;
-
 		if (*pos < file->f_pos)
 			continue;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (29 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table Eric W. Biederman
                   ` (27 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

- In sysctl.h move functions only available if CONFIG_SYSCL
  is defined inside of #ifdef CONFIG_SYSCTL

- Move the stub function definitions for !CONFIG_SYSCTL
  into sysctl.h and make them static inlines.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 include/linux/sysctl.h |   95 +++++++++++++++++++++++++++++++-----------------
 kernel/sysctl.c        |   26 -------------
 2 files changed, 62 insertions(+), 59 deletions(-)

diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index bb9127d..cf3ee7f 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -937,30 +937,8 @@ enum
 struct ctl_table;
 struct nsproxy;
 struct ctl_table_root;
-
-struct ctl_table_set {
-	struct list_head list;
-	struct ctl_table_set *parent;
-	int (*is_seen)(struct ctl_table_set *);
-};
-
-extern void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
-	int (*is_seen)(struct ctl_table_set *));
-
 struct ctl_table_header;
 
-extern void sysctl_head_get(struct ctl_table_header *);
-extern void sysctl_head_put(struct ctl_table_header *);
-extern int sysctl_is_seen(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev);
-extern struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-						struct ctl_table_header *prev);
-extern void sysctl_head_finish(struct ctl_table_header *prev);
-extern int sysctl_perm(struct ctl_table_root *root,
-		struct ctl_table *table, int op);
-
 typedef struct ctl_table ctl_table;
 
 typedef int proc_handler (struct ctl_table *ctl, int write,
@@ -1023,8 +1001,6 @@ static inline void *proc_sys_poll_event(struct ctl_table_poll *poll)
 	return (void *)(unsigned long)atomic_read(&poll->event);
 }
 
-void proc_sys_poll_notify(struct ctl_table_poll *poll);
-
 #define __CTL_TABLE_POLL_INITIALIZER(name) {				\
 	.event = ATOMIC_INIT(0),					\
 	.wait = __WAIT_QUEUE_HEAD_INITIALIZER(name.wait) }
@@ -1047,15 +1023,6 @@ struct ctl_table
 	void *extra2;
 };
 
-struct ctl_table_root {
-	struct list_head root_list;
-	struct ctl_table_set default_set;
-	struct ctl_table_set *(*lookup)(struct ctl_table_root *root,
-					   struct nsproxy *namespaces);
-	int (*permissions)(struct ctl_table_root *root,
-			struct nsproxy *namespaces, struct ctl_table *table);
-};
-
 /* struct ctl_table_header is used to maintain dynamic lists of
    struct ctl_table trees. */
 struct ctl_table_header
@@ -1078,11 +1045,45 @@ struct ctl_table_header
 	struct ctl_table_header *parent;
 };
 
+struct ctl_table_set {
+	struct list_head list;
+	struct ctl_table_set *parent;
+	int (*is_seen)(struct ctl_table_set *);
+};
+
+struct ctl_table_root {
+	struct list_head root_list;
+	struct ctl_table_set default_set;
+	struct ctl_table_set *(*lookup)(struct ctl_table_root *root,
+					   struct nsproxy *namespaces);
+	int (*permissions)(struct ctl_table_root *root,
+			struct nsproxy *namespaces, struct ctl_table *table);
+};
+
 /* struct ctl_path describes where in the hierarchy a table is added */
 struct ctl_path {
 	const char *procname;
 };
 
+#ifdef CONFIG_SYSCTL
+
+void proc_sys_poll_notify(struct ctl_table_poll *poll);
+
+extern void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_set *parent,
+	int (*is_seen)(struct ctl_table_set *));
+
+extern void sysctl_head_get(struct ctl_table_header *);
+extern void sysctl_head_put(struct ctl_table_header *);
+extern int sysctl_is_seen(struct ctl_table_header *);
+extern struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *);
+extern struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev);
+extern struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
+						struct ctl_table_header *prev);
+extern void sysctl_head_finish(struct ctl_table_header *prev);
+extern int sysctl_perm(struct ctl_table_root *root,
+		struct ctl_table *table, int op);
+
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_root *root, struct nsproxy *namespaces,
@@ -1094,6 +1095,34 @@ struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 void unregister_sysctl_table(struct ctl_table_header * table);
 int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table);
 
+#else /* CONFIG_SYSCTL */
+static inline struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
+{
+	return NULL;
+}
+
+static inline struct ctl_table_header *register_sysctl_paths(
+			const struct ctl_path *path, struct ctl_table *table)
+{
+	return NULL;
+}
+
+static inline void unregister_sysctl_table(struct ctl_table_header * table)
+{
+}
+
+static inline void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_set *parent,
+	int (*is_seen)(struct ctl_table_set *))
+{
+}
+
+static inline void sysctl_head_put(struct ctl_table_header *head)
+{
+}
+
+#endif /* CONFIG_SYSCTL */
+
 #endif /* __KERNEL__ */
 
 #endif /* _LINUX_SYSCTL_H */
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index f487f25..d5bbddd 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2017,32 +2017,6 @@ void setup_sysctl_set(struct ctl_table_set *p,
 	p->is_seen = is_seen;
 }
 
-#else /* !CONFIG_SYSCTL */
-struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
-{
-	return NULL;
-}
-
-struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
-						    struct ctl_table *table)
-{
-	return NULL;
-}
-
-void unregister_sysctl_table(struct ctl_table_header * table)
-{
-}
-
-void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
-	int (*is_seen)(struct ctl_table_set *))
-{
-}
-
-void sysctl_head_put(struct ctl_table_header *head)
-{
-}
-
 #endif /* CONFIG_SYSCTL */
 
 /*
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (30 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c Eric W. Biederman
                   ` (26 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Simplify the code by treating the base sysctl table like any other
sysctl table and register it with register_sysctl_table.

To ensure this table is registered early enough to avoid problems
call sysctl_init from proc_sys_init.

Rename sysctl_net.c:sysctl_init() to net_sysctl_init() to avoid
name conflicts now that kernel/sysctl.c:sysctl_init() is no longer
static.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |    3 ++-
 include/linux/sysctl.h |    1 +
 kernel/sysctl.c        |   13 ++++---------
 net/sysctl_net.c       |    4 ++--
 4 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index d82f4a8..9d29d28 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -468,5 +468,6 @@ int __init proc_sys_init(void)
 	proc_sys_root->proc_iops = &proc_sys_dir_operations;
 	proc_sys_root->proc_fops = &proc_sys_dir_file_operations;
 	proc_sys_root->nlink = 0;
-	return 0;
+
+	return sysctl_init();
 }
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index cf3ee7f..5e3532e 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1095,6 +1095,7 @@ struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 void unregister_sysctl_table(struct ctl_table_header * table);
 int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table);
 
+extern int sysctl_init(void);
 #else /* CONFIG_SYSCTL */
 static inline struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
 {
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index d5bbddd..ad46024 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -192,7 +192,7 @@ static int sysrq_sysctl_handler(ctl_table *table, int write,
 
 #endif
 
-static struct ctl_table root_table[];
+static struct ctl_table root_table[1];
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
 	{{.count = 1,
@@ -222,7 +222,7 @@ int sysctl_legacy_va_layout;
 
 /* The default sysctl tables: */
 
-static struct ctl_table root_table[] = {
+static struct ctl_table sysctl_base_table[] = {
 	{
 		.procname	= "kernel",
 		.mode		= 0555,
@@ -1747,17 +1747,12 @@ static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
 	}
 }
 
-static __init int sysctl_init(void)
+int __init sysctl_init(void)
 {
-	sysctl_set_parent(NULL, root_table);
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-	sysctl_check_table(current->nsproxy, root_table);
-#endif
+	register_sysctl_table(sysctl_base_table);
 	return 0;
 }
 
-core_initcall(sysctl_init);
-
 static struct ctl_table *is_branch_in(struct ctl_table *branch,
 				      struct ctl_table *table)
 {
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index e758139..a6bbee2 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -90,7 +90,7 @@ static struct pernet_operations sysctl_pernet_ops = {
 	.exit = sysctl_net_exit,
 };
 
-static __init int sysctl_init(void)
+static __init int net_sysctl_init(void)
 {
 	int ret;
 	ret = register_pernet_subsys(&sysctl_pernet_ops);
@@ -102,7 +102,7 @@ static __init int sysctl_init(void)
 out:
 	return ret;
 }
-subsys_initcall(sysctl_init);
+subsys_initcall(net_sysctl_init);
 
 struct ctl_table_header *register_net_sysctl_table(struct net *net,
 	const struct ctl_path *path, struct ctl_table *table)
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (31 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 05/29] sysctl: Make the directories have nlink == 1 Eric W. Biederman
                   ` (25 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Move the core sysctl code from kernel/sysctl.c and kernel/sysctl_check.c
into fs/proc/proc_sysctl.c.

Currently sysctl maintenance is hampered by the sysctl implementation
being split across 3 files with artificial layering between them.
Consolidate the entire sysctl implementation into 1 file so that
it is easier to see what is going on and hopefully allowing for
simpler maintenance.

For functions that are now only used in fs/proc/proc_sysctl.c remove
their declarations from sysctl.h and make them static in fs/proc/proc_sysctl.c

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/internal.h     |    3 +
 fs/proc/proc_sysctl.c  |  622 ++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/sysctl.h |   16 --
 kernel/Makefile        |    1 -
 kernel/sysctl.c        |  464 ------------------------------------
 kernel/sysctl_check.c  |  160 -------------
 6 files changed, 625 insertions(+), 641 deletions(-)
 delete mode 100644 kernel/sysctl_check.c

diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 2925775..3b5ecd9 100644
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -10,12 +10,15 @@
  */
 
 #include <linux/proc_fs.h>
+struct  ctl_table_header;
 
 extern struct proc_dir_entry proc_root;
 #ifdef CONFIG_PROC_SYSCTL
 extern int proc_sys_init(void);
+extern void sysctl_head_put(struct ctl_table_header *head);
 #else
 static inline void proc_sys_init(void) { }
+static inline void sysctl_head_put(struct ctl_table_header *head) { }
 #endif
 #ifdef CONFIG_NET
 extern int proc_net_init(void);
diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 9d29d28..06e6f10 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -7,6 +7,7 @@
 #include <linux/proc_fs.h>
 #include <linux/security.h>
 #include <linux/namei.h>
+#include <linux/module.h>
 #include "internal.h"
 
 static const struct dentry_operations proc_sys_dentry_operations;
@@ -24,6 +25,209 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
 	wake_up_interruptible(&poll->wait);
 }
 
+static struct ctl_table root_table[1];
+static struct ctl_table_root sysctl_table_root;
+static struct ctl_table_header root_table_header = {
+	{{.count = 1,
+	.ctl_table = root_table,
+	.ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+	.root = &sysctl_table_root,
+	.set = &sysctl_table_root.default_set,
+};
+static struct ctl_table_root sysctl_table_root = {
+	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
+	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+};
+
+static DEFINE_SPINLOCK(sysctl_lock);
+
+/* called under sysctl_lock */
+static int use_table(struct ctl_table_header *p)
+{
+	if (unlikely(p->unregistering))
+		return 0;
+	p->used++;
+	return 1;
+}
+
+/* called under sysctl_lock */
+static void unuse_table(struct ctl_table_header *p)
+{
+	if (!--p->used)
+		if (unlikely(p->unregistering))
+			complete(p->unregistering);
+}
+
+/* called under sysctl_lock, will reacquire if has to wait */
+static void start_unregistering(struct ctl_table_header *p)
+{
+	/*
+	 * if p->used is 0, nobody will ever touch that entry again;
+	 * we'll eliminate all paths to it before dropping sysctl_lock
+	 */
+	if (unlikely(p->used)) {
+		struct completion wait;
+		init_completion(&wait);
+		p->unregistering = &wait;
+		spin_unlock(&sysctl_lock);
+		wait_for_completion(&wait);
+		spin_lock(&sysctl_lock);
+	} else {
+		/* anything non-NULL; we'll never dereference it */
+		p->unregistering = ERR_PTR(-EINVAL);
+	}
+	/*
+	 * do not remove from the list until nobody holds it; walking the
+	 * list in do_sysctl() relies on that.
+	 */
+	list_del_init(&p->ctl_entry);
+}
+
+static void sysctl_head_get(struct ctl_table_header *head)
+{
+	spin_lock(&sysctl_lock);
+	head->count++;
+	spin_unlock(&sysctl_lock);
+}
+
+void sysctl_head_put(struct ctl_table_header *head)
+{
+	spin_lock(&sysctl_lock);
+	if (!--head->count)
+		kfree_rcu(head, rcu);
+	spin_unlock(&sysctl_lock);
+}
+
+static struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
+{
+	if (!head)
+		BUG();
+	spin_lock(&sysctl_lock);
+	if (!use_table(head))
+		head = ERR_PTR(-ENOENT);
+	spin_unlock(&sysctl_lock);
+	return head;
+}
+
+static void sysctl_head_finish(struct ctl_table_header *head)
+{
+	if (!head)
+		return;
+	spin_lock(&sysctl_lock);
+	unuse_table(head);
+	spin_unlock(&sysctl_lock);
+}
+
+static struct ctl_table_set *
+lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
+{
+	struct ctl_table_set *set = &root->default_set;
+	if (root->lookup)
+		set = root->lookup(root, namespaces);
+	return set;
+}
+
+static struct list_head *
+lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
+{
+	struct ctl_table_set *set = lookup_header_set(root, namespaces);
+	return &set->list;
+}
+
+static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
+						struct ctl_table_header *prev)
+{
+	struct ctl_table_root *root;
+	struct list_head *header_list;
+	struct ctl_table_header *head;
+	struct list_head *tmp;
+
+	spin_lock(&sysctl_lock);
+	if (prev) {
+		head = prev;
+		tmp = &prev->ctl_entry;
+		unuse_table(prev);
+		goto next;
+	}
+	tmp = &root_table_header.ctl_entry;
+	for (;;) {
+		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
+
+		if (!use_table(head))
+			goto next;
+		spin_unlock(&sysctl_lock);
+		return head;
+	next:
+		root = head->root;
+		tmp = tmp->next;
+		header_list = lookup_header_list(root, namespaces);
+		if (tmp != header_list)
+			continue;
+
+		do {
+			root = list_entry(root->root_list.next,
+					struct ctl_table_root, root_list);
+			if (root == &sysctl_table_root)
+				goto out;
+			header_list = lookup_header_list(root, namespaces);
+		} while (list_empty(header_list));
+		tmp = header_list->next;
+	}
+out:
+	spin_unlock(&sysctl_lock);
+	return NULL;
+}
+
+static struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
+{
+	return __sysctl_head_next(current->nsproxy, prev);
+}
+
+void register_sysctl_root(struct ctl_table_root *root)
+{
+	spin_lock(&sysctl_lock);
+	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
+	spin_unlock(&sysctl_lock);
+}
+
+/*
+ * sysctl_perm does NOT grant the superuser all rights automatically, because
+ * some sysctl variables are readonly even to root.
+ */
+
+static int test_perm(int mode, int op)
+{
+	if (!current_euid())
+		mode >>= 6;
+	else if (in_egroup_p(0))
+		mode >>= 3;
+	if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
+		return 0;
+	return -EACCES;
+}
+
+static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
+{
+	int mode;
+
+	if (root->permissions)
+		mode = root->permissions(root, current->nsproxy, table);
+	else
+		mode = table->mode;
+
+	return test_perm(mode, op);
+}
+
+static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
+{
+	for (; table->procname; table++) {
+		table->parent = parent;
+		if (table->child)
+			sysctl_set_parent(table, table->child);
+	}
+}
+
+
 static struct inode *proc_sys_make_inode(struct super_block *sb,
 		struct ctl_table_header *head, struct ctl_table *table)
 {
@@ -435,6 +639,21 @@ static int proc_sys_delete(const struct dentry *dentry)
 	return !!PROC_I(dentry->d_inode)->sysctl->unregistering;
 }
 
+static int sysctl_is_seen(struct ctl_table_header *p)
+{
+	struct ctl_table_set *set = p->set;
+	int res;
+	spin_lock(&sysctl_lock);
+	if (p->unregistering)
+		res = 0;
+	else if (!set->is_seen)
+		res = 1;
+	else
+		res = set->is_seen(set);
+	spin_unlock(&sysctl_lock);
+	return res;
+}
+
 static int proc_sys_compare(const struct dentry *parent,
 		const struct inode *pinode,
 		const struct dentry *dentry, const struct inode *inode,
@@ -460,6 +679,409 @@ static const struct dentry_operations proc_sys_dentry_operations = {
 	.d_compare	= proc_sys_compare,
 };
 
+static struct ctl_table *is_branch_in(struct ctl_table *branch,
+				      struct ctl_table *table)
+{
+	struct ctl_table *p;
+	const char *s = branch->procname;
+
+	/* branch should have named subdirectory as its first element */
+	if (!s || !branch->child)
+		return NULL;
+
+	/* ... and nothing else */
+	if (branch[1].procname)
+		return NULL;
+
+	/* table should contain subdirectory with the same name */
+	for (p = table; p->procname; p++) {
+		if (!p->child)
+			continue;
+		if (p->procname && strcmp(p->procname, s) == 0)
+			return p;
+	}
+	return NULL;
+}
+
+/* see if attaching q to p would be an improvement */
+static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
+{
+	struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
+	struct ctl_table *next;
+	int is_better = 0;
+	int not_in_parent = !p->attached_by;
+
+	while ((next = is_branch_in(by, to)) != NULL) {
+		if (by == q->attached_by)
+			is_better = 1;
+		if (to == p->attached_by)
+			not_in_parent = 1;
+		by = by->child;
+		to = next->child;
+	}
+
+	if (is_better && not_in_parent) {
+		q->attached_by = by;
+		q->attached_to = to;
+		q->parent = p;
+	}
+}
+
+#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
+static int sysctl_depth(struct ctl_table *table)
+{
+	struct ctl_table *tmp;
+	int depth;
+
+	depth = 0;
+	for (tmp = table; tmp->parent; tmp = tmp->parent)
+		depth++;
+
+	return depth;
+}
+
+static struct ctl_table *sysctl_parent(struct ctl_table *table, int n)
+{
+	int i;
+
+	for (i = 0; table && i < n; i++)
+		table = table->parent;
+
+	return table;
+}
+
+
+static void sysctl_print_path(struct ctl_table *table)
+{
+	struct ctl_table *tmp;
+	int depth, i;
+	depth = sysctl_depth(table);
+	if (table->procname) {
+		for (i = depth; i >= 0; i--) {
+			tmp = sysctl_parent(table, i);
+			printk("/%s", tmp->procname?tmp->procname:"");
+		}
+	}
+	printk(" ");
+}
+
+static struct ctl_table *sysctl_check_lookup(struct nsproxy *namespaces,
+						struct ctl_table *table)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *ref, *test;
+	int depth, cur_depth;
+
+	depth = sysctl_depth(table);
+
+	for (head = __sysctl_head_next(namespaces, NULL); head;
+	     head = __sysctl_head_next(namespaces, head)) {
+		cur_depth = depth;
+		ref = head->ctl_table;
+repeat:
+		test = sysctl_parent(table, cur_depth);
+		for (; ref->procname; ref++) {
+			int match = 0;
+			if (cur_depth && !ref->child)
+				continue;
+
+			if (test->procname && ref->procname &&
+			    (strcmp(test->procname, ref->procname) == 0))
+					match++;
+
+			if (match) {
+				if (cur_depth != 0) {
+					cur_depth--;
+					ref = ref->child;
+					goto repeat;
+				}
+				goto out;
+			}
+		}
+	}
+	ref = NULL;
+out:
+	sysctl_head_finish(head);
+	return ref;
+}
+
+static void set_fail(const char **fail, struct ctl_table *table, const char *str)
+{
+	if (*fail) {
+		printk(KERN_ERR "sysctl table check failed: ");
+		sysctl_print_path(table);
+		printk(" %s\n", *fail);
+		dump_stack();
+	}
+	*fail = str;
+}
+
+static void sysctl_check_leaf(struct nsproxy *namespaces,
+				struct ctl_table *table, const char **fail)
+{
+	struct ctl_table *ref;
+
+	ref = sysctl_check_lookup(namespaces, table);
+	if (ref && (ref != table))
+		set_fail(fail, table, "Sysctl already exists");
+}
+
+static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
+{
+	int error = 0;
+	for (; table->procname; table++) {
+		const char *fail = NULL;
+
+		if (table->parent) {
+			if (!table->parent->procname)
+				set_fail(&fail, table, "Parent without procname");
+		}
+		if (table->child) {
+			if (table->data)
+				set_fail(&fail, table, "Directory with data?");
+			if (table->maxlen)
+				set_fail(&fail, table, "Directory with maxlen?");
+			if ((table->mode & (S_IRUGO|S_IXUGO)) != table->mode)
+				set_fail(&fail, table, "Writable sysctl directory");
+			if (table->proc_handler)
+				set_fail(&fail, table, "Directory with proc_handler");
+			if (table->extra1)
+				set_fail(&fail, table, "Directory with extra1");
+			if (table->extra2)
+				set_fail(&fail, table, "Directory with extra2");
+		} else {
+			if ((table->proc_handler == proc_dostring) ||
+			    (table->proc_handler == proc_dointvec) ||
+			    (table->proc_handler == proc_dointvec_minmax) ||
+			    (table->proc_handler == proc_dointvec_jiffies) ||
+			    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
+			    (table->proc_handler == proc_dointvec_ms_jiffies) ||
+			    (table->proc_handler == proc_doulongvec_minmax) ||
+			    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
+				if (!table->data)
+					set_fail(&fail, table, "No data");
+				if (!table->maxlen)
+					set_fail(&fail, table, "No maxlen");
+			}
+#ifdef CONFIG_PROC_SYSCTL
+			if (!table->proc_handler)
+				set_fail(&fail, table, "No proc_handler");
+#endif
+			sysctl_check_leaf(namespaces, table, &fail);
+		}
+		if (table->mode > 0777)
+			set_fail(&fail, table, "bogus .mode");
+		if (fail) {
+			set_fail(&fail, table, NULL);
+			error = -EINVAL;
+		}
+		if (table->child)
+			error |= sysctl_check_table(namespaces, table->child);
+	}
+	return error;
+}
+#endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
+
+/**
+ * __register_sysctl_paths - register a sysctl hierarchy
+ * @root: List of sysctl headers to register on
+ * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * The members of the &struct ctl_table structure are used as follows:
+ *
+ * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
+ *            enter a sysctl file
+ *
+ * data - a pointer to data for use by proc_handler
+ *
+ * maxlen - the maximum size in bytes of the data
+ *
+ * mode - the file permissions for the /proc/sys file, and for sysctl(2)
+ *
+ * child - a pointer to the child sysctl table if this entry is a directory, or
+ *         %NULL.
+ *
+ * proc_handler - the text handler routine (described below)
+ *
+ * de - for internal use by the sysctl routines
+ *
+ * extra1, extra2 - extra pointers usable by the proc handler routines
+ *
+ * Leaf nodes in the sysctl tree will be represented by a single file
+ * under /proc; non-leaf nodes will be represented by directories.
+ *
+ * sysctl(2) can automatically manage read and write requests through
+ * the sysctl table.  The data and maxlen fields of the ctl_table
+ * struct enable minimal validation of the values being written to be
+ * performed, and the mode field allows minimal authentication.
+ *
+ * There must be a proc_handler routine for any terminal nodes
+ * mirrored under /proc/sys (non-terminals are handled by a built-in
+ * directory handler).  Several default handlers are available to
+ * cover common cases -
+ *
+ * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
+ * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
+ * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
+ *
+ * It is the handler's job to read the input buffer from user memory
+ * and process it. The handler should return 0 on success.
+ *
+ * This routine returns %NULL on a failure to register, and a pointer
+ * to the table header on success.
+ */
+struct ctl_table_header *__register_sysctl_paths(
+	struct ctl_table_root *root,
+	struct nsproxy *namespaces,
+	const struct ctl_path *path, struct ctl_table *table)
+{
+	struct ctl_table_header *header;
+	struct ctl_table *new, **prevp;
+	unsigned int n, npath;
+	struct ctl_table_set *set;
+
+	/* Count the path components */
+	for (npath = 0; path[npath].procname; ++npath)
+		;
+
+	/*
+	 * For each path component, allocate a 2-element ctl_table array.
+	 * The first array element will be filled with the sysctl entry
+	 * for this, the second will be the sentinel (procname == 0).
+	 *
+	 * We allocate everything in one go so that we don't have to
+	 * worry about freeing additional memory in unregister_sysctl_table.
+	 */
+	header = kzalloc(sizeof(struct ctl_table_header) +
+			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
+	if (!header)
+		return NULL;
+
+	new = (struct ctl_table *) (header + 1);
+
+	/* Now connect the dots */
+	prevp = &header->ctl_table;
+	for (n = 0; n < npath; ++n, ++path) {
+		/* Copy the procname */
+		new->procname = path->procname;
+		new->mode     = 0555;
+
+		*prevp = new;
+		prevp = &new->child;
+
+		new += 2;
+	}
+	*prevp = table;
+	header->ctl_table_arg = table;
+
+	INIT_LIST_HEAD(&header->ctl_entry);
+	header->used = 0;
+	header->unregistering = NULL;
+	header->root = root;
+	sysctl_set_parent(NULL, header->ctl_table);
+	header->count = 1;
+#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
+	if (sysctl_check_table(namespaces, header->ctl_table)) {
+		kfree(header);
+		return NULL;
+	}
+#endif
+	spin_lock(&sysctl_lock);
+	header->set = lookup_header_set(root, namespaces);
+	header->attached_by = header->ctl_table;
+	header->attached_to = root_table;
+	header->parent = &root_table_header;
+	for (set = header->set; set; set = set->parent) {
+		struct ctl_table_header *p;
+		list_for_each_entry(p, &set->list, ctl_entry) {
+			if (p->unregistering)
+				continue;
+			try_attach(p, header);
+		}
+	}
+	header->parent->count++;
+	list_add_tail(&header->ctl_entry, &header->set->list);
+	spin_unlock(&sysctl_lock);
+
+	return header;
+}
+
+/**
+ * register_sysctl_table_path - register a sysctl table hierarchy
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See __register_sysctl_paths for more details.
+ */
+struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
+						struct ctl_table *table)
+{
+	return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
+					path, table);
+}
+EXPORT_SYMBOL(register_sysctl_paths);
+
+/**
+ * register_sysctl_table - register a sysctl table hierarchy
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See register_sysctl_paths for more details.
+ */
+struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
+{
+	static const struct ctl_path null_path[] = { {} };
+
+	return register_sysctl_paths(null_path, table);
+}
+EXPORT_SYMBOL(register_sysctl_table);
+
+/**
+ * unregister_sysctl_table - unregister a sysctl table hierarchy
+ * @header: the header returned from register_sysctl_table
+ *
+ * Unregisters the sysctl table and all children. proc entries may not
+ * actually be removed until they are no longer used by anyone.
+ */
+void unregister_sysctl_table(struct ctl_table_header * header)
+{
+	might_sleep();
+
+	if (header == NULL)
+		return;
+
+	spin_lock(&sysctl_lock);
+	start_unregistering(header);
+	if (!--header->parent->count) {
+		WARN_ON(1);
+		kfree_rcu(header->parent, rcu);
+	}
+	if (!--header->count)
+		kfree_rcu(header, rcu);
+	spin_unlock(&sysctl_lock);
+}
+EXPORT_SYMBOL(unregister_sysctl_table);
+
+void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_set *parent,
+	int (*is_seen)(struct ctl_table_set *))
+{
+	INIT_LIST_HEAD(&p->list);
+	p->parent = parent ? parent : &sysctl_table_root.default_set;
+	p->is_seen = is_seen;
+}
+
+
 int __init proc_sys_init(void)
 {
 	struct proc_dir_entry *proc_sys_root;
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 5e3532e..08cabbf 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1073,17 +1073,6 @@ extern void setup_sysctl_set(struct ctl_table_set *p,
 	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *));
 
-extern void sysctl_head_get(struct ctl_table_header *);
-extern void sysctl_head_put(struct ctl_table_header *);
-extern int sysctl_is_seen(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *);
-extern struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev);
-extern struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-						struct ctl_table_header *prev);
-extern void sysctl_head_finish(struct ctl_table_header *prev);
-extern int sysctl_perm(struct ctl_table_root *root,
-		struct ctl_table *table, int op);
-
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_root *root, struct nsproxy *namespaces,
@@ -1093,7 +1082,6 @@ struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 						struct ctl_table *table);
 
 void unregister_sysctl_table(struct ctl_table_header * table);
-int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table);
 
 extern int sysctl_init(void);
 #else /* CONFIG_SYSCTL */
@@ -1118,10 +1106,6 @@ static inline void setup_sysctl_set(struct ctl_table_set *p,
 {
 }
 
-static inline void sysctl_head_put(struct ctl_table_header *head)
-{
-}
-
 #endif /* CONFIG_SYSCTL */
 
 #endif /* __KERNEL__ */
diff --git a/kernel/Makefile b/kernel/Makefile
index 2d9de86..cb41b95 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -27,7 +27,6 @@ obj-y += power/
 
 obj-$(CONFIG_FREEZER) += freezer.o
 obj-$(CONFIG_PROFILING) += profile.o
-obj-$(CONFIG_SYSCTL_SYSCALL_CHECK) += sysctl_check.o
 obj-$(CONFIG_STACKTRACE) += stacktrace.o
 obj-y += time/
 obj-$(CONFIG_DEBUG_MUTEXES) += mutex-debug.o
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index ad46024..b774909 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -192,20 +192,6 @@ static int sysrq_sysctl_handler(ctl_table *table, int write,
 
 #endif
 
-static struct ctl_table root_table[1];
-static struct ctl_table_root sysctl_table_root;
-static struct ctl_table_header root_table_header = {
-	{{.count = 1,
-	.ctl_table = root_table,
-	.ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
-	.root = &sysctl_table_root,
-	.set = &sysctl_table_root.default_set,
-};
-static struct ctl_table_root sysctl_table_root = {
-	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
-};
-
 static struct ctl_table kern_table[];
 static struct ctl_table vm_table[];
 static struct ctl_table fs_table[];
@@ -1559,459 +1545,12 @@ static struct ctl_table dev_table[] = {
 	{ }
 };
 
-static DEFINE_SPINLOCK(sysctl_lock);
-
-/* called under sysctl_lock */
-static int use_table(struct ctl_table_header *p)
-{
-	if (unlikely(p->unregistering))
-		return 0;
-	p->used++;
-	return 1;
-}
-
-/* called under sysctl_lock */
-static void unuse_table(struct ctl_table_header *p)
-{
-	if (!--p->used)
-		if (unlikely(p->unregistering))
-			complete(p->unregistering);
-}
-
-/* called under sysctl_lock, will reacquire if has to wait */
-static void start_unregistering(struct ctl_table_header *p)
-{
-	/*
-	 * if p->used is 0, nobody will ever touch that entry again;
-	 * we'll eliminate all paths to it before dropping sysctl_lock
-	 */
-	if (unlikely(p->used)) {
-		struct completion wait;
-		init_completion(&wait);
-		p->unregistering = &wait;
-		spin_unlock(&sysctl_lock);
-		wait_for_completion(&wait);
-		spin_lock(&sysctl_lock);
-	} else {
-		/* anything non-NULL; we'll never dereference it */
-		p->unregistering = ERR_PTR(-EINVAL);
-	}
-	/*
-	 * do not remove from the list until nobody holds it; walking the
-	 * list in do_sysctl() relies on that.
-	 */
-	list_del_init(&p->ctl_entry);
-}
-
-void sysctl_head_get(struct ctl_table_header *head)
-{
-	spin_lock(&sysctl_lock);
-	head->count++;
-	spin_unlock(&sysctl_lock);
-}
-
-void sysctl_head_put(struct ctl_table_header *head)
-{
-	spin_lock(&sysctl_lock);
-	if (!--head->count)
-		kfree_rcu(head, rcu);
-	spin_unlock(&sysctl_lock);
-}
-
-struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
-{
-	if (!head)
-		BUG();
-	spin_lock(&sysctl_lock);
-	if (!use_table(head))
-		head = ERR_PTR(-ENOENT);
-	spin_unlock(&sysctl_lock);
-	return head;
-}
-
-void sysctl_head_finish(struct ctl_table_header *head)
-{
-	if (!head)
-		return;
-	spin_lock(&sysctl_lock);
-	unuse_table(head);
-	spin_unlock(&sysctl_lock);
-}
-
-static struct ctl_table_set *
-lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
-{
-	struct ctl_table_set *set = &root->default_set;
-	if (root->lookup)
-		set = root->lookup(root, namespaces);
-	return set;
-}
-
-static struct list_head *
-lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
-{
-	struct ctl_table_set *set = lookup_header_set(root, namespaces);
-	return &set->list;
-}
-
-struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-					    struct ctl_table_header *prev)
-{
-	struct ctl_table_root *root;
-	struct list_head *header_list;
-	struct ctl_table_header *head;
-	struct list_head *tmp;
-
-	spin_lock(&sysctl_lock);
-	if (prev) {
-		head = prev;
-		tmp = &prev->ctl_entry;
-		unuse_table(prev);
-		goto next;
-	}
-	tmp = &root_table_header.ctl_entry;
-	for (;;) {
-		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
-
-		if (!use_table(head))
-			goto next;
-		spin_unlock(&sysctl_lock);
-		return head;
-	next:
-		root = head->root;
-		tmp = tmp->next;
-		header_list = lookup_header_list(root, namespaces);
-		if (tmp != header_list)
-			continue;
-
-		do {
-			root = list_entry(root->root_list.next,
-					struct ctl_table_root, root_list);
-			if (root == &sysctl_table_root)
-				goto out;
-			header_list = lookup_header_list(root, namespaces);
-		} while (list_empty(header_list));
-		tmp = header_list->next;
-	}
-out:
-	spin_unlock(&sysctl_lock);
-	return NULL;
-}
-
-struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
-{
-	return __sysctl_head_next(current->nsproxy, prev);
-}
-
-void register_sysctl_root(struct ctl_table_root *root)
-{
-	spin_lock(&sysctl_lock);
-	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
-	spin_unlock(&sysctl_lock);
-}
-
-/*
- * sysctl_perm does NOT grant the superuser all rights automatically, because
- * some sysctl variables are readonly even to root.
- */
-
-static int test_perm(int mode, int op)
-{
-	if (!current_euid())
-		mode >>= 6;
-	else if (in_egroup_p(0))
-		mode >>= 3;
-	if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
-		return 0;
-	return -EACCES;
-}
-
-int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
-{
-	int mode;
-
-	if (root->permissions)
-		mode = root->permissions(root, current->nsproxy, table);
-	else
-		mode = table->mode;
-
-	return test_perm(mode, op);
-}
-
-static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
-{
-	for (; table->procname; table++) {
-		table->parent = parent;
-		if (table->child)
-			sysctl_set_parent(table, table->child);
-	}
-}
-
 int __init sysctl_init(void)
 {
 	register_sysctl_table(sysctl_base_table);
 	return 0;
 }
 
-static struct ctl_table *is_branch_in(struct ctl_table *branch,
-				      struct ctl_table *table)
-{
-	struct ctl_table *p;
-	const char *s = branch->procname;
-
-	/* branch should have named subdirectory as its first element */
-	if (!s || !branch->child)
-		return NULL;
-
-	/* ... and nothing else */
-	if (branch[1].procname)
-		return NULL;
-
-	/* table should contain subdirectory with the same name */
-	for (p = table; p->procname; p++) {
-		if (!p->child)
-			continue;
-		if (p->procname && strcmp(p->procname, s) == 0)
-			return p;
-	}
-	return NULL;
-}
-
-/* see if attaching q to p would be an improvement */
-static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
-{
-	struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
-	struct ctl_table *next;
-	int is_better = 0;
-	int not_in_parent = !p->attached_by;
-
-	while ((next = is_branch_in(by, to)) != NULL) {
-		if (by == q->attached_by)
-			is_better = 1;
-		if (to == p->attached_by)
-			not_in_parent = 1;
-		by = by->child;
-		to = next->child;
-	}
-
-	if (is_better && not_in_parent) {
-		q->attached_by = by;
-		q->attached_to = to;
-		q->parent = p;
-	}
-}
-
-/**
- * __register_sysctl_paths - register a sysctl hierarchy
- * @root: List of sysctl headers to register on
- * @namespaces: Data to compute which lists of sysctl entries are visible
- * @path: The path to the directory the sysctl table is in.
- * @table: the top-level table structure
- *
- * Register a sysctl table hierarchy. @table should be a filled in ctl_table
- * array. A completely 0 filled entry terminates the table.
- *
- * The members of the &struct ctl_table structure are used as follows:
- *
- * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
- *            enter a sysctl file
- *
- * data - a pointer to data for use by proc_handler
- *
- * maxlen - the maximum size in bytes of the data
- *
- * mode - the file permissions for the /proc/sys file, and for sysctl(2)
- *
- * child - a pointer to the child sysctl table if this entry is a directory, or
- *         %NULL.
- *
- * proc_handler - the text handler routine (described below)
- *
- * de - for internal use by the sysctl routines
- *
- * extra1, extra2 - extra pointers usable by the proc handler routines
- *
- * Leaf nodes in the sysctl tree will be represented by a single file
- * under /proc; non-leaf nodes will be represented by directories.
- *
- * sysctl(2) can automatically manage read and write requests through
- * the sysctl table.  The data and maxlen fields of the ctl_table
- * struct enable minimal validation of the values being written to be
- * performed, and the mode field allows minimal authentication.
- *
- * There must be a proc_handler routine for any terminal nodes
- * mirrored under /proc/sys (non-terminals are handled by a built-in
- * directory handler).  Several default handlers are available to
- * cover common cases -
- *
- * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
- * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), 
- * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
- *
- * It is the handler's job to read the input buffer from user memory
- * and process it. The handler should return 0 on success.
- *
- * This routine returns %NULL on a failure to register, and a pointer
- * to the table header on success.
- */
-struct ctl_table_header *__register_sysctl_paths(
-	struct ctl_table_root *root,
-	struct nsproxy *namespaces,
-	const struct ctl_path *path, struct ctl_table *table)
-{
-	struct ctl_table_header *header;
-	struct ctl_table *new, **prevp;
-	unsigned int n, npath;
-	struct ctl_table_set *set;
-
-	/* Count the path components */
-	for (npath = 0; path[npath].procname; ++npath)
-		;
-
-	/*
-	 * For each path component, allocate a 2-element ctl_table array.
-	 * The first array element will be filled with the sysctl entry
-	 * for this, the second will be the sentinel (procname == 0).
-	 *
-	 * We allocate everything in one go so that we don't have to
-	 * worry about freeing additional memory in unregister_sysctl_table.
-	 */
-	header = kzalloc(sizeof(struct ctl_table_header) +
-			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
-	if (!header)
-		return NULL;
-
-	new = (struct ctl_table *) (header + 1);
-
-	/* Now connect the dots */
-	prevp = &header->ctl_table;
-	for (n = 0; n < npath; ++n, ++path) {
-		/* Copy the procname */
-		new->procname = path->procname;
-		new->mode     = 0555;
-
-		*prevp = new;
-		prevp = &new->child;
-
-		new += 2;
-	}
-	*prevp = table;
-	header->ctl_table_arg = table;
-
-	INIT_LIST_HEAD(&header->ctl_entry);
-	header->used = 0;
-	header->unregistering = NULL;
-	header->root = root;
-	sysctl_set_parent(NULL, header->ctl_table);
-	header->count = 1;
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-	if (sysctl_check_table(namespaces, header->ctl_table)) {
-		kfree(header);
-		return NULL;
-	}
-#endif
-	spin_lock(&sysctl_lock);
-	header->set = lookup_header_set(root, namespaces);
-	header->attached_by = header->ctl_table;
-	header->attached_to = root_table;
-	header->parent = &root_table_header;
-	for (set = header->set; set; set = set->parent) {
-		struct ctl_table_header *p;
-		list_for_each_entry(p, &set->list, ctl_entry) {
-			if (p->unregistering)
-				continue;
-			try_attach(p, header);
-		}
-	}
-	header->parent->count++;
-	list_add_tail(&header->ctl_entry, &header->set->list);
-	spin_unlock(&sysctl_lock);
-
-	return header;
-}
-
-/**
- * register_sysctl_table_path - register a sysctl table hierarchy
- * @path: The path to the directory the sysctl table is in.
- * @table: the top-level table structure
- *
- * Register a sysctl table hierarchy. @table should be a filled in ctl_table
- * array. A completely 0 filled entry terminates the table.
- *
- * See __register_sysctl_paths for more details.
- */
-struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
-						struct ctl_table *table)
-{
-	return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
-					path, table);
-}
-
-/**
- * register_sysctl_table - register a sysctl table hierarchy
- * @table: the top-level table structure
- *
- * Register a sysctl table hierarchy. @table should be a filled in ctl_table
- * array. A completely 0 filled entry terminates the table.
- *
- * See register_sysctl_paths for more details.
- */
-struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
-{
-	static const struct ctl_path null_path[] = { {} };
-
-	return register_sysctl_paths(null_path, table);
-}
-
-/**
- * unregister_sysctl_table - unregister a sysctl table hierarchy
- * @header: the header returned from register_sysctl_table
- *
- * Unregisters the sysctl table and all children. proc entries may not
- * actually be removed until they are no longer used by anyone.
- */
-void unregister_sysctl_table(struct ctl_table_header * header)
-{
-	might_sleep();
-
-	if (header == NULL)
-		return;
-
-	spin_lock(&sysctl_lock);
-	start_unregistering(header);
-	if (!--header->parent->count) {
-		WARN_ON(1);
-		kfree_rcu(header->parent, rcu);
-	}
-	if (!--header->count)
-		kfree_rcu(header, rcu);
-	spin_unlock(&sysctl_lock);
-}
-
-int sysctl_is_seen(struct ctl_table_header *p)
-{
-	struct ctl_table_set *set = p->set;
-	int res;
-	spin_lock(&sysctl_lock);
-	if (p->unregistering)
-		res = 0;
-	else if (!set->is_seen)
-		res = 1;
-	else
-		res = set->is_seen(set);
-	spin_unlock(&sysctl_lock);
-	return res;
-}
-
-void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
-	int (*is_seen)(struct ctl_table_set *))
-{
-	INIT_LIST_HEAD(&p->list);
-	p->parent = parent ? parent : &sysctl_table_root.default_set;
-	p->is_seen = is_seen;
-}
-
 #endif /* CONFIG_SYSCTL */
 
 /*
@@ -2977,6 +2516,3 @@ EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
 EXPORT_SYMBOL(proc_dostring);
 EXPORT_SYMBOL(proc_doulongvec_minmax);
 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
-EXPORT_SYMBOL(register_sysctl_table);
-EXPORT_SYMBOL(register_sysctl_paths);
-EXPORT_SYMBOL(unregister_sysctl_table);
diff --git a/kernel/sysctl_check.c b/kernel/sysctl_check.c
deleted file mode 100644
index 362da65..0000000
--- a/kernel/sysctl_check.c
+++ /dev/null
@@ -1,160 +0,0 @@
-#include <linux/stat.h>
-#include <linux/sysctl.h>
-#include "../fs/xfs/xfs_sysctl.h"
-#include <linux/sunrpc/debug.h>
-#include <linux/string.h>
-#include <net/ip_vs.h>
-
-
-static int sysctl_depth(struct ctl_table *table)
-{
-	struct ctl_table *tmp;
-	int depth;
-
-	depth = 0;
-	for (tmp = table; tmp->parent; tmp = tmp->parent)
-		depth++;
-
-	return depth;
-}
-
-static struct ctl_table *sysctl_parent(struct ctl_table *table, int n)
-{
-	int i;
-
-	for (i = 0; table && i < n; i++)
-		table = table->parent;
-
-	return table;
-}
-
-
-static void sysctl_print_path(struct ctl_table *table)
-{
-	struct ctl_table *tmp;
-	int depth, i;
-	depth = sysctl_depth(table);
-	if (table->procname) {
-		for (i = depth; i >= 0; i--) {
-			tmp = sysctl_parent(table, i);
-			printk("/%s", tmp->procname?tmp->procname:"");
-		}
-	}
-	printk(" ");
-}
-
-static struct ctl_table *sysctl_check_lookup(struct nsproxy *namespaces,
-						struct ctl_table *table)
-{
-	struct ctl_table_header *head;
-	struct ctl_table *ref, *test;
-	int depth, cur_depth;
-
-	depth = sysctl_depth(table);
-
-	for (head = __sysctl_head_next(namespaces, NULL); head;
-	     head = __sysctl_head_next(namespaces, head)) {
-		cur_depth = depth;
-		ref = head->ctl_table;
-repeat:
-		test = sysctl_parent(table, cur_depth);
-		for (; ref->procname; ref++) {
-			int match = 0;
-			if (cur_depth && !ref->child)
-				continue;
-
-			if (test->procname && ref->procname &&
-			    (strcmp(test->procname, ref->procname) == 0))
-					match++;
-
-			if (match) {
-				if (cur_depth != 0) {
-					cur_depth--;
-					ref = ref->child;
-					goto repeat;
-				}
-				goto out;
-			}
-		}
-	}
-	ref = NULL;
-out:
-	sysctl_head_finish(head);
-	return ref;
-}
-
-static void set_fail(const char **fail, struct ctl_table *table, const char *str)
-{
-	if (*fail) {
-		printk(KERN_ERR "sysctl table check failed: ");
-		sysctl_print_path(table);
-		printk(" %s\n", *fail);
-		dump_stack();
-	}
-	*fail = str;
-}
-
-static void sysctl_check_leaf(struct nsproxy *namespaces,
-				struct ctl_table *table, const char **fail)
-{
-	struct ctl_table *ref;
-
-	ref = sysctl_check_lookup(namespaces, table);
-	if (ref && (ref != table))
-		set_fail(fail, table, "Sysctl already exists");
-}
-
-int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
-{
-	int error = 0;
-	for (; table->procname; table++) {
-		const char *fail = NULL;
-
-		if (table->parent) {
-			if (!table->parent->procname)
-				set_fail(&fail, table, "Parent without procname");
-		}
-		if (table->child) {
-			if (table->data)
-				set_fail(&fail, table, "Directory with data?");
-			if (table->maxlen)
-				set_fail(&fail, table, "Directory with maxlen?");
-			if ((table->mode & (S_IRUGO|S_IXUGO)) != table->mode)
-				set_fail(&fail, table, "Writable sysctl directory");
-			if (table->proc_handler)
-				set_fail(&fail, table, "Directory with proc_handler");
-			if (table->extra1)
-				set_fail(&fail, table, "Directory with extra1");
-			if (table->extra2)
-				set_fail(&fail, table, "Directory with extra2");
-		} else {
-			if ((table->proc_handler == proc_dostring) ||
-			    (table->proc_handler == proc_dointvec) ||
-			    (table->proc_handler == proc_dointvec_minmax) ||
-			    (table->proc_handler == proc_dointvec_jiffies) ||
-			    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
-			    (table->proc_handler == proc_dointvec_ms_jiffies) ||
-			    (table->proc_handler == proc_doulongvec_minmax) ||
-			    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
-				if (!table->data)
-					set_fail(&fail, table, "No data");
-				if (!table->maxlen)
-					set_fail(&fail, table, "No maxlen");
-			}
-#ifdef CONFIG_PROC_SYSCTL
-			if (!table->proc_handler)
-				set_fail(&fail, table, "No proc_handler");
-#endif
-			sysctl_check_leaf(namespaces, table, &fail);
-		}
-		if (table->mode > 0777)
-			set_fail(&fail, table, "bogus .mode");
-		if (fail) {
-			set_fail(&fail, table, NULL);
-			error = -EINVAL;
-		}
-		if (table->child)
-			error |= sysctl_check_table(namespaces, table->child);
-	}
-	return error;
-}
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 05/29] sysctl: Make the directories have nlink == 1
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (32 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 06/29] sysctl: Implement retire_sysctl_set Eric W. Biederman
                   ` (24 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

I goofed when I made sysctl directories have nlink == 0.
nlink == 0 means the directory has been deleted.
nlink == 1 meands a directory does not count subdirectories.

Use the default nlink == 1 for sysctl directories.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 06e6f10..f6aa751 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -253,7 +253,6 @@ static struct inode *proc_sys_make_inode(struct super_block *sb,
 		inode->i_fop = &proc_sys_file_operations;
 	} else {
 		inode->i_mode |= S_IFDIR;
-		clear_nlink(inode);
 		inode->i_op = &proc_sys_dir_operations;
 		inode->i_fop = &proc_sys_dir_file_operations;
 	}
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 06/29] sysctl: Implement retire_sysctl_set
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (33 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 05/29] sysctl: Make the directories have nlink == 1 Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept Eric W. Biederman
                   ` (23 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

This adds a small helper retire_sysctl_set to remove the intimate knowledge about
the how a sysctl_set is implemented from net/sysct_net.c

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |    4 ++++
 include/linux/sysctl.h |    1 +
 net/sysctl_net.c       |    2 +-
 3 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index f6aa751..9d8223c 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1080,6 +1080,10 @@ void setup_sysctl_set(struct ctl_table_set *p,
 	p->is_seen = is_seen;
 }
 
+void retire_sysctl_set(struct ctl_table_set *set)
+{
+	WARN_ON(!list_empty(&set->list));
+}
 
 int __init proc_sys_init(void)
 {
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 08cabbf..475ff0e 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1072,6 +1072,7 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll);
 extern void setup_sysctl_set(struct ctl_table_set *p,
 	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *));
+extern void retire_sysctl_set(struct ctl_table_set *set);
 
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_paths(
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index a6bbee2..ffd67a6 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -82,7 +82,7 @@ static int __net_init sysctl_net_init(struct net *net)
 
 static void __net_exit sysctl_net_exit(struct net *net)
 {
-	WARN_ON(!list_empty(&net->sysctls.list));
+	retire_sysctl_set(&net->sysctls);
 }
 
 static struct pernet_operations sysctl_pernet_ops = {
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (34 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 06/29] sysctl: Implement retire_sysctl_set Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 08/29] sysctl: Create local copies of directory names used in paths Eric W. Biederman
                   ` (22 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

In sysctl_net register the two networking roots in the proper order.

In register_sysctl walk the sysctl sets in the reverse order of the
sysctl roots.

Remove parent from ctl_table_set and setup_sysctl_set as it is no
longer needed.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   11 ++++++++---
 include/linux/sysctl.h |    3 ---
 net/sysctl_net.c       |    5 ++---
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 9d8223c..86d32a3 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -995,13 +995,20 @@ struct ctl_table_header *__register_sysctl_paths(
 	header->attached_by = header->ctl_table;
 	header->attached_to = root_table;
 	header->parent = &root_table_header;
-	for (set = header->set; set; set = set->parent) {
+	set = header->set;
+	root = header->root;
+	for (;;) {
 		struct ctl_table_header *p;
 		list_for_each_entry(p, &set->list, ctl_entry) {
 			if (p->unregistering)
 				continue;
 			try_attach(p, header);
 		}
+		if (root == &sysctl_table_root)
+			break;
+		root = list_entry(root->root_list.prev,
+				  struct ctl_table_root, root_list);
+		set = lookup_header_set(root, namespaces);
 	}
 	header->parent->count++;
 	list_add_tail(&header->ctl_entry, &header->set->list);
@@ -1072,11 +1079,9 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 EXPORT_SYMBOL(unregister_sysctl_table);
 
 void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *))
 {
 	INIT_LIST_HEAD(&p->list);
-	p->parent = parent ? parent : &sysctl_table_root.default_set;
 	p->is_seen = is_seen;
 }
 
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 475ff0e..43c36ac 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1047,7 +1047,6 @@ struct ctl_table_header
 
 struct ctl_table_set {
 	struct list_head list;
-	struct ctl_table_set *parent;
 	int (*is_seen)(struct ctl_table_set *);
 };
 
@@ -1070,7 +1069,6 @@ struct ctl_path {
 void proc_sys_poll_notify(struct ctl_table_poll *poll);
 
 extern void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *));
 extern void retire_sysctl_set(struct ctl_table_set *set);
 
@@ -1102,7 +1100,6 @@ static inline void unregister_sysctl_table(struct ctl_table_header * table)
 }
 
 static inline void setup_sysctl_set(struct ctl_table_set *p,
-	struct ctl_table_set *parent,
 	int (*is_seen)(struct ctl_table_set *))
 {
 }
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index ffd67a6..07c6b87 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -75,7 +75,6 @@ static struct ctl_table_root net_sysctl_ro_root = {
 static int __net_init sysctl_net_init(struct net *net)
 {
 	setup_sysctl_set(&net->sysctls,
-			 &net_sysctl_ro_root.default_set,
 			 is_seen);
 	return 0;
 }
@@ -96,9 +95,9 @@ static __init int net_sysctl_init(void)
 	ret = register_pernet_subsys(&sysctl_pernet_ops);
 	if (ret)
 		goto out;
-	register_sysctl_root(&net_sysctl_root);
-	setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL, NULL);
+	setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL);
 	register_sysctl_root(&net_sysctl_ro_root);
+	register_sysctl_root(&net_sysctl_root);
 out:
 	return ret;
 }
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 08/29] sysctl: Create local copies of directory names used in paths
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (35 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path Eric W. Biederman
                   ` (21 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Creating local copies of directory names is a good idea for
two reasons.
- The dynamic names used by callers must be copied into new
  strings by the callers today to ensure the strings do not
  change between register and unregister of the sysctl table.

- Sysctl directories have a potentially different lifetime
  than the time between register and unregister of any
  particular sysctl table.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   11 ++++++++---
 1 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 86d32a3..bcf60fb 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -943,10 +943,12 @@ struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table *new, **prevp;
 	unsigned int n, npath;
 	struct ctl_table_set *set;
+	size_t path_bytes = 0;
+	char *new_name;
 
 	/* Count the path components */
 	for (npath = 0; path[npath].procname; ++npath)
-		;
+		path_bytes += strlen(path[npath].procname) + 1;
 
 	/*
 	 * For each path component, allocate a 2-element ctl_table array.
@@ -956,24 +958,27 @@ struct ctl_table_header *__register_sysctl_paths(
 	 * We allocate everything in one go so that we don't have to
 	 * worry about freeing additional memory in unregister_sysctl_table.
 	 */
-	header = kzalloc(sizeof(struct ctl_table_header) +
+	header = kzalloc(sizeof(struct ctl_table_header) + path_bytes +
 			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
 	if (!header)
 		return NULL;
 
 	new = (struct ctl_table *) (header + 1);
+	new_name = (char *)(new + (2 * npath));
 
 	/* Now connect the dots */
 	prevp = &header->ctl_table;
 	for (n = 0; n < npath; ++n, ++path) {
 		/* Copy the procname */
-		new->procname = path->procname;
+		strcpy(new_name, path->procname);
+		new->procname = new_name;
 		new->mode     = 0555;
 
 		*prevp = new;
 		prevp = &new->child;
 
 		new += 2;
+		new_name += strlen(new_name) + 1;
 	}
 	*prevp = table;
 	header->ctl_table_arg = table;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (36 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 08/29] sysctl: Create local copies of directory names used in paths Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths Eric W. Biederman
                   ` (20 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Make __register_sysctl_table the core sysctl registration operation and
make it take a char * string as path.

Now that binary paths have been banished into the real of backwards
compatibility in kernel/binary_sysctl.c where they can be safely
ignored there is no longer a need to use struct ctl_path to represent
path names when registering ctl_tables.

Start the transition to using normal char * strings to represent
pathnames when registering sysctl tables.  Normal strings are easier
to deal with both in the internal sysctl implementation and for
programmers registering sysctl tables.

__register_sysctl_paths is turned into a backwards compatibility wrapper
that converts a ctl_path array into a normal char * string.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   94 ++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/sysctl.h |    3 ++
 2 files changed, 87 insertions(+), 10 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index bcf60fb..5704ff0 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -882,7 +882,7 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
 #endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
 
 /**
- * __register_sysctl_paths - register a sysctl hierarchy
+ * __register_sysctl_table - register a sysctl table
  * @root: List of sysctl headers to register on
  * @namespaces: Data to compute which lists of sysctl entries are visible
  * @path: The path to the directory the sysctl table is in.
@@ -934,21 +934,34 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
  * This routine returns %NULL on a failure to register, and a pointer
  * to the table header on success.
  */
-struct ctl_table_header *__register_sysctl_paths(
+struct ctl_table_header *__register_sysctl_table(
 	struct ctl_table_root *root,
 	struct nsproxy *namespaces,
-	const struct ctl_path *path, struct ctl_table *table)
+	const char *path, struct ctl_table *table)
 {
 	struct ctl_table_header *header;
 	struct ctl_table *new, **prevp;
-	unsigned int n, npath;
+	const char *name, *nextname;
+	unsigned int npath = 0;
 	struct ctl_table_set *set;
 	size_t path_bytes = 0;
 	char *new_name;
 
 	/* Count the path components */
-	for (npath = 0; path[npath].procname; ++npath)
-		path_bytes += strlen(path[npath].procname) + 1;
+	for (name = path; name; name = nextname) {
+		int namelen;
+		nextname = strchr(name, '/');
+		if (nextname) {
+			namelen = nextname - name;
+			nextname++;
+		} else {
+			namelen = strlen(name);
+		}
+		if (namelen == 0)
+			continue;
+		path_bytes += namelen + 1;
+		npath++;
+	}
 
 	/*
 	 * For each path component, allocate a 2-element ctl_table array.
@@ -968,9 +981,20 @@ struct ctl_table_header *__register_sysctl_paths(
 
 	/* Now connect the dots */
 	prevp = &header->ctl_table;
-	for (n = 0; n < npath; ++n, ++path) {
-		/* Copy the procname */
-		strcpy(new_name, path->procname);
+	for (name = path; name; name = nextname) {
+		int namelen;
+		nextname = strchr(name, '/');
+		if (nextname) {
+			namelen = nextname - name;
+			nextname++;
+		} else {
+			namelen = strlen(name);
+		}
+		if (namelen == 0)
+			continue;
+		memcpy(new_name, name, namelen);
+		new_name[namelen] = '\0';
+
 		new->procname = new_name;
 		new->mode     = 0555;
 
@@ -978,7 +1002,7 @@ struct ctl_table_header *__register_sysctl_paths(
 		prevp = &new->child;
 
 		new += 2;
-		new_name += strlen(new_name) + 1;
+		new_name += namelen + 1;
 	}
 	*prevp = table;
 	header->ctl_table_arg = table;
@@ -1022,6 +1046,56 @@ struct ctl_table_header *__register_sysctl_paths(
 	return header;
 }
 
+static char *append_path(const char *path, char *pos, const char *name)
+{
+	int namelen;
+	namelen = strlen(name);
+	if (((pos - path) + namelen + 2) >= PATH_MAX)
+		return NULL;
+	memcpy(pos, name, namelen);
+	pos[namelen] = '/';
+	pos[namelen + 1] = '\0';
+	pos += namelen + 1;
+	return pos;
+}
+
+/**
+ * __register_sysctl_paths - register a sysctl table hierarchy
+ * @root: List of sysctl headers to register on
+ * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the top-level table structure
+ *
+ * Register a sysctl table hierarchy. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See __register_sysctl_table for more details.
+ */
+struct ctl_table_header *__register_sysctl_paths(
+	struct ctl_table_root *root,
+	struct nsproxy *namespaces,
+	const struct ctl_path *path, struct ctl_table *table)
+{
+	struct ctl_table_header *header = NULL;
+	const struct ctl_path *component;
+	char *new_path, *pos;
+
+	pos = new_path = kmalloc(PATH_MAX, GFP_KERNEL);
+	if (!new_path)
+		return NULL;
+
+	pos[0] = '\0';
+	for (component = path; component->procname; component++) {
+		pos = append_path(new_path, pos, component->procname);
+		if (!pos)
+			goto out;
+	}
+	header = __register_sysctl_table(root, namespaces, new_path, table);
+out:
+	kfree(new_path);
+	return header;
+}
+
 /**
  * register_sysctl_table_path - register a sysctl table hierarchy
  * @path: The path to the directory the sysctl table is in.
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 43c36ac..a514e0f 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1073,6 +1073,9 @@ extern void setup_sysctl_set(struct ctl_table_set *p,
 extern void retire_sysctl_set(struct ctl_table_set *set);
 
 void register_sysctl_root(struct ctl_table_root *root);
+struct ctl_table_header *__register_sysctl_table(
+	struct ctl_table_root *root, struct nsproxy *namespaces,
+	const char *path, struct ctl_table *table);
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_root *root, struct nsproxy *namespaces,
 	const struct ctl_path *path, struct ctl_table *table);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (37 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 11/29] sysctl: register only tables of sysctl files Eric W. Biederman
                   ` (19 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

For any component of table passed to __register_sysctl_paths
that actually serves as a path, add that to the cstring path
that is passed to __register_sysctl_table.

The result is that for most calls to __register_sysctl_paths
we only pass a table to __register_sysctl_table that contains
no child directories.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 5704ff0..9b91dee 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1076,6 +1076,7 @@ struct ctl_table_header *__register_sysctl_paths(
 	struct nsproxy *namespaces,
 	const struct ctl_path *path, struct ctl_table *table)
 {
+	struct ctl_table *ctl_table_arg = table;
 	struct ctl_table_header *header = NULL;
 	const struct ctl_path *component;
 	char *new_path, *pos;
@@ -1090,7 +1091,15 @@ struct ctl_table_header *__register_sysctl_paths(
 		if (!pos)
 			goto out;
 	}
+	while (table->procname && table->child && !table[1].procname) {
+		pos = append_path(new_path, pos, table->procname);
+		if (!pos)
+			goto out;
+		table = table->child;
+	}
 	header = __register_sysctl_table(root, namespaces, new_path, table);
+	if (header)
+		header->ctl_table_arg = ctl_table_arg;
 out:
 	kfree(new_path);
 	return header;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 11/29] sysctl: register only tables of sysctl files
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (38 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 12/29] sysctl: Improve the sysctl sanity checks Eric W. Biederman
                   ` (18 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Split the registration of a complex ctl_table array which may have
arbitrary numbers of directories (->child != NULL) and tables of files
into a series of simpler registrations that only register tables of files.

Graphically:

   register('dir', { + file-a
                     + file-b
                     + subdir1
                       + file-c
                     + subdir2
                       + file-d
                       + file-e })

is transformed into:
   wrapper->subheaders[0] = register('dir', {file1-a, file1-b})
   wrapper->subheaders[1] = register('dir/subdir1', {file-c})
   wrapper->subheaders[2] = register('dir/subdir2', {file-d, file-e})
   return wrapper

This guarantees that __register_sysctl_table will only see a simple
ctl_table array with all entries having (->child == NULL).

Care was taken to pass the original simple ctl_table arrays to
__register_sysctl_table whenever possible.

This change is derived from a similar patch written
by Lucrian Grijincu.

Inspired-by: Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  165 ++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/sysctl.h |    2 +-
 2 files changed, 148 insertions(+), 19 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 9b91dee..6bab2ae 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -882,7 +882,7 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
 #endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
 
 /**
- * __register_sysctl_table - register a sysctl table
+ * __register_sysctl_table - register a leaf sysctl table
  * @root: List of sysctl headers to register on
  * @namespaces: Data to compute which lists of sysctl entries are visible
  * @path: The path to the directory the sysctl table is in.
@@ -900,29 +900,19 @@ static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *tabl
  *
  * maxlen - the maximum size in bytes of the data
  *
- * mode - the file permissions for the /proc/sys file, and for sysctl(2)
+ * mode - the file permissions for the /proc/sys file
  *
- * child - a pointer to the child sysctl table if this entry is a directory, or
- *         %NULL.
+ * child - must be %NULL.
  *
  * proc_handler - the text handler routine (described below)
  *
- * de - for internal use by the sysctl routines
- *
  * extra1, extra2 - extra pointers usable by the proc handler routines
  *
  * Leaf nodes in the sysctl tree will be represented by a single file
  * under /proc; non-leaf nodes will be represented by directories.
  *
- * sysctl(2) can automatically manage read and write requests through
- * the sysctl table.  The data and maxlen fields of the ctl_table
- * struct enable minimal validation of the values being written to be
- * performed, and the mode field allows minimal authentication.
- *
- * There must be a proc_handler routine for any terminal nodes
- * mirrored under /proc/sys (non-terminals are handled by a built-in
- * directory handler).  Several default handlers are available to
- * cover common cases -
+ * There must be a proc_handler routine for any terminal nodes.
+ * Several default handlers are available to cover common cases -
  *
  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
@@ -1059,6 +1049,100 @@ static char *append_path(const char *path, char *pos, const char *name)
 	return pos;
 }
 
+static int count_subheaders(struct ctl_table *table)
+{
+	int has_files = 0;
+	int nr_subheaders = 0;
+	struct ctl_table *entry;
+
+	/* special case: no directory and empty directory */
+	if (!table || !table->procname)
+		return 1;
+
+	for (entry = table; entry->procname; entry++) {
+		if (entry->child)
+			nr_subheaders += count_subheaders(entry->child);
+		else
+			has_files = 1;
+	}
+	return nr_subheaders + has_files;
+}
+
+static int register_leaf_sysctl_tables(const char *path, char *pos,
+	struct ctl_table_header ***subheader,
+	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table *table)
+{
+	struct ctl_table *ctl_table_arg = NULL;
+	struct ctl_table *entry, *files;
+	int nr_files = 0;
+	int nr_dirs = 0;
+	int err = -ENOMEM;
+
+	for (entry = table; entry->procname; entry++) {
+		if (entry->child)
+			nr_dirs++;
+		else
+			nr_files++;
+	}
+
+	files = table;
+	/* If there are mixed files and directories we need a new table */
+	if (nr_dirs && nr_files) {
+		struct ctl_table *new;
+		files = kzalloc(sizeof(struct ctl_table) * (nr_files + 1),
+				GFP_KERNEL);
+		if (!files)
+			goto out;
+
+		ctl_table_arg = files;
+		for (new = files, entry = table; entry->procname; entry++) {
+			if (entry->child)
+				continue;
+			*new = *entry;
+			new++;
+		}
+	}
+
+	/* Register everything except a directory full of subdirectories */
+	if (nr_files || !nr_dirs) {
+		struct ctl_table_header *header;
+		header = __register_sysctl_table(root, namespaces, path, files);
+		if (!header) {
+			kfree(ctl_table_arg);
+			goto out;
+		}
+
+		/* Remember if we need to free the file table */
+		header->ctl_table_arg = ctl_table_arg;
+		**subheader = header;
+		(*subheader)++;
+	}
+
+	/* Recurse into the subdirectories. */
+	for (entry = table; entry->procname; entry++) {
+		char *child_pos;
+
+		if (!entry->child)
+			continue;
+
+		err = -ENAMETOOLONG;
+		child_pos = append_path(path, pos, entry->procname);
+		if (!child_pos)
+			goto out;
+
+		err = register_leaf_sysctl_tables(path, child_pos, subheader,
+						  root, namespaces, entry->child);
+		pos[0] = '\0';
+		if (err)
+			goto out;
+	}
+	err = 0;
+out:
+	/* On failure our caller will unregister all registered subheaders */
+	return err;
+}
+
 /**
  * __register_sysctl_paths - register a sysctl table hierarchy
  * @root: List of sysctl headers to register on
@@ -1077,7 +1161,8 @@ struct ctl_table_header *__register_sysctl_paths(
 	const struct ctl_path *path, struct ctl_table *table)
 {
 	struct ctl_table *ctl_table_arg = table;
-	struct ctl_table_header *header = NULL;
+	int nr_subheaders = count_subheaders(table);
+	struct ctl_table_header *header = NULL, **subheaders, **subheader;
 	const struct ctl_path *component;
 	char *new_path, *pos;
 
@@ -1097,12 +1182,39 @@ struct ctl_table_header *__register_sysctl_paths(
 			goto out;
 		table = table->child;
 	}
-	header = __register_sysctl_table(root, namespaces, new_path, table);
-	if (header)
+	if (nr_subheaders == 1) {
+		header = __register_sysctl_table(root, namespaces, new_path, table);
+		if (header)
+			header->ctl_table_arg = ctl_table_arg;
+	} else {
+		header = kzalloc(sizeof(*header) +
+				 sizeof(*subheaders)*nr_subheaders, GFP_KERNEL);
+		if (!header)
+			goto out;
+
+		subheaders = (struct ctl_table_header **) (header + 1);
+		subheader = subheaders;
 		header->ctl_table_arg = ctl_table_arg;
+
+		if (register_leaf_sysctl_tables(new_path, pos, &subheader,
+						root, namespaces, table))
+			goto err_register_leaves;
+	}
+
 out:
 	kfree(new_path);
 	return header;
+
+err_register_leaves:
+	while (subheader > subheaders) {
+		struct ctl_table_header *subh = *(--subheader);
+		struct ctl_table *table = subh->ctl_table_arg;
+		unregister_sysctl_table(subh);
+		kfree(table);
+	}
+	kfree(header);
+	header = NULL;
+	goto out;
 }
 
 /**
@@ -1149,11 +1261,28 @@ EXPORT_SYMBOL(register_sysctl_table);
  */
 void unregister_sysctl_table(struct ctl_table_header * header)
 {
+	int nr_subheaders;
 	might_sleep();
 
 	if (header == NULL)
 		return;
 
+	nr_subheaders = count_subheaders(header->ctl_table_arg);
+	if (unlikely(nr_subheaders > 1)) {
+		struct ctl_table_header **subheaders;
+		int i;
+
+		subheaders = (struct ctl_table_header **)(header + 1);
+		for (i = nr_subheaders -1; i >= 0; i--) {
+			struct ctl_table_header *subh = subheaders[i];
+			struct ctl_table *table = subh->ctl_table_arg;
+			unregister_sysctl_table(subh);
+			kfree(table);
+		}
+		kfree(header);
+		return;
+	}
+
 	spin_lock(&sysctl_lock);
 	start_unregistering(header);
 	if (!--header->parent->count) {
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index a514e0f..25c7dfe 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1015,7 +1015,7 @@ struct ctl_table
 	void *data;
 	int maxlen;
 	umode_t mode;
-	struct ctl_table *child;
+	struct ctl_table *child;	/* Deprecated */
 	struct ctl_table *parent;	/* Automatically set */
 	proc_handler *proc_handler;	/* Callback for text formatting */
 	struct ctl_table_poll *poll;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 12/29] sysctl: Improve the sysctl sanity checks
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (39 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 11/29] sysctl: register only tables of sysctl files Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field Eric W. Biederman
                   ` (17 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

- Stop validating subdirectories now that we only register leaf tables

- Cleanup and improve the duplicate filename check.
  * Run the duplicate filename check under the sysctl_lock to guarantee
    we never add duplicate names.
  * Reduce the duplicate filename check to nearly O(M*N) where M is the
    number of entries in tthe table we are registering and N is the
    number of entries in the directory before we got there.

- Move the duplicate filename check into it's own function and call
  it directtly from __register_sysctl_table

- Kill the config option as the sanity checks are now cheap enough
  the config option is unnecessary. The original reason for the config
  option was because we had a huge table used to verify the proc filename
  to binary sysctl mapping.  That table has now evolved into the binary_sysctl
  translation layer and is no longer part of the sysctl_check code.

- Tighten up the permission checks.  Guarnateeing that files only have read
  or write permissions.

- Removed redudant check for parents having a procname as now everything has
  a procname.

- Generalize the backtrace logic so that we print a backtrace from
  any failure of __register_sysctl_table that was not caused by
  a memmory allocation failure.  The backtrace allows us to track
  down who erroneously registered a sysctl table.

Bechmark before (CONFIG_SYSCTL_CHECK=y):
    make-dummies 0 999 -> 12s
    rmmod dummy        -> 0.08s

Bechmark before (CONFIG_SYSCTL_CHECK=n):
    make-dummies 0 999 -> 0.7s
    rmmod dummy        -> 0.06s
    make-dummies 0 99999 -> 1m13s
    rmmod dummy          -> 0.38s

Benchmark after:
    make-dummies 0 999 -> 0.65s
    rmmod dummy        -> 0.055s
    make-dummies 0 9999 -> 1m10s
    rmmod dummy         -> 0.39s

The sysctl sanity checks now impose no measurable cost.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |  222 +++++++++++++++++++------------------------------
 lib/Kconfig.debug     |    8 --
 2 files changed, 86 insertions(+), 144 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 6bab2ae..a492ff6 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -726,160 +726,106 @@ static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
 	}
 }
 
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-static int sysctl_depth(struct ctl_table *table)
+static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
+	struct ctl_table *table)
 {
-	struct ctl_table *tmp;
-	int depth;
-
-	depth = 0;
-	for (tmp = table; tmp->parent; tmp = tmp->parent)
-		depth++;
+	struct ctl_table *entry, *test;
+	int error = 0;
 
-	return depth;
+	for (entry = old; entry->procname; entry++) {
+		for (test = table; test->procname; test++) {
+			if (strcmp(entry->procname, test->procname) == 0) {
+				printk(KERN_ERR "sysctl duplicate entry: %s/%s\n",
+					path, test->procname);
+				error = -EEXIST;
+			}
+		}
+	}
+	return error;
 }
 
-static struct ctl_table *sysctl_parent(struct ctl_table *table, int n)
+static int sysctl_check_dups(struct nsproxy *namespaces,
+	struct ctl_table_header *header,
+	const char *path, struct ctl_table *table)
 {
-	int i;
+	struct ctl_table_root *root;
+	struct ctl_table_set *set;
+	struct ctl_table_header *dir_head, *head;
+	struct ctl_table *dir_table;
+	int error = 0;
 
-	for (i = 0; table && i < n; i++)
-		table = table->parent;
+	/* No dups if we are the only member of our directory */
+	if (header->attached_by != table)
+		return 0;
 
-	return table;
-}
+	dir_head = header->parent;
+	dir_table = header->attached_to;
 
+	error = sysctl_check_table_dups(path, dir_table, table);
 
-static void sysctl_print_path(struct ctl_table *table)
-{
-	struct ctl_table *tmp;
-	int depth, i;
-	depth = sysctl_depth(table);
-	if (table->procname) {
-		for (i = depth; i >= 0; i--) {
-			tmp = sysctl_parent(table, i);
-			printk("/%s", tmp->procname?tmp->procname:"");
-		}
-	}
-	printk(" ");
-}
+	root = &sysctl_table_root;
+	do {
+		set = lookup_header_set(root, namespaces);
 
-static struct ctl_table *sysctl_check_lookup(struct nsproxy *namespaces,
-						struct ctl_table *table)
-{
-	struct ctl_table_header *head;
-	struct ctl_table *ref, *test;
-	int depth, cur_depth;
-
-	depth = sysctl_depth(table);
-
-	for (head = __sysctl_head_next(namespaces, NULL); head;
-	     head = __sysctl_head_next(namespaces, head)) {
-		cur_depth = depth;
-		ref = head->ctl_table;
-repeat:
-		test = sysctl_parent(table, cur_depth);
-		for (; ref->procname; ref++) {
-			int match = 0;
-			if (cur_depth && !ref->child)
+		list_for_each_entry(head, &set->list, ctl_entry) {
+			if (head->unregistering)
 				continue;
-
-			if (test->procname && ref->procname &&
-			    (strcmp(test->procname, ref->procname) == 0))
-					match++;
-
-			if (match) {
-				if (cur_depth != 0) {
-					cur_depth--;
-					ref = ref->child;
-					goto repeat;
-				}
-				goto out;
-			}
+			if (head->attached_to != dir_table)
+				continue;
+			error = sysctl_check_table_dups(path, head->attached_by,
+							table);
 		}
-	}
-	ref = NULL;
-out:
-	sysctl_head_finish(head);
-	return ref;
+		root = list_entry(root->root_list.next,
+				  struct ctl_table_root, root_list);
+	} while (root != &sysctl_table_root);
+	return error;
 }
 
-static void set_fail(const char **fail, struct ctl_table *table, const char *str)
+static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
 {
-	if (*fail) {
-		printk(KERN_ERR "sysctl table check failed: ");
-		sysctl_print_path(table);
-		printk(" %s\n", *fail);
-		dump_stack();
-	}
-	*fail = str;
-}
+	struct va_format vaf;
+	va_list args;
 
-static void sysctl_check_leaf(struct nsproxy *namespaces,
-				struct ctl_table *table, const char **fail)
-{
-	struct ctl_table *ref;
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	printk(KERN_ERR "sysctl table check failed: %s/%s %pV\n",
+		path, table->procname, &vaf);
 
-	ref = sysctl_check_lookup(namespaces, table);
-	if (ref && (ref != table))
-		set_fail(fail, table, "Sysctl already exists");
+	va_end(args);
+	return -EINVAL;
 }
 
-static int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
+static int sysctl_check_table(const char *path, struct ctl_table *table)
 {
-	int error = 0;
+	int err = 0;
 	for (; table->procname; table++) {
-		const char *fail = NULL;
-
-		if (table->parent) {
-			if (!table->parent->procname)
-				set_fail(&fail, table, "Parent without procname");
-		}
-		if (table->child) {
-			if (table->data)
-				set_fail(&fail, table, "Directory with data?");
-			if (table->maxlen)
-				set_fail(&fail, table, "Directory with maxlen?");
-			if ((table->mode & (S_IRUGO|S_IXUGO)) != table->mode)
-				set_fail(&fail, table, "Writable sysctl directory");
-			if (table->proc_handler)
-				set_fail(&fail, table, "Directory with proc_handler");
-			if (table->extra1)
-				set_fail(&fail, table, "Directory with extra1");
-			if (table->extra2)
-				set_fail(&fail, table, "Directory with extra2");
-		} else {
-			if ((table->proc_handler == proc_dostring) ||
-			    (table->proc_handler == proc_dointvec) ||
-			    (table->proc_handler == proc_dointvec_minmax) ||
-			    (table->proc_handler == proc_dointvec_jiffies) ||
-			    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
-			    (table->proc_handler == proc_dointvec_ms_jiffies) ||
-			    (table->proc_handler == proc_doulongvec_minmax) ||
-			    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
-				if (!table->data)
-					set_fail(&fail, table, "No data");
-				if (!table->maxlen)
-					set_fail(&fail, table, "No maxlen");
-			}
-#ifdef CONFIG_PROC_SYSCTL
-			if (!table->proc_handler)
-				set_fail(&fail, table, "No proc_handler");
-#endif
-			sysctl_check_leaf(namespaces, table, &fail);
-		}
-		if (table->mode > 0777)
-			set_fail(&fail, table, "bogus .mode");
-		if (fail) {
-			set_fail(&fail, table, NULL);
-			error = -EINVAL;
-		}
 		if (table->child)
-			error |= sysctl_check_table(namespaces, table->child);
+			err = sysctl_err(path, table, "Not a file");
+
+		if ((table->proc_handler == proc_dostring) ||
+		    (table->proc_handler == proc_dointvec) ||
+		    (table->proc_handler == proc_dointvec_minmax) ||
+		    (table->proc_handler == proc_dointvec_jiffies) ||
+		    (table->proc_handler == proc_dointvec_userhz_jiffies) ||
+		    (table->proc_handler == proc_dointvec_ms_jiffies) ||
+		    (table->proc_handler == proc_doulongvec_minmax) ||
+		    (table->proc_handler == proc_doulongvec_ms_jiffies_minmax)) {
+			if (!table->data)
+				err = sysctl_err(path, table, "No data");
+			if (!table->maxlen)
+				err = sysctl_err(path, table, "No maxlen");
+		}
+		if (!table->proc_handler)
+			err = sysctl_err(path, table, "No proc_handler");
+
+		if ((table->mode & (S_IRUGO|S_IWUGO)) != table->mode)
+			err = sysctl_err(path, table, "bogus .mode 0%o",
+				table->mode);
 	}
-	return error;
+	return err;
 }
-#endif /* CONFIG_SYSCTL_SYSCALL_CHECK */
 
 /**
  * __register_sysctl_table - register a leaf sysctl table
@@ -1003,12 +949,8 @@ struct ctl_table_header *__register_sysctl_table(
 	header->root = root;
 	sysctl_set_parent(NULL, header->ctl_table);
 	header->count = 1;
-#ifdef CONFIG_SYSCTL_SYSCALL_CHECK
-	if (sysctl_check_table(namespaces, header->ctl_table)) {
-		kfree(header);
-		return NULL;
-	}
-#endif
+	if (sysctl_check_table(path, table))
+		goto fail;
 	spin_lock(&sysctl_lock);
 	header->set = lookup_header_set(root, namespaces);
 	header->attached_by = header->ctl_table;
@@ -1029,11 +971,19 @@ struct ctl_table_header *__register_sysctl_table(
 				  struct ctl_table_root, root_list);
 		set = lookup_header_set(root, namespaces);
 	}
+	if (sysctl_check_dups(namespaces, header, path, table))
+		goto fail_locked;
 	header->parent->count++;
 	list_add_tail(&header->ctl_entry, &header->set->list);
 	spin_unlock(&sysctl_lock);
 
 	return header;
+fail_locked:
+	spin_unlock(&sysctl_lock);
+fail:
+	kfree(header);
+	dump_stack();
+	return NULL;
 }
 
 static char *append_path(const char *path, char *pos, const char *name)
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8745ac7..943a618 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1113,14 +1113,6 @@ config LATENCYTOP
 	  Enable this option if you want to use the LatencyTOP tool
 	  to find out which userspace is blocking on what kernel operations.
 
-config SYSCTL_SYSCALL_CHECK
-	bool "Sysctl checks"
-	depends on SYSCTL
-	---help---
-	  sys_sysctl uses binary paths that have been found challenging
-	  to properly maintain and use. This enables checks that help
-	  you to keep things correct.
-
 source mm/Kconfig.debug
 source kernel/trace/Kconfig
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (40 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 12/29] sysctl: Improve the sysctl sanity checks Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 14/29] sysctl: A more obvious version of grab_header Eric W. Biederman
                   ` (16 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

While useful at one time for selinux and the sysctl sanity
checks those users no longer use the parent field and we can
safely remove it.

Inspired-by: Lucian Adrian Grijincu <lucian.grijincu@gmil.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   12 +-----------
 include/linux/sysctl.h |    1 -
 2 files changed, 1 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a492ff6..e573f9b 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -218,16 +218,6 @@ static int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int
 	return test_perm(mode, op);
 }
 
-static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
-{
-	for (; table->procname; table++) {
-		table->parent = parent;
-		if (table->child)
-			sysctl_set_parent(table, table->child);
-	}
-}
-
-
 static struct inode *proc_sys_make_inode(struct super_block *sb,
 		struct ctl_table_header *head, struct ctl_table *table)
 {
@@ -947,10 +937,10 @@ struct ctl_table_header *__register_sysctl_table(
 	header->used = 0;
 	header->unregistering = NULL;
 	header->root = root;
-	sysctl_set_parent(NULL, header->ctl_table);
 	header->count = 1;
 	if (sysctl_check_table(path, table))
 		goto fail;
+
 	spin_lock(&sysctl_lock);
 	header->set = lookup_header_set(root, namespaces);
 	header->attached_by = header->ctl_table;
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 25c7dfe..094bc5c 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1016,7 +1016,6 @@ struct ctl_table
 	int maxlen;
 	umode_t mode;
 	struct ctl_table *child;	/* Deprecated */
-	struct ctl_table *parent;	/* Automatically set */
 	proc_handler *proc_handler;	/* Callback for text formatting */
 	struct ctl_table_poll *poll;
 	void *extra1;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 14/29] sysctl: A more obvious version of grab_header.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (41 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables Eric W. Biederman
                   ` (15 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Instead of relying on sysct_head_next(NULL) to magically
return the right header for the root directory instead
explicitly transform NULL into the root directories header.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e573f9b..1544485 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -267,10 +267,10 @@ static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
 
 static struct ctl_table_header *grab_header(struct inode *inode)
 {
-	if (PROC_I(inode)->sysctl)
-		return sysctl_head_grab(PROC_I(inode)->sysctl);
-	else
-		return sysctl_head_next(NULL);
+	struct ctl_table_header *head = PROC_I(inode)->sysctl;
+	if (!head)
+		head = &root_table_header;
+	return sysctl_head_grab(head);
 }
 
 static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (42 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 14/29] sysctl: A more obvious version of grab_header Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths Eric W. Biederman
                   ` (14 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Add nreg to ctl_table_header.  When nreg drops to 0 the ctl_table_header
will be unregistered.

Factor out drop_sysctl_table from unregister_sysctl_table, and add
the logic for decrementing nreg.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   28 +++++++++++++++++++---------
 include/linux/sysctl.h |    1 +
 2 files changed, 20 insertions(+), 9 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 1544485..13faa48 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -29,8 +29,9 @@ static struct ctl_table root_table[1];
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
 	{{.count = 1,
-	.ctl_table = root_table,
-	.ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+	  .nreg = 1,
+	  .ctl_table = root_table,
+	  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
 	.root = &sysctl_table_root,
 	.set = &sysctl_table_root.default_set,
 };
@@ -938,6 +939,7 @@ struct ctl_table_header *__register_sysctl_table(
 	header->unregistering = NULL;
 	header->root = root;
 	header->count = 1;
+	header->nreg = 1;
 	if (sysctl_check_table(path, table))
 		goto fail;
 
@@ -1192,6 +1194,20 @@ struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
 }
 EXPORT_SYMBOL(register_sysctl_table);
 
+static void drop_sysctl_table(struct ctl_table_header *header)
+{
+	if (--header->nreg)
+		return;
+
+	start_unregistering(header);
+	if (!--header->parent->count) {
+		WARN_ON(1);
+		kfree_rcu(header->parent, rcu);
+	}
+	if (!--header->count)
+		kfree_rcu(header, rcu);
+}
+
 /**
  * unregister_sysctl_table - unregister a sysctl table hierarchy
  * @header: the header returned from register_sysctl_table
@@ -1224,13 +1240,7 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 	}
 
 	spin_lock(&sysctl_lock);
-	start_unregistering(header);
-	if (!--header->parent->count) {
-		WARN_ON(1);
-		kfree_rcu(header->parent, rcu);
-	}
-	if (!--header->count)
-		kfree_rcu(header, rcu);
+	drop_sysctl_table(header);
 	spin_unlock(&sysctl_lock);
 }
 EXPORT_SYMBOL(unregister_sysctl_table);
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 094bc5c..e40b8f6 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1032,6 +1032,7 @@ struct ctl_table_header
 			struct list_head ctl_entry;
 			int used;
 			int count;
+			int nreg;
 		};
 		struct rcu_head rcu;
 	};
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (43 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 17/29] sysctl: Factor out insert_header and erase_header Eric W. Biederman
                   ` (13 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Factor out a routing to initialize the sysctl_table_header.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   25 +++++++++++++++++--------
 1 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 13faa48..4979925 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -42,6 +42,21 @@ static struct ctl_table_root sysctl_table_root = {
 
 static DEFINE_SPINLOCK(sysctl_lock);
 
+static void init_header(struct ctl_table_header *head,
+	struct ctl_table_root *root, struct ctl_table_set *set,
+	struct ctl_table *table)
+{
+	head->ctl_table_arg = table;
+	INIT_LIST_HEAD(&head->ctl_entry);
+	head->used = 0;
+	head->count = 1;
+	head->nreg = 1;
+	head->unregistering = NULL;
+	head->root = root;
+	head->set = set;
+	head->parent = NULL;
+}
+
 /* called under sysctl_lock */
 static int use_table(struct ctl_table_header *p)
 {
@@ -932,14 +947,8 @@ struct ctl_table_header *__register_sysctl_table(
 		new_name += namelen + 1;
 	}
 	*prevp = table;
-	header->ctl_table_arg = table;
-
-	INIT_LIST_HEAD(&header->ctl_entry);
-	header->used = 0;
-	header->unregistering = NULL;
-	header->root = root;
-	header->count = 1;
-	header->nreg = 1;
+
+	init_header(header, root, NULL, table);
 	if (sysctl_check_table(path, table))
 		goto fail;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 17/29] sysctl: Factor out insert_header and erase_header
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (44 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
                   ` (12 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   16 +++++++++++++---
 1 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 4979925..7e96a26 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -57,6 +57,17 @@ static void init_header(struct ctl_table_header *head,
 	head->parent = NULL;
 }
 
+static void erase_header(struct ctl_table_header *head)
+{
+	list_del_init(&head->ctl_entry);
+}
+
+static void insert_header(struct ctl_table_header *header)
+{
+	header->parent->count++;
+	list_add_tail(&header->ctl_entry, &header->set->list);
+}
+
 /* called under sysctl_lock */
 static int use_table(struct ctl_table_header *p)
 {
@@ -96,7 +107,7 @@ static void start_unregistering(struct ctl_table_header *p)
 	 * do not remove from the list until nobody holds it; walking the
 	 * list in do_sysctl() relies on that.
 	 */
-	list_del_init(&p->ctl_entry);
+	erase_header(p);
 }
 
 static void sysctl_head_get(struct ctl_table_header *head)
@@ -974,8 +985,7 @@ struct ctl_table_header *__register_sysctl_table(
 	}
 	if (sysctl_check_dups(namespaces, header, path, table))
 		goto fail_locked;
-	header->parent->count++;
-	list_add_tail(&header->ctl_entry, &header->set->list);
+	insert_header(header);
 	spin_unlock(&sysctl_lock);
 
 	return header;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 18/29] sysctl: Normalize the root_table data structure.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (45 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 17/29] sysctl: Factor out insert_header and erase_header Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-29 17:36   ` Lucian Adrian Grijincu
  2012-01-27  4:51 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
                   ` (11 subsequent siblings)
  58 siblings, 1 reply; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Every other directory has a .child member and we look at the .child
for our entries.  Do the same for the root_table.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   15 +++++++++++----
 1 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 7e96a26..88d1b06 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -25,7 +25,14 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
 	wake_up_interruptible(&poll->wait);
 }
 
-static struct ctl_table root_table[1];
+static struct ctl_table root_table[] = {
+	{
+		.procname = "",
+		.mode = S_IRUGO|S_IXUGO,
+		.child = &root_table[1],
+	},
+	{ }
+};
 static struct ctl_table_root sysctl_table_root;
 static struct ctl_table_header root_table_header = {
 	{{.count = 1,
@@ -319,7 +326,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 		goto out;
 	}
 
-	table = table ? table->child : head->ctl_table;
+	table = table ? table->child : &head->ctl_table[1];
 
 	p = find_in_table(table, name);
 	if (!p) {
@@ -510,7 +517,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 		goto out;
 	}
 
-	table = table ? table->child : head->ctl_table;
+	table = table ? table->child : &head->ctl_table[1];
 
 	ret = 0;
 	/* Avoid a switch here: arm builds fail with missing __cmpdi2 */
@@ -966,7 +973,7 @@ struct ctl_table_header *__register_sysctl_table(
 	spin_lock(&sysctl_lock);
 	header->set = lookup_header_set(root, namespaces);
 	header->attached_by = header->ctl_table;
-	header->attached_to = root_table;
+	header->attached_to = &root_table[1];
 	header->parent = &root_table_header;
 	set = header->set;
 	root = header->root;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (46 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-29 15:49   ` Lucian Adrian Grijincu
  2012-01-27  4:51 ` [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry Eric W. Biederman
                   ` (10 subsequent siblings)
  58 siblings, 1 reply; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Replace the helpers that proc_sys_lookup uses with helpers that work
in terms of an entire sysctl directory.  This is worse for sysctl_lock
hold times but it is much better for code clarity and the code cleanups
to come.

find_in_table is no longer needed so it is removed.

find_entry a general helper to find entries in a directory is added.

lookup_entry is a simple wrapper around find_entry that takes the
sysctl_lock increases the use count if an entry is found and drops
the sysctl_lock.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |  102 ++++++++++++++++++++++++++++++++++++------------
 1 files changed, 76 insertions(+), 26 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 88d1b06..3b63f29 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -49,6 +49,55 @@ static struct ctl_table_root sysctl_table_root = {
 
 static DEFINE_SPINLOCK(sysctl_lock);
 
+static int namecmp(const char *name1, int len1, const char *name2, int len2)
+{
+	int minlen;
+	int cmp;
+
+	minlen = len1;
+	if (minlen > len2)
+		minlen = len2;
+
+	cmp = memcmp(name1, name2, minlen);
+	if (cmp == 0)
+		cmp = len1 - len2;
+	return cmp;
+}
+
+static struct ctl_table *find_entry(struct ctl_table_header **phead,
+	struct ctl_table_set *set,
+	struct ctl_table_header *dir_head, struct ctl_table *dir,
+	const char *name, int namelen)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *entry;
+
+	if (dir_head->set == set) {
+		for (entry = dir; entry->procname; entry++) {
+			const char *procname = entry->procname;
+			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
+				*phead = dir_head;
+				return entry;
+			}
+		}
+	}
+
+	list_for_each_entry(head, &set->list, ctl_entry) {
+		if (head->unregistering)
+			continue;
+		if (head->attached_to != dir)
+			continue;
+		for (entry = head->attached_by; entry->procname; entry++) {
+			const char *procname = entry->procname;
+			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
+				*phead = head;
+				return entry;
+			}
+		}
+	}
+	return NULL;
+}
+
 static void init_header(struct ctl_table_header *head,
 	struct ctl_table_root *root, struct ctl_table_set *set,
 	struct ctl_table *table)
@@ -168,6 +217,32 @@ lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
 	return &set->list;
 }
 
+static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
+				      struct ctl_table_header *dir_head,
+				      struct ctl_table *dir,
+				      const char *name, int namelen)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *entry;
+	struct ctl_table_root *root;
+	struct ctl_table_set *set;
+
+	spin_lock(&sysctl_lock);
+	root = &sysctl_table_root;
+	do {
+		set = lookup_header_set(root, current->nsproxy);
+		entry = find_entry(&head, set, dir_head, dir, name, namelen);
+		if (entry && use_table(head))
+			*phead = head;
+		else
+			entry = NULL;
+		root = list_entry(root->root_list.next,
+				  struct ctl_table_root, root_list);
+	} while (!entry && root != &sysctl_table_root);
+	spin_unlock(&sysctl_lock);
+	return entry;
+}
+
 static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
 						struct ctl_table_header *prev)
 {
@@ -284,21 +359,6 @@ out:
 	return inode;
 }
 
-static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
-{
-	for ( ; p->procname; p++) {
-		if (strlen(p->procname) != name->len)
-			continue;
-
-		if (memcmp(p->procname, name->name, name->len) != 0)
-			continue;
-
-		/* I have a match */
-		return p;
-	}
-	return NULL;
-}
-
 static struct ctl_table_header *grab_header(struct inode *inode)
 {
 	struct ctl_table_header *head = PROC_I(inode)->sysctl;
@@ -328,17 +388,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 
 	table = table ? table->child : &head->ctl_table[1];
 
-	p = find_in_table(table, name);
-	if (!p) {
-		for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
-			if (h->attached_to != table)
-				continue;
-			p = find_in_table(h->attached_by, name);
-			if (p)
-				break;
-		}
-	}
-
+	p = lookup_entry(&h, head, table, name->name, name->len);
 	if (!p)
 		goto out;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (47 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:51 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
                   ` (9 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Replace sysctl_head_next with first_entry and next_entry.  These new
iterators operate at the level of sysctl table entries and filter
out any sysctl tables that should not be shown.

Utilizing two specialized functions instead of a single function removes
conditionals for handling awkward special cases that only come up
at the beginning of iteration, making the iterators easier to read
and understand.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   98 +++++++++++++++++++++++++++++++------------------
 1 files changed, 62 insertions(+), 36 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 3b63f29..d9c3ae6 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -243,31 +243,25 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	return entry;
 }
 
-static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
-						struct ctl_table_header *prev)
+static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
+	struct ctl_table_root *root, struct list_head *tmp)
 {
-	struct ctl_table_root *root;
+	struct nsproxy *namespaces = current->nsproxy;
 	struct list_head *header_list;
 	struct ctl_table_header *head;
-	struct list_head *tmp;
 
-	spin_lock(&sysctl_lock);
-	if (prev) {
-		head = prev;
-		tmp = &prev->ctl_entry;
-		unuse_table(prev);
-		goto next;
-	}
-	tmp = &root_table_header.ctl_entry;
+	goto next;
 	for (;;) {
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
+		root = head->root;
 
-		if (!use_table(head))
+		if (head->attached_to != dir ||
+		    !head->attached_by->procname ||
+		    !use_table(head))
 			goto next;
-		spin_unlock(&sysctl_lock);
+
 		return head;
 	next:
-		root = head->root;
 		tmp = tmp->next;
 		header_list = lookup_header_list(root, namespaces);
 		if (tmp != header_list)
@@ -283,13 +277,53 @@ static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
 		tmp = header_list->next;
 	}
 out:
-	spin_unlock(&sysctl_lock);
 	return NULL;
 }
 
-static struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
+static void first_entry(
+	struct ctl_table_header *dir_head, struct ctl_table *dir,
+	struct ctl_table_header **phead, struct ctl_table **pentry)
 {
-	return __sysctl_head_next(current->nsproxy, prev);
+	struct ctl_table_header *head = dir_head;
+	struct ctl_table *entry = dir;
+
+	spin_lock(&sysctl_lock);
+	if (entry->procname) {
+		use_table(head);
+	} else {
+		head = next_usable_entry(dir, &sysctl_table_root,
+					 &sysctl_table_root.default_set.list);
+		if (head)
+			entry = head->attached_by;
+	}
+	spin_unlock(&sysctl_lock);
+	*phead = head;
+	*pentry = entry;
+}
+
+static void next_entry(struct ctl_table *dir,
+	struct ctl_table_header **phead, struct ctl_table **pentry)
+{
+	struct ctl_table_header *head = *phead;
+	struct ctl_table *entry = *pentry;
+
+	entry++;
+	if (!entry->procname) {
+		struct ctl_table_root *root = head->root;
+		struct list_head *tmp = &head->ctl_entry;
+		if (head->attached_to != dir) {
+			root = &sysctl_table_root;
+			tmp = &sysctl_table_root.default_set.list;
+		}
+		spin_lock(&sysctl_lock);
+		unuse_table(head);
+		head = next_usable_entry(dir, root, tmp);
+		spin_unlock(&sysctl_lock);
+		if (head)
+			entry = head->attached_by;
+	}
+	*phead = head;
+	*pentry = entry;
 }
 
 void register_sysctl_root(struct ctl_table_root *root)
@@ -533,20 +567,17 @@ static int scan(struct ctl_table_header *head, ctl_table *table,
 		unsigned long *pos, struct file *file,
 		void *dirent, filldir_t filldir)
 {
+	int res;
 
-	for (; table->procname; table++, (*pos)++) {
-		int res;
+	if ((*pos)++ < file->f_pos)
+		return 0;
 
-		if (*pos < file->f_pos)
-			continue;
+	res = proc_sys_fill_cache(file, dirent, filldir, head, table);
 
-		res = proc_sys_fill_cache(file, dirent, filldir, head, table);
-		if (res)
-			return res;
+	if (res == 0)
+		file->f_pos = *pos;
 
-		file->f_pos = *pos + 1;
-	}
-	return 0;
+	return res;
 }
 
 static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
@@ -556,6 +587,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct ctl_table_header *head = grab_header(inode);
 	struct ctl_table *table = PROC_I(inode)->sysctl_entry;
 	struct ctl_table_header *h = NULL;
+	struct ctl_table *entry;
 	unsigned long pos;
 	int ret = -EINVAL;
 
@@ -585,14 +617,8 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	}
 	pos = 2;
 
-	ret = scan(head, table, &pos, filp, dirent, filldir);
-	if (ret)
-		goto out;
-
-	for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
-		if (h->attached_to != table)
-			continue;
-		ret = scan(h, h->attached_by, &pos, filp, dirent, filldir);
+	for (first_entry(head, table, &h, &entry); h; next_entry(table, &h, &entry)) {
+		ret = scan(h, entry, &pos, filp, dirent, filldir);
 		if (ret) {
 			sysctl_head_finish(h);
 			break;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (48 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-29 17:19   ` Lucian Adrian Grijincu
  2012-01-27  4:51 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
                   ` (8 subsequent siblings)
  58 siblings, 1 reply; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Add a ctl_table_root pointer to ctl_table set so it is easy to
go from a ctl_table_set to a ctl_table_root.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |    3 +++
 include/linux/sysctl.h |    3 +++
 net/sysctl_net.c       |    5 ++---
 3 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index d9c3ae6..65c13dd 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -45,6 +45,7 @@ static struct ctl_table_header root_table_header = {
 static struct ctl_table_root sysctl_table_root = {
 	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+	.default_set.root = &sysctl_table_root,
 };
 
 static DEFINE_SPINLOCK(sysctl_lock);
@@ -1348,9 +1349,11 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 EXPORT_SYMBOL(unregister_sysctl_table);
 
 void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *))
 {
 	INIT_LIST_HEAD(&p->list);
+	p->root = root;
 	p->is_seen = is_seen;
 }
 
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index e40b8f6..e73ba33 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1047,6 +1047,7 @@ struct ctl_table_header
 
 struct ctl_table_set {
 	struct list_head list;
+	struct ctl_table_root *root;
 	int (*is_seen)(struct ctl_table_set *);
 };
 
@@ -1069,6 +1070,7 @@ struct ctl_path {
 void proc_sys_poll_notify(struct ctl_table_poll *poll);
 
 extern void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *));
 extern void retire_sysctl_set(struct ctl_table_set *set);
 
@@ -1103,6 +1105,7 @@ static inline void unregister_sysctl_table(struct ctl_table_header * table)
 }
 
 static inline void setup_sysctl_set(struct ctl_table_set *p,
+	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *))
 {
 }
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index 07c6b87..e998c64 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -74,8 +74,7 @@ static struct ctl_table_root net_sysctl_ro_root = {
 
 static int __net_init sysctl_net_init(struct net *net)
 {
-	setup_sysctl_set(&net->sysctls,
-			 is_seen);
+	setup_sysctl_set(&net->sysctls, &net_sysctl_root, is_seen);
 	return 0;
 }
 
@@ -95,7 +94,7 @@ static __init int net_sysctl_init(void)
 	ret = register_pernet_subsys(&sysctl_pernet_ops);
 	if (ret)
 		goto out;
-	setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL);
+	setup_sysctl_set(&net_sysctl_ro_root.default_set, &net_sysctl_ro_root, NULL);
 	register_sysctl_root(&net_sysctl_ro_root);
 	register_sysctl_root(&net_sysctl_root);
 out:
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (49 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-29 19:31   ` Lucian Adrian Grijincu
  2012-01-27  4:51 ` [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir Eric W. Biederman
                   ` (7 subsequent siblings)
  58 siblings, 1 reply; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Simplify the code and the sysctl semantics by autogenerating
sysctl directories when a sysctl table is registered that needs
the directories and autodeleting the directories when there are
no more sysctl tables registered that need them.

Autogenerating directories keeps sysctl tables from depending
on each other, removing all of the arcane register/unregister
ordering constraints and makes it impossible to get the order
wrong when reigsering and unregistering sysctl tables.

Autogenerating directories yields one unique entity that dentries
can point to, retaining the current effective use of the dcache.

Add struct ctl_dir as the type of these new autogenerated
directories.

The attached_by and attached_to fields in ctl_table_header are
removed as they are no longer needed.

The child field in ctl_table is no longer needed by the core of
the sysctl code.  ctl_table.child can be removed once all of the
existing users have been updated.

Benchmark before:
    make-dummies 0 999 -> 0.7s
    rmmod dummy        -> 0.07s
    make-dummies 0 9999 -> 1m10s
    rmmod dummy         -> 0.4s

Benchmark after:
    make-dummies 0 999 -> 0.44s
    rmmod dummy        -> 0.065s
    make-dummies 0 9999 -> 1m36s
    rmmod dummy         -> 0.4s

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  342 ++++++++++++++++++++----------------------------
 include/linux/sysctl.h |   10 +-
 2 files changed, 150 insertions(+), 202 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 65c13dd..3c0767d 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -28,28 +28,31 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
 static struct ctl_table root_table[] = {
 	{
 		.procname = "",
-		.mode = S_IRUGO|S_IXUGO,
-		.child = &root_table[1],
+		.mode = S_IFDIR|S_IRUGO|S_IXUGO,
 	},
 	{ }
 };
 static struct ctl_table_root sysctl_table_root;
-static struct ctl_table_header root_table_header = {
-	{{.count = 1,
-	  .nreg = 1,
-	  .ctl_table = root_table,
-	  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
-	.root = &sysctl_table_root,
-	.set = &sysctl_table_root.default_set,
+static struct ctl_dir sysctl_root_dir = {
+	.header = {
+		{{.count = 1,
+		  .nreg = 1,
+		  .ctl_table = root_table,
+		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+		.root = &sysctl_table_root,
+		.set = &sysctl_table_root.default_set,
+	},
 };
 static struct ctl_table_root sysctl_table_root = {
 	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-	.default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
+	.default_set.list = LIST_HEAD_INIT(sysctl_root_dir.header.ctl_entry),
 	.default_set.root = &sysctl_table_root,
 };
 
 static DEFINE_SPINLOCK(sysctl_lock);
 
+static void drop_sysctl_table(struct ctl_table_header *header);
+
 static int namecmp(const char *name1, int len1, const char *name2, int len2)
 {
 	int minlen;
@@ -66,29 +69,18 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 }
 
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
-	struct ctl_table_set *set,
-	struct ctl_table_header *dir_head, struct ctl_table *dir,
+	struct ctl_table_set *set, struct ctl_dir *dir,
 	const char *name, int namelen)
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
-	if (dir_head->set == set) {
-		for (entry = dir; entry->procname; entry++) {
-			const char *procname = entry->procname;
-			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
-				*phead = dir_head;
-				return entry;
-			}
-		}
-	}
-
 	list_for_each_entry(head, &set->list, ctl_entry) {
 		if (head->unregistering)
 			continue;
-		if (head->attached_to != dir)
+		if (head->parent != dir)
 			continue;
-		for (entry = head->attached_by; entry->procname; entry++) {
+		for (entry = head->ctl_table; entry->procname; entry++) {
 			const char *procname = entry->procname;
 			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
 				*phead = head;
@@ -103,6 +95,7 @@ static void init_header(struct ctl_table_header *head,
 	struct ctl_table_root *root, struct ctl_table_set *set,
 	struct ctl_table *table)
 {
+	head->ctl_table = table;
 	head->ctl_table_arg = table;
 	INIT_LIST_HEAD(&head->ctl_entry);
 	head->used = 0;
@@ -119,9 +112,10 @@ static void erase_header(struct ctl_table_header *head)
 	list_del_init(&head->ctl_entry);
 }
 
-static void insert_header(struct ctl_table_header *header)
+static void insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
-	header->parent->count++;
+	header->parent = dir;
+	header->parent->header.nreg++;
 	list_add_tail(&header->ctl_entry, &header->set->list);
 }
 
@@ -219,8 +213,7 @@ lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
 }
 
 static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
-				      struct ctl_table_header *dir_head,
-				      struct ctl_table *dir,
+				      struct ctl_dir *dir,
 				      const char *name, int namelen)
 {
 	struct ctl_table_header *head;
@@ -232,7 +225,7 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	root = &sysctl_table_root;
 	do {
 		set = lookup_header_set(root, current->nsproxy);
-		entry = find_entry(&head, set, dir_head, dir, name, namelen);
+		entry = find_entry(&head, set, dir, name, namelen);
 		if (entry && use_table(head))
 			*phead = head;
 		else
@@ -244,7 +237,7 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	return entry;
 }
 
-static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
+static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
 	struct ctl_table_root *root, struct list_head *tmp)
 {
 	struct nsproxy *namespaces = current->nsproxy;
@@ -256,8 +249,8 @@ static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
 		root = head->root;
 
-		if (head->attached_to != dir ||
-		    !head->attached_by->procname ||
+		if (head->parent != dir ||
+		    !head->ctl_table->procname ||
 		    !use_table(head))
 			goto next;
 
@@ -281,47 +274,35 @@ out:
 	return NULL;
 }
 
-static void first_entry(
-	struct ctl_table_header *dir_head, struct ctl_table *dir,
+static void first_entry(struct ctl_dir *dir,
 	struct ctl_table_header **phead, struct ctl_table **pentry)
 {
-	struct ctl_table_header *head = dir_head;
-	struct ctl_table *entry = dir;
+	struct ctl_table_header *head;
+	struct ctl_table *entry = NULL;
 
 	spin_lock(&sysctl_lock);
-	if (entry->procname) {
-		use_table(head);
-	} else {
-		head = next_usable_entry(dir, &sysctl_table_root,
-					 &sysctl_table_root.default_set.list);
-		if (head)
-			entry = head->attached_by;
-	}
+	head = next_usable_entry(dir, &sysctl_table_root,
+				 &sysctl_table_root.default_set.list);
 	spin_unlock(&sysctl_lock);
+	if (head)
+		entry = head->ctl_table;
 	*phead = head;
 	*pentry = entry;
 }
 
-static void next_entry(struct ctl_table *dir,
-	struct ctl_table_header **phead, struct ctl_table **pentry)
+static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry)
 {
 	struct ctl_table_header *head = *phead;
 	struct ctl_table *entry = *pentry;
 
 	entry++;
 	if (!entry->procname) {
-		struct ctl_table_root *root = head->root;
-		struct list_head *tmp = &head->ctl_entry;
-		if (head->attached_to != dir) {
-			root = &sysctl_table_root;
-			tmp = &sysctl_table_root.default_set.list;
-		}
 		spin_lock(&sysctl_lock);
 		unuse_table(head);
-		head = next_usable_entry(dir, root, tmp);
+		head = next_usable_entry(head->parent, head->root, &head->ctl_entry);
 		spin_unlock(&sysctl_lock);
 		if (head)
-			entry = head->attached_by;
+			entry = head->ctl_table;
 	}
 	*phead = head;
 	*pentry = entry;
@@ -381,7 +362,7 @@ static struct inode *proc_sys_make_inode(struct super_block *sb,
 
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 	inode->i_mode = table->mode;
-	if (!table->child) {
+	if (!S_ISDIR(table->mode)) {
 		inode->i_mode |= S_IFREG;
 		inode->i_op = &proc_sys_inode_operations;
 		inode->i_fop = &proc_sys_file_operations;
@@ -398,7 +379,7 @@ static struct ctl_table_header *grab_header(struct inode *inode)
 {
 	struct ctl_table_header *head = PROC_I(inode)->sysctl;
 	if (!head)
-		head = &root_table_header;
+		head = &sysctl_root_dir.header;
 	return sysctl_head_grab(head);
 }
 
@@ -406,24 +387,19 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 					struct nameidata *nd)
 {
 	struct ctl_table_header *head = grab_header(dir);
-	struct ctl_table *table = PROC_I(dir)->sysctl_entry;
 	struct ctl_table_header *h = NULL;
 	struct qstr *name = &dentry->d_name;
 	struct ctl_table *p;
 	struct inode *inode;
 	struct dentry *err = ERR_PTR(-ENOENT);
+	struct ctl_dir *ctl_dir;
 
 	if (IS_ERR(head))
 		return ERR_CAST(head);
 
-	if (table && !table->child) {
-		WARN_ON(1);
-		goto out;
-	}
+	ctl_dir = container_of(head, struct ctl_dir, header);
 
-	table = table ? table->child : &head->ctl_table[1];
-
-	p = lookup_entry(&h, head, table, name->name, name->len);
+	p = lookup_entry(&h, ctl_dir, name->name, name->len);
 	if (!p)
 		goto out;
 
@@ -586,21 +562,16 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct dentry *dentry = filp->f_path.dentry;
 	struct inode *inode = dentry->d_inode;
 	struct ctl_table_header *head = grab_header(inode);
-	struct ctl_table *table = PROC_I(inode)->sysctl_entry;
 	struct ctl_table_header *h = NULL;
 	struct ctl_table *entry;
+	struct ctl_dir *ctl_dir;
 	unsigned long pos;
 	int ret = -EINVAL;
 
 	if (IS_ERR(head))
 		return PTR_ERR(head);
 
-	if (table && !table->child) {
-		WARN_ON(1);
-		goto out;
-	}
-
-	table = table ? table->child : &head->ctl_table[1];
+	ctl_dir = container_of(head, struct ctl_dir, header);
 
 	ret = 0;
 	/* Avoid a switch here: arm builds fail with missing __cmpdi2 */
@@ -618,7 +589,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	}
 	pos = 2;
 
-	for (first_entry(head, table, &h, &entry); h; next_entry(table, &h, &entry)) {
+	for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) {
 		ret = scan(h, entry, &pos, filp, dirent, filldir);
 		if (ret) {
 			sysctl_head_finish(h);
@@ -779,52 +750,86 @@ static const struct dentry_operations proc_sys_dentry_operations = {
 	.d_compare	= proc_sys_compare,
 };
 
-static struct ctl_table *is_branch_in(struct ctl_table *branch,
-				      struct ctl_table *table)
+static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *dir,
+	const char *name, int namelen)
 {
-	struct ctl_table *p;
-	const char *s = branch->procname;
+	struct ctl_table_header *head;
+	struct ctl_table *entry;
 
-	/* branch should have named subdirectory as its first element */
-	if (!s || !branch->child)
-		return NULL;
+	entry = find_entry(&head, set, dir, name, namelen);
+	if (!entry)
+		return ERR_PTR(-ENOENT);
+	if (S_ISDIR(entry->mode))
+		return container_of(head, struct ctl_dir, header);
+	return ERR_PTR(-ENOTDIR);
+}
+
+static struct ctl_dir *new_dir(struct ctl_table_set *set,
+	const char *name, int namelen)
+{
+	struct ctl_table *table;
+	struct ctl_dir *new;
+	char *new_name;
 
-	/* ... and nothing else */
-	if (branch[1].procname)
+	new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 +
+		      namelen + 1, GFP_KERNEL);
+	if (!new)
 		return NULL;
 
-	/* table should contain subdirectory with the same name */
-	for (p = table; p->procname; p++) {
-		if (!p->child)
-			continue;
-		if (p->procname && strcmp(p->procname, s) == 0)
-			return p;
-	}
-	return NULL;
+	table = (struct ctl_table *)(new + 1);
+	new_name = (char *)(table + 2);
+	memcpy(new_name, name, namelen);
+	new_name[namelen] = '\0';
+	table[0].procname = new_name;
+	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
+	init_header(&new->header, set->root, set, table);
+
+	return new;
 }
 
-/* see if attaching q to p would be an improvement */
-static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
+static struct ctl_dir *get_subdir(struct ctl_table_set *set,
+	struct ctl_dir *dir, const char *name, int namelen)
 {
-	struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
-	struct ctl_table *next;
-	int is_better = 0;
-	int not_in_parent = !p->attached_by;
-
-	while ((next = is_branch_in(by, to)) != NULL) {
-		if (by == q->attached_by)
-			is_better = 1;
-		if (to == p->attached_by)
-			not_in_parent = 1;
-		by = by->child;
-		to = next->child;
-	}
+	struct ctl_dir *subdir, *new = NULL;
 
-	if (is_better && not_in_parent) {
-		q->attached_by = by;
-		q->attached_to = to;
-		q->parent = p;
+	spin_lock(&sysctl_lock);
+	subdir = find_subdir(dir->header.set, dir, name, namelen);
+	if (!IS_ERR(subdir))
+		goto found;
+	if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)
+		subdir = find_subdir(set, dir, name, namelen);
+	if (!IS_ERR(subdir))
+		goto found;
+	if (PTR_ERR(subdir) != -ENOENT)
+		goto failed;
+
+	spin_unlock(&sysctl_lock);
+	new = new_dir(set, name, namelen);
+	spin_lock(&sysctl_lock);
+	subdir = ERR_PTR(-ENOMEM);
+	if (!new)
+		goto failed;
+
+	subdir = find_subdir(set, dir, name, namelen);
+	if (!IS_ERR(subdir))
+		goto found;
+	if (PTR_ERR(subdir) != -ENOENT)
+		goto failed;
+
+	insert_header(dir, &new->header);
+	subdir = new;
+found:
+	subdir->header.nreg++;
+failed:
+	if (unlikely(IS_ERR(subdir))) {
+		printk(KERN_ERR "sysctl could not get directory: %*.*s %ld\n",
+			namelen, namelen, name, PTR_ERR(subdir));
 	}
+	drop_sysctl_table(&dir->header);
+	if (new)
+		drop_sysctl_table(&new->header);
+	spin_unlock(&sysctl_lock);
+	return subdir;
 }
 
 static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
@@ -846,24 +851,14 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 }
 
 static int sysctl_check_dups(struct nsproxy *namespaces,
-	struct ctl_table_header *header,
+	struct ctl_dir *dir,
 	const char *path, struct ctl_table *table)
 {
 	struct ctl_table_root *root;
 	struct ctl_table_set *set;
-	struct ctl_table_header *dir_head, *head;
-	struct ctl_table *dir_table;
+	struct ctl_table_header *head;
 	int error = 0;
 
-	/* No dups if we are the only member of our directory */
-	if (header->attached_by != table)
-		return 0;
-
-	dir_head = header->parent;
-	dir_table = header->attached_to;
-
-	error = sysctl_check_table_dups(path, dir_table, table);
-
 	root = &sysctl_table_root;
 	do {
 		set = lookup_header_set(root, namespaces);
@@ -871,9 +866,9 @@ static int sysctl_check_dups(struct nsproxy *namespaces,
 		list_for_each_entry(head, &set->list, ctl_entry) {
 			if (head->unregistering)
 				continue;
-			if (head->attached_to != dir_table)
+			if (head->parent != dir)
 				continue;
-			error = sysctl_check_table_dups(path, head->attached_by,
+			error = sysctl_check_table_dups(path, head->ctl_table,
 							table);
 		}
 		root = list_entry(root->root_list.next,
@@ -977,47 +972,25 @@ struct ctl_table_header *__register_sysctl_table(
 	const char *path, struct ctl_table *table)
 {
 	struct ctl_table_header *header;
-	struct ctl_table *new, **prevp;
 	const char *name, *nextname;
-	unsigned int npath = 0;
 	struct ctl_table_set *set;
-	size_t path_bytes = 0;
-	char *new_name;
-
-	/* Count the path components */
-	for (name = path; name; name = nextname) {
-		int namelen;
-		nextname = strchr(name, '/');
-		if (nextname) {
-			namelen = nextname - name;
-			nextname++;
-		} else {
-			namelen = strlen(name);
-		}
-		if (namelen == 0)
-			continue;
-		path_bytes += namelen + 1;
-		npath++;
-	}
+	struct ctl_dir *dir;
 
-	/*
-	 * For each path component, allocate a 2-element ctl_table array.
-	 * The first array element will be filled with the sysctl entry
-	 * for this, the second will be the sentinel (procname == 0).
-	 *
-	 * We allocate everything in one go so that we don't have to
-	 * worry about freeing additional memory in unregister_sysctl_table.
-	 */
-	header = kzalloc(sizeof(struct ctl_table_header) + path_bytes +
-			 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
+	header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
 	if (!header)
 		return NULL;
 
-	new = (struct ctl_table *) (header + 1);
-	new_name = (char *)(new + (2 * npath));
+	init_header(header, root, NULL, table);
+	if (sysctl_check_table(path, table))
+		goto fail;
+
+	spin_lock(&sysctl_lock);
+	header->set = set = lookup_header_set(root, namespaces);
+	dir = &sysctl_root_dir;
+	dir->header.nreg++;
+	spin_unlock(&sysctl_lock);
 
-	/* Now connect the dots */
-	prevp = &header->ctl_table;
+	/* Find the directory for the ctl_table */
 	for (name = path; name; name = nextname) {
 		int namelen;
 		nextname = strchr(name, '/');
@@ -1029,51 +1002,21 @@ struct ctl_table_header *__register_sysctl_table(
 		}
 		if (namelen == 0)
 			continue;
-		memcpy(new_name, name, namelen);
-		new_name[namelen] = '\0';
-
-		new->procname = new_name;
-		new->mode     = 0555;
-
-		*prevp = new;
-		prevp = &new->child;
 
-		new += 2;
-		new_name += namelen + 1;
+		dir = get_subdir(set, dir, name, namelen);
+		if (IS_ERR(dir))
+			goto fail;
 	}
-	*prevp = table;
-
-	init_header(header, root, NULL, table);
-	if (sysctl_check_table(path, table))
-		goto fail;
-
 	spin_lock(&sysctl_lock);
-	header->set = lookup_header_set(root, namespaces);
-	header->attached_by = header->ctl_table;
-	header->attached_to = &root_table[1];
-	header->parent = &root_table_header;
-	set = header->set;
-	root = header->root;
-	for (;;) {
-		struct ctl_table_header *p;
-		list_for_each_entry(p, &set->list, ctl_entry) {
-			if (p->unregistering)
-				continue;
-			try_attach(p, header);
-		}
-		if (root == &sysctl_table_root)
-			break;
-		root = list_entry(root->root_list.prev,
-				  struct ctl_table_root, root_list);
-		set = lookup_header_set(root, namespaces);
-	}
-	if (sysctl_check_dups(namespaces, header, path, table))
-		goto fail_locked;
-	insert_header(header);
+	if (sysctl_check_dups(namespaces, dir, path, table))
+		goto fail_put_dir_locked;
+	insert_header(dir, header);
+	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 
 	return header;
-fail_locked:
+fail_put_dir_locked:
+	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 fail:
 	kfree(header);
@@ -1299,16 +1242,17 @@ EXPORT_SYMBOL(register_sysctl_table);
 
 static void drop_sysctl_table(struct ctl_table_header *header)
 {
+	struct ctl_dir *parent = header->parent;
+
 	if (--header->nreg)
 		return;
 
 	start_unregistering(header);
-	if (!--header->parent->count) {
-		WARN_ON(1);
-		kfree_rcu(header->parent, rcu);
-	}
 	if (!--header->count)
 		kfree_rcu(header, rcu);
+
+	if (parent)
+		drop_sysctl_table(&parent->header);
 }
 
 /**
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index e73ba33..3084b62 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -938,6 +938,7 @@ struct ctl_table;
 struct nsproxy;
 struct ctl_table_root;
 struct ctl_table_header;
+struct ctl_dir;
 
 typedef struct ctl_table ctl_table;
 
@@ -1040,9 +1041,12 @@ struct ctl_table_header
 	struct ctl_table *ctl_table_arg;
 	struct ctl_table_root *root;
 	struct ctl_table_set *set;
-	struct ctl_table *attached_by;
-	struct ctl_table *attached_to;
-	struct ctl_table_header *parent;
+	struct ctl_dir *parent;
+};
+
+struct ctl_dir {
+	/* Header must be at the start of ctl_dir */
+	struct ctl_table_header header;
 };
 
 struct ctl_table_set {
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (50 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
@ 2012-01-27  4:51 ` Eric W. Biederman
  2012-01-27  4:52 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
                   ` (6 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:51 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

When there are errors it is very nice to know the full sysctl path.
Add a simple function that computes the sysctl path and prints it
out.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 3c0767d..a785565 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -53,6 +53,13 @@ static DEFINE_SPINLOCK(sysctl_lock);
 
 static void drop_sysctl_table(struct ctl_table_header *header);
 
+static void sysctl_print_dir(struct ctl_dir *dir)
+{
+	if (dir->header.parent)
+		sysctl_print_dir(dir->header.parent);
+	printk(KERN_CONT "%s/", dir->header.ctl_table[0].procname);
+}
+
 static int namecmp(const char *name1, int len1, const char *name2, int len2)
 {
 	int minlen;
@@ -822,7 +829,9 @@ found:
 	subdir->header.nreg++;
 failed:
 	if (unlikely(IS_ERR(subdir))) {
-		printk(KERN_ERR "sysctl could not get directory: %*.*s %ld\n",
+		printk(KERN_ERR "sysctl could not get directory: ");
+		sysctl_print_dir(dir);
+		printk(KERN_CONT "/%*.*s %ld\n",
 			namelen, namelen, name, PTR_ERR(subdir));
 	}
 	drop_sysctl_table(&dir->header);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (51 preceding siblings ...)
  2012-01-27  4:51 ` [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir Eric W. Biederman
@ 2012-01-27  4:52 ` Eric W. Biederman
  2012-01-30  0:01   ` Lucian Adrian Grijincu
  2012-01-30 17:51   ` Dave Jones
  2012-01-27  4:52 ` [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy Eric W. Biederman
                   ` (5 subsequent siblings)
  58 siblings, 2 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:52 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Piecing together directories by looking first in one directory
tree, than in another directory tree and finally in a third
directory tree makes it hard to verify that some directory
entries are not multiply defined and makes it hard to create
efficient implementations the sysctl filesystem.

Replace the sysctl wide list of roots with autogenerated
links from the core sysctl directory tree to the other
sysctl directory trees.

This simplifies sysctl directory reading and lookups as now
only entries in a single sysctl directory tree need to be
considered.

Benchmark before:
    make-dummies 0 999 -> 0.44s
    rmmod dummy        -> 0.065s
    make-dummies 0 9999 -> 1m36s
    rmmod dummy         -> 0.4s

Benchmark after:
    make-dummies 0 999 -> 0.63s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 2m35s
    rmmod dummy         -> 18s

The slowdown is caused by the lookups used in insert_headers
and put_links to see if we need to add links or remove links.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  397 +++++++++++++++++++++++++++++++++++------------
 include/linux/sysctl.h |    3 +-
 2 files changed, 296 insertions(+), 104 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index a785565..ec54a57 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -32,26 +32,26 @@ static struct ctl_table root_table[] = {
 	},
 	{ }
 };
-static struct ctl_table_root sysctl_table_root;
-static struct ctl_dir sysctl_root_dir = {
-	.header = {
+static struct ctl_table_root sysctl_table_root = {
+	.default_set.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),
+	.default_set.dir.header = {
 		{{.count = 1,
 		  .nreg = 1,
 		  .ctl_table = root_table,
 		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+		.ctl_table_arg = root_table,
 		.root = &sysctl_table_root,
 		.set = &sysctl_table_root.default_set,
 	},
 };
-static struct ctl_table_root sysctl_table_root = {
-	.root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
-	.default_set.list = LIST_HEAD_INIT(sysctl_root_dir.header.ctl_entry),
-	.default_set.root = &sysctl_table_root,
-};
 
 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, struct nsproxy *namespaces);
+static int insert_links(struct ctl_table_header *head);
+static void put_links(struct ctl_table_header *header);
 
 static void sysctl_print_dir(struct ctl_dir *dir)
 {
@@ -76,9 +76,9 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 }
 
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
-	struct ctl_table_set *set, struct ctl_dir *dir,
-	const char *name, int namelen)
+	struct ctl_dir *dir, const char *name, int namelen)
 {
+	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
@@ -119,11 +119,21 @@ static void erase_header(struct ctl_table_header *head)
 	list_del_init(&head->ctl_entry);
 }
 
-static void insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
+static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
+	int err;
+
+	dir->header.nreg++;
 	header->parent = dir;
-	header->parent->header.nreg++;
+	err = insert_links(header);
+	if (err)
+		goto fail_links;
 	list_add_tail(&header->ctl_entry, &header->set->list);
+	return 0;
+fail_links:
+	header->parent = NULL;
+	drop_sysctl_table(&dir->header);
+	return err;
 }
 
 /* called under sysctl_lock */
@@ -212,72 +222,39 @@ lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
 	return set;
 }
 
-static struct list_head *
-lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
-{
-	struct ctl_table_set *set = lookup_header_set(root, namespaces);
-	return &set->list;
-}
-
 static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 				      struct ctl_dir *dir,
 				      const char *name, int namelen)
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
-	struct ctl_table_root *root;
-	struct ctl_table_set *set;
 
 	spin_lock(&sysctl_lock);
-	root = &sysctl_table_root;
-	do {
-		set = lookup_header_set(root, current->nsproxy);
-		entry = find_entry(&head, set, dir, name, namelen);
-		if (entry && use_table(head))
-			*phead = head;
-		else
-			entry = NULL;
-		root = list_entry(root->root_list.next,
-				  struct ctl_table_root, root_list);
-	} while (!entry && root != &sysctl_table_root);
+	entry = find_entry(&head, dir, name, namelen);
+	if (entry && use_table(head))
+		*phead = head;
+	else
+		entry = NULL;
 	spin_unlock(&sysctl_lock);
 	return entry;
 }
 
 static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
-	struct ctl_table_root *root, struct list_head *tmp)
+						  struct list_head *tmp)
 {
-	struct nsproxy *namespaces = current->nsproxy;
-	struct list_head *header_list;
+	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 
-	goto next;
-	for (;;) {
+	for (tmp = tmp->next; tmp != &set->list; tmp = tmp->next) {
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
-		root = head->root;
 
 		if (head->parent != dir ||
 		    !head->ctl_table->procname ||
 		    !use_table(head))
-			goto next;
-
-		return head;
-	next:
-		tmp = tmp->next;
-		header_list = lookup_header_list(root, namespaces);
-		if (tmp != header_list)
 			continue;
 
-		do {
-			root = list_entry(root->root_list.next,
-					struct ctl_table_root, root_list);
-			if (root == &sysctl_table_root)
-				goto out;
-			header_list = lookup_header_list(root, namespaces);
-		} while (list_empty(header_list));
-		tmp = header_list->next;
+		return head;
 	}
-out:
 	return NULL;
 }
 
@@ -288,8 +265,7 @@ static void first_entry(struct ctl_dir *dir,
 	struct ctl_table *entry = NULL;
 
 	spin_lock(&sysctl_lock);
-	head = next_usable_entry(dir, &sysctl_table_root,
-				 &sysctl_table_root.default_set.list);
+	head = next_usable_entry(dir, &dir->header.set->list);
 	spin_unlock(&sysctl_lock);
 	if (head)
 		entry = head->ctl_table;
@@ -306,7 +282,7 @@ static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentr
 	if (!entry->procname) {
 		spin_lock(&sysctl_lock);
 		unuse_table(head);
-		head = next_usable_entry(head->parent, head->root, &head->ctl_entry);
+		head = next_usable_entry(head->parent, &head->ctl_entry);
 		spin_unlock(&sysctl_lock);
 		if (head)
 			entry = head->ctl_table;
@@ -317,9 +293,6 @@ static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentr
 
 void register_sysctl_root(struct ctl_table_root *root)
 {
-	spin_lock(&sysctl_lock);
-	list_add_tail(&root->root_list, &sysctl_table_root.root_list);
-	spin_unlock(&sysctl_lock);
 }
 
 /*
@@ -386,7 +359,7 @@ static struct ctl_table_header *grab_header(struct inode *inode)
 {
 	struct ctl_table_header *head = PROC_I(inode)->sysctl;
 	if (!head)
-		head = &sysctl_root_dir.header;
+		head = &sysctl_table_root.default_set.dir.header;
 	return sysctl_head_grab(head);
 }
 
@@ -400,6 +373,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 	struct inode *inode;
 	struct dentry *err = ERR_PTR(-ENOENT);
 	struct ctl_dir *ctl_dir;
+	int ret;
 
 	if (IS_ERR(head))
 		return ERR_CAST(head);
@@ -410,6 +384,11 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 	if (!p)
 		goto out;
 
+	ret = sysctl_follow_link(&h, &p, current->nsproxy);
+	err = ERR_PTR(ret);
+	if (ret)
+		goto out;
+
 	err = ERR_PTR(-ENOMEM);
 	inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
 	if (h)
@@ -547,6 +526,25 @@ static int proc_sys_fill_cache(struct file *filp, void *dirent,
 	return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
 }
 
+static int proc_sys_link_fill_cache(struct file *filp, void *dirent,
+				    filldir_t filldir,
+				    struct ctl_table_header *head,
+				    struct ctl_table *table)
+{
+	int err, ret = 0;
+	head = sysctl_head_grab(head);
+
+	/* It is not an error if we can not follow the link ignore it */
+	err = sysctl_follow_link(&head, &table, current->nsproxy);
+	if (err)
+		goto out;
+
+	ret = proc_sys_fill_cache(filp, dirent, filldir, head, table);
+out:
+	sysctl_head_finish(head);
+	return ret;
+}
+
 static int scan(struct ctl_table_header *head, ctl_table *table,
 		unsigned long *pos, struct file *file,
 		void *dirent, filldir_t filldir)
@@ -556,7 +554,10 @@ static int scan(struct ctl_table_header *head, ctl_table *table,
 	if ((*pos)++ < file->f_pos)
 		return 0;
 
-	res = proc_sys_fill_cache(file, dirent, filldir, head, table);
+	if (unlikely(S_ISLNK(table->mode)))
+		res = proc_sys_link_fill_cache(file, dirent, filldir, head, table);
+	else
+		res = proc_sys_fill_cache(file, dirent, filldir, head, table);
 
 	if (res == 0)
 		file->f_pos = *pos;
@@ -757,13 +758,13 @@ static const struct dentry_operations proc_sys_dentry_operations = {
 	.d_compare	= proc_sys_compare,
 };
 
-static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *dir,
-	const char *name, int namelen)
+static struct ctl_dir *find_subdir(struct ctl_dir *dir,
+				   const char *name, int namelen)
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
-	entry = find_entry(&head, set, dir, name, namelen);
+	entry = find_entry(&head, dir, name, namelen);
 	if (!entry)
 		return ERR_PTR(-ENOENT);
 	if (S_ISDIR(entry->mode))
@@ -772,7 +773,7 @@ static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *di
 }
 
 static struct ctl_dir *new_dir(struct ctl_table_set *set,
-	const char *name, int namelen)
+			       const char *name, int namelen)
 {
 	struct ctl_table *table;
 	struct ctl_dir *new;
@@ -789,22 +790,19 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 	new_name[namelen] = '\0';
 	table[0].procname = new_name;
 	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
-	init_header(&new->header, set->root, set, table);
+	init_header(&new->header, set->dir.header.root, set, table);
 
 	return new;
 }
 
-static struct ctl_dir *get_subdir(struct ctl_table_set *set,
-	struct ctl_dir *dir, const char *name, int namelen)
+static struct ctl_dir *get_subdir(struct ctl_dir *dir,
+				  const char *name, int namelen)
 {
+	struct ctl_table_set *set = dir->header.set;
 	struct ctl_dir *subdir, *new = NULL;
 
 	spin_lock(&sysctl_lock);
-	subdir = find_subdir(dir->header.set, dir, name, namelen);
-	if (!IS_ERR(subdir))
-		goto found;
-	if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)
-		subdir = find_subdir(set, dir, name, namelen);
+	subdir = find_subdir(dir, name, namelen);
 	if (!IS_ERR(subdir))
 		goto found;
 	if (PTR_ERR(subdir) != -ENOENT)
@@ -817,13 +815,14 @@ static struct ctl_dir *get_subdir(struct ctl_table_set *set,
 	if (!new)
 		goto failed;
 
-	subdir = find_subdir(set, dir, name, namelen);
+	subdir = find_subdir(dir, name, namelen);
 	if (!IS_ERR(subdir))
 		goto found;
 	if (PTR_ERR(subdir) != -ENOENT)
 		goto failed;
 
-	insert_header(dir, &new->header);
+	if (insert_header(dir, &new->header))
+		goto failed;
 	subdir = new;
 found:
 	subdir->header.nreg++;
@@ -841,6 +840,57 @@ failed:
 	return subdir;
 }
 
+static struct ctl_dir *xlate_dir(struct ctl_table_set *set, struct ctl_dir *dir)
+{
+	struct ctl_dir *parent;
+	const char *procname;
+	if (!dir->header.parent)
+		return &set->dir;
+	parent = xlate_dir(set, dir->header.parent);
+	if (IS_ERR(parent))
+		return parent;
+	procname = dir->header.ctl_table[0].procname;
+	return find_subdir(parent, procname, strlen(procname));
+}
+
+static int sysctl_follow_link(struct ctl_table_header **phead,
+	struct ctl_table **pentry, struct nsproxy *namespaces)
+{
+	struct ctl_table_header *head;
+	struct ctl_table_root *root;
+	struct ctl_table_set *set;
+	struct ctl_table *entry;
+	struct ctl_dir *dir;
+	int ret;
+
+	/* Get out quickly if not a link */
+	if (!S_ISLNK((*pentry)->mode))
+		return 0;
+
+	ret = 0;
+	spin_lock(&sysctl_lock);
+	root = (*pentry)->data;
+	set = lookup_header_set(root, namespaces);
+	dir = xlate_dir(set, (*phead)->parent);
+	if (IS_ERR(dir))
+		ret = PTR_ERR(dir);
+	else {
+		const char *procname = (*pentry)->procname;
+		head = NULL;
+		entry = find_entry(&head, dir, procname, strlen(procname));
+		ret = -ENOENT;
+		if (entry && use_table(head)) {
+			unuse_table(*phead);
+			*phead = head;
+			*pentry = entry;
+			ret = 0;
+		}
+	}
+
+	spin_unlock(&sysctl_lock);
+	return ret;
+}
+
 static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	struct ctl_table *table)
 {
@@ -859,30 +909,21 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	return error;
 }
 
-static int sysctl_check_dups(struct nsproxy *namespaces,
-	struct ctl_dir *dir,
+static int sysctl_check_dups(struct ctl_dir *dir,
 	const char *path, struct ctl_table *table)
 {
-	struct ctl_table_root *root;
 	struct ctl_table_set *set;
 	struct ctl_table_header *head;
 	int error = 0;
 
-	root = &sysctl_table_root;
-	do {
-		set = lookup_header_set(root, namespaces);
-
-		list_for_each_entry(head, &set->list, ctl_entry) {
-			if (head->unregistering)
-				continue;
-			if (head->parent != dir)
-				continue;
-			error = sysctl_check_table_dups(path, head->ctl_table,
-							table);
-		}
-		root = list_entry(root->root_list.next,
-				  struct ctl_table_root, root_list);
-	} while (root != &sysctl_table_root);
+	set = dir->header.set;
+	list_for_each_entry(head, &set->list, ctl_entry) {
+		if (head->unregistering)
+			continue;
+		if (head->parent != dir)
+			continue;
+		error = sysctl_check_table_dups(path, head->ctl_table, table);
+	}
 	return error;
 }
 
@@ -932,6 +973,115 @@ static int sysctl_check_table(const char *path, struct ctl_table *table)
 	return err;
 }
 
+static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table *table,
+	struct ctl_table_root *link_root)
+{
+	struct ctl_table *link_table, *entry, *link;
+	struct ctl_table_header *links;
+	char *link_name;
+	int nr_entries, name_bytes;
+
+	name_bytes = 0;
+	nr_entries = 0;
+	for (entry = table; entry->procname; entry++) {
+		nr_entries++;
+		name_bytes += strlen(entry->procname) + 1;
+	}
+
+	links = kzalloc(sizeof(struct ctl_table_header) +
+			sizeof(struct ctl_table)*(nr_entries + 1) +
+			name_bytes,
+			GFP_KERNEL);
+
+	if (!links)
+		return NULL;
+
+	link_table = (struct ctl_table *)(links + 1);
+	link_name = (char *)&link_table[nr_entries + 1];
+
+	for (link = link_table, entry = table; entry->procname; link++, entry++) {
+		int len = strlen(entry->procname) + 1;
+		memcpy(link_name, entry->procname, len);
+		link->procname = link_name;
+		link->mode = S_IFLNK|S_IRWXUGO;
+		link->data = link_root;
+		link_name += len;
+	}
+	init_header(links, dir->header.root, dir->header.set, link_table);
+	links->nreg = nr_entries;
+
+	return links;
+}
+
+static bool get_links(struct ctl_dir *dir,
+	struct ctl_table *table, struct ctl_table_root *link_root)
+{
+	struct ctl_table_header *head;
+	struct ctl_table *entry, *link;
+
+	/* Are there links available for every entry in table? */
+	for (entry = table; entry->procname; entry++) {
+		const char *procname = entry->procname;
+		link = find_entry(&head, dir, procname, strlen(procname));
+		if (!link)
+			return false;
+		if (S_ISDIR(link->mode) && S_ISDIR(entry->mode))
+			continue;
+		if (S_ISLNK(link->mode) && (link->data == link_root))
+			continue;
+		return false;
+	}
+
+	/* The checks passed.  Increase the registration count on the links */
+	for (entry = table; entry->procname; entry++) {
+		const char *procname = entry->procname;
+		link = find_entry(&head, dir, procname, strlen(procname));
+		head->nreg++;
+	}
+	return true;
+}
+
+static int insert_links(struct ctl_table_header *head)
+{
+	struct ctl_table_set *root_set = &sysctl_table_root.default_set;
+	struct ctl_dir *core_parent = NULL;
+	struct ctl_table_header *links;
+	int err;
+
+	if (head->set == root_set)
+		return 0;
+
+	core_parent = xlate_dir(root_set, head->parent);
+	if (IS_ERR(core_parent))
+		return 0;
+
+	if (get_links(core_parent, head->ctl_table, head->root))
+		return 0;
+
+	core_parent->header.nreg++;
+	spin_unlock(&sysctl_lock);
+
+	links = new_links(core_parent, head->ctl_table, head->root);
+
+	spin_lock(&sysctl_lock);
+	err = -ENOMEM;
+	if (!links)
+		goto out;
+
+	err = 0;
+	if (get_links(core_parent, head->ctl_table, head->root)) {
+		kfree(links);
+		goto out;
+	}
+
+	err = insert_header(core_parent, links);
+	if (err)
+		kfree(links);
+out:
+	drop_sysctl_table(&core_parent->header);
+	return err;
+}
+
 /**
  * __register_sysctl_table - register a leaf sysctl table
  * @root: List of sysctl headers to register on
@@ -980,6 +1130,7 @@ struct ctl_table_header *__register_sysctl_table(
 	struct nsproxy *namespaces,
 	const char *path, struct ctl_table *table)
 {
+	struct ctl_table_header *links = NULL;
 	struct ctl_table_header *header;
 	const char *name, *nextname;
 	struct ctl_table_set *set;
@@ -995,7 +1146,7 @@ struct ctl_table_header *__register_sysctl_table(
 
 	spin_lock(&sysctl_lock);
 	header->set = set = lookup_header_set(root, namespaces);
-	dir = &sysctl_root_dir;
+	dir = &set->dir;
 	dir->header.nreg++;
 	spin_unlock(&sysctl_lock);
 
@@ -1012,22 +1163,28 @@ struct ctl_table_header *__register_sysctl_table(
 		if (namelen == 0)
 			continue;
 
-		dir = get_subdir(set, dir, name, namelen);
+		dir = get_subdir(dir, name, namelen);
 		if (IS_ERR(dir))
 			goto fail;
 	}
+
 	spin_lock(&sysctl_lock);
-	if (sysctl_check_dups(namespaces, dir, path, table))
+	if (sysctl_check_dups(dir, path, table))
+		goto fail_put_dir_locked;
+
+	if (insert_header(dir, header))
 		goto fail_put_dir_locked;
-	insert_header(dir, header);
+
 	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 
 	return header;
+
 fail_put_dir_locked:
 	drop_sysctl_table(&dir->header);
 	spin_unlock(&sysctl_lock);
 fail:
+	kfree(links);
 	kfree(header);
 	dump_stack();
 	return NULL;
@@ -1249,6 +1406,40 @@ struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
 }
 EXPORT_SYMBOL(register_sysctl_table);
 
+static void put_links(struct ctl_table_header *header)
+{
+	struct ctl_table_set *root_set = &sysctl_table_root.default_set;
+	struct ctl_table_root *root = header->root;
+	struct ctl_dir *parent = header->parent;
+	struct ctl_dir *core_parent;
+	struct ctl_table *entry;
+
+	if (header->set == root_set)
+		return;
+
+	core_parent = xlate_dir(root_set, parent);
+	if (IS_ERR(core_parent))
+		return;
+
+	for (entry = header->ctl_table; entry->procname; entry++) {
+		struct ctl_table_header *link_head;
+		struct ctl_table *link;
+		const char *name = entry->procname;
+
+		link = find_entry(&link_head, core_parent, name, strlen(name));
+		if (link &&
+		    ((S_ISDIR(link->mode) && S_ISDIR(entry->mode)) ||
+		     (S_ISLNK(link->mode) && (link->data == root)))) {
+			drop_sysctl_table(link_head);
+		}
+		else {
+			printk(KERN_ERR "sysctl link missing during unregister: ");
+			sysctl_print_dir(parent);
+			printk(KERN_CONT "/%s\n", name);
+		}
+	}
+}
+
 static void drop_sysctl_table(struct ctl_table_header *header)
 {
 	struct ctl_dir *parent = header->parent;
@@ -1256,6 +1447,7 @@ static void drop_sysctl_table(struct ctl_table_header *header)
 	if (--header->nreg)
 		return;
 
+	put_links(header);
 	start_unregistering(header);
 	if (!--header->count)
 		kfree_rcu(header, rcu);
@@ -1301,13 +1493,14 @@ void unregister_sysctl_table(struct ctl_table_header * header)
 }
 EXPORT_SYMBOL(unregister_sysctl_table);
 
-void setup_sysctl_set(struct ctl_table_set *p,
+void setup_sysctl_set(struct ctl_table_set *set,
 	struct ctl_table_root *root,
 	int (*is_seen)(struct ctl_table_set *))
 {
-	INIT_LIST_HEAD(&p->list);
-	p->root = root;
-	p->is_seen = is_seen;
+	memset(set, sizeof(*set), 0);
+	INIT_LIST_HEAD(&set->list);
+	set->is_seen = is_seen;
+	init_header(&set->dir.header, root, set, root_table);
 }
 
 void retire_sysctl_set(struct ctl_table_set *set)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 3084b62..2a1446a 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1051,12 +1051,11 @@ struct ctl_dir {
 
 struct ctl_table_set {
 	struct list_head list;
-	struct ctl_table_root *root;
 	int (*is_seen)(struct ctl_table_set *);
+	struct ctl_dir dir;
 };
 
 struct ctl_table_root {
-	struct list_head root_list;
 	struct ctl_table_set default_set;
 	struct ctl_table_set *(*lookup)(struct ctl_table_root *root,
 					   struct nsproxy *namespaces);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (52 preceding siblings ...)
  2012-01-27  4:52 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
@ 2012-01-27  4:52 ` Eric W. Biederman
  2012-01-27  4:52 ` [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header Eric W. Biederman
                   ` (4 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:52 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

An nsproxy argument here has always been awkard and now the nsproxy argument
is completely unnecessary so remove it, replacing it with the set we want
the registered tables to show up in.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   30 ++++++++++++------------------
 include/linux/sysctl.h |    4 ++--
 net/sysctl_net.c       |   10 +++-------
 3 files changed, 17 insertions(+), 27 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index ec54a57..e0d3e7e 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1084,8 +1084,7 @@ out:
 
 /**
  * __register_sysctl_table - register a leaf sysctl table
- * @root: List of sysctl headers to register on
- * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @set: Sysctl tree to register on
  * @path: The path to the directory the sysctl table is in.
  * @table: the top-level table structure
  *
@@ -1126,26 +1125,24 @@ out:
  * to the table header on success.
  */
 struct ctl_table_header *__register_sysctl_table(
-	struct ctl_table_root *root,
-	struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const char *path, struct ctl_table *table)
 {
+	struct ctl_table_root *root = set->dir.header.root;
 	struct ctl_table_header *links = NULL;
 	struct ctl_table_header *header;
 	const char *name, *nextname;
-	struct ctl_table_set *set;
 	struct ctl_dir *dir;
 
 	header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
 	if (!header)
 		return NULL;
 
-	init_header(header, root, NULL, table);
+	init_header(header, root, set, table);
 	if (sysctl_check_table(path, table))
 		goto fail;
 
 	spin_lock(&sysctl_lock);
-	header->set = set = lookup_header_set(root, namespaces);
 	dir = &set->dir;
 	dir->header.nreg++;
 	spin_unlock(&sysctl_lock);
@@ -1223,8 +1220,7 @@ static int count_subheaders(struct ctl_table *table)
 }
 
 static int register_leaf_sysctl_tables(const char *path, char *pos,
-	struct ctl_table_header ***subheader,
-	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table_header ***subheader, struct ctl_table_set *set,
 	struct ctl_table *table)
 {
 	struct ctl_table *ctl_table_arg = NULL;
@@ -1261,7 +1257,7 @@ static int register_leaf_sysctl_tables(const char *path, char *pos,
 	/* Register everything except a directory full of subdirectories */
 	if (nr_files || !nr_dirs) {
 		struct ctl_table_header *header;
-		header = __register_sysctl_table(root, namespaces, path, files);
+		header = __register_sysctl_table(set, path, files);
 		if (!header) {
 			kfree(ctl_table_arg);
 			goto out;
@@ -1286,7 +1282,7 @@ static int register_leaf_sysctl_tables(const char *path, char *pos,
 			goto out;
 
 		err = register_leaf_sysctl_tables(path, child_pos, subheader,
-						  root, namespaces, entry->child);
+						  set, entry->child);
 		pos[0] = '\0';
 		if (err)
 			goto out;
@@ -1299,8 +1295,7 @@ out:
 
 /**
  * __register_sysctl_paths - register a sysctl table hierarchy
- * @root: List of sysctl headers to register on
- * @namespaces: Data to compute which lists of sysctl entries are visible
+ * @set: Sysctl tree to register on
  * @path: The path to the directory the sysctl table is in.
  * @table: the top-level table structure
  *
@@ -1310,8 +1305,7 @@ out:
  * See __register_sysctl_table for more details.
  */
 struct ctl_table_header *__register_sysctl_paths(
-	struct ctl_table_root *root,
-	struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const struct ctl_path *path, struct ctl_table *table)
 {
 	struct ctl_table *ctl_table_arg = table;
@@ -1337,7 +1331,7 @@ struct ctl_table_header *__register_sysctl_paths(
 		table = table->child;
 	}
 	if (nr_subheaders == 1) {
-		header = __register_sysctl_table(root, namespaces, new_path, table);
+		header = __register_sysctl_table(set, new_path, table);
 		if (header)
 			header->ctl_table_arg = ctl_table_arg;
 	} else {
@@ -1351,7 +1345,7 @@ struct ctl_table_header *__register_sysctl_paths(
 		header->ctl_table_arg = ctl_table_arg;
 
 		if (register_leaf_sysctl_tables(new_path, pos, &subheader,
-						root, namespaces, table))
+						set, table))
 			goto err_register_leaves;
 	}
 
@@ -1384,7 +1378,7 @@ err_register_leaves:
 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 						struct ctl_table *table)
 {
-	return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
+	return __register_sysctl_paths(&sysctl_table_root.default_set,
 					path, table);
 }
 EXPORT_SYMBOL(register_sysctl_paths);
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 2a1446a..cec5941 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1079,10 +1079,10 @@ extern void retire_sysctl_set(struct ctl_table_set *set);
 
 void register_sysctl_root(struct ctl_table_root *root);
 struct ctl_table_header *__register_sysctl_table(
-	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const char *path, struct ctl_table *table);
 struct ctl_table_header *__register_sysctl_paths(
-	struct ctl_table_root *root, struct nsproxy *namespaces,
+	struct ctl_table_set *set,
 	const struct ctl_path *path, struct ctl_table *table);
 struct ctl_table_header *register_sysctl_table(struct ctl_table * table);
 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
diff --git a/net/sysctl_net.c b/net/sysctl_net.c
index e998c64..c3e65ae 100644
--- a/net/sysctl_net.c
+++ b/net/sysctl_net.c
@@ -105,19 +105,15 @@ subsys_initcall(net_sysctl_init);
 struct ctl_table_header *register_net_sysctl_table(struct net *net,
 	const struct ctl_path *path, struct ctl_table *table)
 {
-	struct nsproxy namespaces;
-	namespaces = *current->nsproxy;
-	namespaces.net_ns = net;
-	return __register_sysctl_paths(&net_sysctl_root,
-					&namespaces, path, table);
+	return __register_sysctl_paths(&net->sysctls, path, table);
 }
 EXPORT_SYMBOL_GPL(register_net_sysctl_table);
 
 struct ctl_table_header *register_net_sysctl_rotable(const
 		struct ctl_path *path, struct ctl_table *table)
 {
-	return __register_sysctl_paths(&net_sysctl_ro_root,
-			&init_nsproxy, path, table);
+	return __register_sysctl_paths(&net_sysctl_ro_root.default_set,
+					path, table);
 }
 EXPORT_SYMBOL_GPL(register_net_sysctl_rotable);
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (53 preceding siblings ...)
  2012-01-27  4:52 ` [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy Eric W. Biederman
@ 2012-01-27  4:52 ` Eric W. Biederman
  2012-01-27  4:52 ` [PATCH 27/29] sysctl: Make the header lists per directory Eric W. Biederman
                   ` (3 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:52 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Simplify the callers of insert_header by removing explicit calls to check
for duplicates and instead have insert_header do the work.

This makes the code slightly more maintainable by enabling changes to
data structures where the insertion of new entries without duplicate
suppression is not possible.

There is not always a convenient path string where insert_header
is called so modify sysctl_check_dups to use sysctl_print_dir
when printing the full path when a duplicate is discovered.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   20 +++++++++++---------
 1 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e0d3e7e..160d578 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -52,6 +52,7 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	struct ctl_table **pentry, struct nsproxy *namespaces);
 static int insert_links(struct ctl_table_header *head);
 static void put_links(struct ctl_table_header *header);
+static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table);
 
 static void sysctl_print_dir(struct ctl_dir *dir)
 {
@@ -123,6 +124,10 @@ static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
 	int err;
 
+	err = sysctl_check_dups(dir, header->ctl_table);
+	if (err)
+		return err;
+
 	dir->header.nreg++;
 	header->parent = dir;
 	err = insert_links(header);
@@ -891,7 +896,7 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	return ret;
 }
 
-static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
+static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
 	struct ctl_table *table)
 {
 	struct ctl_table *entry, *test;
@@ -900,8 +905,9 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	for (entry = old; entry->procname; entry++) {
 		for (test = table; test->procname; test++) {
 			if (strcmp(entry->procname, test->procname) == 0) {
-				printk(KERN_ERR "sysctl duplicate entry: %s/%s\n",
-					path, test->procname);
+				printk(KERN_ERR "sysctl duplicate entry: ");
+				sysctl_print_dir(dir);
+				printk(KERN_CONT "/%s\n", test->procname);
 				error = -EEXIST;
 			}
 		}
@@ -909,8 +915,7 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
 	return error;
 }
 
-static int sysctl_check_dups(struct ctl_dir *dir,
-	const char *path, struct ctl_table *table)
+static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
 {
 	struct ctl_table_set *set;
 	struct ctl_table_header *head;
@@ -922,7 +927,7 @@ static int sysctl_check_dups(struct ctl_dir *dir,
 			continue;
 		if (head->parent != dir)
 			continue;
-		error = sysctl_check_table_dups(path, head->ctl_table, table);
+		error = sysctl_check_table_dups(dir, head->ctl_table, table);
 	}
 	return error;
 }
@@ -1166,9 +1171,6 @@ struct ctl_table_header *__register_sysctl_table(
 	}
 
 	spin_lock(&sysctl_lock);
-	if (sysctl_check_dups(dir, path, table))
-		goto fail_put_dir_locked;
-
 	if (insert_header(dir, header))
 		goto fail_put_dir_locked;
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 27/29] sysctl: Make the header lists per directory.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (54 preceding siblings ...)
  2012-01-27  4:52 ` [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header Eric W. Biederman
@ 2012-01-27  4:52 ` Eric W. Biederman
  2012-01-27  4:52 ` [PATCH 28/29] sysctl: Index sysctl directories with rbtrees Eric W. Biederman
                   ` (2 subsequent siblings)
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:52 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

Slightly enhance efficiency and clarity of the code by making the
header list per directory instead of per set.

Benchmark before:
    make-dummies 0 999 -> 0.63s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 2m35s
    rmmod dummy         -> 18s

Benchmark after:
    make-dummies 0 999 -> 0.32s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 1m17s
    rmmod dummy         -> 17s

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   28 +++++++++++-----------------
 include/linux/sysctl.h |    2 +-
 2 files changed, 12 insertions(+), 18 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 160d578..e971ccc 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -33,12 +33,12 @@ static struct ctl_table root_table[] = {
 	{ }
 };
 static struct ctl_table_root sysctl_table_root = {
-	.default_set.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),
+	.default_set.dir.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.list),
 	.default_set.dir.header = {
 		{{.count = 1,
 		  .nreg = 1,
 		  .ctl_table = root_table,
-		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
+		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),}},
 		.ctl_table_arg = root_table,
 		.root = &sysctl_table_root,
 		.set = &sysctl_table_root.default_set,
@@ -79,15 +79,12 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
 	struct ctl_dir *dir, const char *name, int namelen)
 {
-	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
 
-	list_for_each_entry(head, &set->list, ctl_entry) {
+	list_for_each_entry(head, &dir->list, ctl_entry) {
 		if (head->unregistering)
 			continue;
-		if (head->parent != dir)
-			continue;
 		for (entry = head->ctl_table; entry->procname; entry++) {
 			const char *procname = entry->procname;
 			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
@@ -133,7 +130,7 @@ static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 	err = insert_links(header);
 	if (err)
 		goto fail_links;
-	list_add_tail(&header->ctl_entry, &header->set->list);
+	list_add_tail(&header->ctl_entry, &header->parent->list);
 	return 0;
 fail_links:
 	header->parent = NULL;
@@ -247,14 +244,12 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
 						  struct list_head *tmp)
 {
-	struct ctl_table_set *set = dir->header.set;
 	struct ctl_table_header *head;
 
-	for (tmp = tmp->next; tmp != &set->list; tmp = tmp->next) {
+	for (tmp = tmp->next; tmp != &dir->list; tmp = tmp->next) {
 		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
 
-		if (head->parent != dir ||
-		    !head->ctl_table->procname ||
+		if (!head->ctl_table->procname ||
 		    !use_table(head))
 			continue;
 
@@ -270,7 +265,7 @@ static void first_entry(struct ctl_dir *dir,
 	struct ctl_table *entry = NULL;
 
 	spin_lock(&sysctl_lock);
-	head = next_usable_entry(dir, &dir->header.set->list);
+	head = next_usable_entry(dir, &dir->list);
 	spin_unlock(&sysctl_lock);
 	if (head)
 		entry = head->ctl_table;
@@ -793,6 +788,7 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 	new_name = (char *)(table + 2);
 	memcpy(new_name, name, namelen);
 	new_name[namelen] = '\0';
+	INIT_LIST_HEAD(&new->list);
 	table[0].procname = new_name;
 	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
 	init_header(&new->header, set->dir.header.root, set, table);
@@ -917,12 +913,10 @@ static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
 
 static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
 {
-	struct ctl_table_set *set;
 	struct ctl_table_header *head;
 	int error = 0;
 
-	set = dir->header.set;
-	list_for_each_entry(head, &set->list, ctl_entry) {
+	list_for_each_entry(head, &dir->list, ctl_entry) {
 		if (head->unregistering)
 			continue;
 		if (head->parent != dir)
@@ -1494,14 +1488,14 @@ void setup_sysctl_set(struct ctl_table_set *set,
 	int (*is_seen)(struct ctl_table_set *))
 {
 	memset(set, sizeof(*set), 0);
-	INIT_LIST_HEAD(&set->list);
 	set->is_seen = is_seen;
+	INIT_LIST_HEAD(&set->dir.list);
 	init_header(&set->dir.header, root, set, root_table);
 }
 
 void retire_sysctl_set(struct ctl_table_set *set)
 {
-	WARN_ON(!list_empty(&set->list));
+	WARN_ON(!list_empty(&set->dir.list));
 }
 
 int __init proc_sys_init(void)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index cec5941..36dec75 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1047,10 +1047,10 @@ struct ctl_table_header
 struct ctl_dir {
 	/* Header must be at the start of ctl_dir */
 	struct ctl_table_header header;
+	struct list_head list;
 };
 
 struct ctl_table_set {
-	struct list_head list;
 	int (*is_seen)(struct ctl_table_set *);
 	struct ctl_dir dir;
 };
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 28/29] sysctl: Index sysctl directories with rbtrees.
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (55 preceding siblings ...)
  2012-01-27  4:52 ` [PATCH 27/29] sysctl: Make the header lists per directory Eric W. Biederman
@ 2012-01-27  4:52 ` Eric W. Biederman
  2012-01-27  4:52 ` [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users Eric W. Biederman
  2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:52 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

One of the most important jobs of sysctl is to export network stack
tunables.  Several of those tunables are per network device.  In
several instances people are running with 1000+ network devices in
there network stacks, which makes the simple per directory linked list
in sysctl a scaling bottleneck.   Replace O(N^2) sysctl insertion and
lookup times with O(NlogN) by using an rbtree to index the sysctl
directories.

Benchmark before:
    make-dummies 0 999 -> 0.32s
    rmmod dummy        -> 0.12s
    make-dummies 0 9999 -> 1m17s
    rmmod dummy         -> 17s

Benchmark after:
    make-dummies 0 999 -> 0.074s
    rmmod dummy        -> 0.070s
    make-dummies 0 9999 -> 3.4s
    rmmod dummy         -> 0.44s

Benchmark after (without dev_snmp6):
    make-dummies 0 9999 -> 0.75s
    rmmod dummy         -> 0.44s
    make-dummies 0 99999 -> 11s
    rmmod dummy          -> 4.3s

At 10,000 dummy devices the bottleneck becomes the time to add and
remove the files under /proc/sys/net/dev_snmp6.  I have commented
out the code that adds and removes files under /proc/sys/net/dev_snmp6
and taken measurments of creating and destroying 100,000 dummies to
verify the sysctl continues to scale.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |  224 +++++++++++++++++++++++++++++-------------------
 include/linux/sysctl.h |   10 ++-
 2 files changed, 142 insertions(+), 92 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e971ccc..05c393a 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -33,12 +33,10 @@ static struct ctl_table root_table[] = {
 	{ }
 };
 static struct ctl_table_root sysctl_table_root = {
-	.default_set.dir.list = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.list),
 	.default_set.dir.header = {
 		{{.count = 1,
 		  .nreg = 1,
-		  .ctl_table = root_table,
-		  .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.dir.header.ctl_entry),}},
+		  .ctl_table = root_table }},
 		.ctl_table_arg = root_table,
 		.root = &sysctl_table_root,
 		.set = &sysctl_table_root.default_set,
@@ -52,7 +50,6 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	struct ctl_table **pentry, struct nsproxy *namespaces);
 static int insert_links(struct ctl_table_header *head);
 static void put_links(struct ctl_table_header *header);
-static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table);
 
 static void sysctl_print_dir(struct ctl_dir *dir)
 {
@@ -81,28 +78,83 @@ static struct ctl_table *find_entry(struct ctl_table_header **phead,
 {
 	struct ctl_table_header *head;
 	struct ctl_table *entry;
+	struct rb_node *node = dir->root.rb_node;
 
-	list_for_each_entry(head, &dir->list, ctl_entry) {
-		if (head->unregistering)
-			continue;
-		for (entry = head->ctl_table; entry->procname; entry++) {
-			const char *procname = entry->procname;
-			if (namecmp(procname, strlen(procname), name, namelen) == 0) {
-				*phead = head;
-				return entry;
-			}
+	while (node)
+	{
+		struct ctl_node *ctl_node;
+		const char *procname;
+		int cmp;
+
+		ctl_node = rb_entry(node, struct ctl_node, node);
+		head = ctl_node->header;
+		entry = &head->ctl_table[ctl_node - head->node];
+		procname = entry->procname;
+
+		cmp = namecmp(name, namelen, procname, strlen(procname));
+		if (cmp < 0)
+			node = node->rb_left;
+		else if (cmp > 0)
+			node = node->rb_right;
+		else {
+			*phead = head;
+			return entry;
 		}
 	}
 	return NULL;
 }
 
+static int insert_entry(struct ctl_table_header *head, struct ctl_table *entry)
+{
+	struct rb_node *node = &head->node[entry - head->ctl_table].node;
+	struct rb_node **p = &head->parent->root.rb_node;
+	struct rb_node *parent = NULL;
+	const char *name = entry->procname;
+	int namelen = strlen(name);
+
+	while (*p) {
+		struct ctl_table_header *parent_head;
+		struct ctl_table *parent_entry;
+		struct ctl_node *parent_node;
+		const char *parent_name;
+		int cmp;
+
+		parent = *p;
+		parent_node = rb_entry(parent, struct ctl_node, node);
+		parent_head = parent_node->header;
+		parent_entry = &parent_head->ctl_table[parent_node - parent_head->node];
+		parent_name = parent_entry->procname;
+
+		cmp = namecmp(name, namelen, parent_name, strlen(parent_name));
+		if (cmp < 0)
+			p = &(*p)->rb_left;
+		else if (cmp > 0)
+			p = &(*p)->rb_right;
+		else {
+			printk(KERN_ERR "sysctl duplicate entry: ");
+			sysctl_print_dir(head->parent);
+			printk(KERN_CONT "/%s\n", entry->procname);
+			return -EEXIST;
+		}
+	}
+
+	rb_link_node(node, parent, p);
+	return 0;
+}
+
+static void erase_entry(struct ctl_table_header *head, struct ctl_table *entry)
+{
+	struct rb_node *node = &head->node[entry - head->ctl_table].node;
+
+	rb_erase(node, &head->parent->root);
+}
+
 static void init_header(struct ctl_table_header *head,
 	struct ctl_table_root *root, struct ctl_table_set *set,
-	struct ctl_table *table)
+	struct ctl_node *node, struct ctl_table *table)
 {
 	head->ctl_table = table;
 	head->ctl_table_arg = table;
-	INIT_LIST_HEAD(&head->ctl_entry);
 	head->used = 0;
 	head->count = 1;
 	head->nreg = 1;
@@ -110,28 +162,42 @@ static void init_header(struct ctl_table_header *head,
 	head->root = root;
 	head->set = set;
 	head->parent = NULL;
+	head->node = node;
+	if (node) {
+		struct ctl_table *entry;
+		for (entry = table; entry->procname; entry++, node++) {
+			rb_init_node(&node->node);
+			node->header = head;
+		}
+	}
 }
 
 static void erase_header(struct ctl_table_header *head)
 {
-	list_del_init(&head->ctl_entry);
+	struct ctl_table *entry;
+	for (entry = head->ctl_table; entry->procname; entry++)
+		erase_entry(head, entry);
 }
 
 static int insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
 {
+	struct ctl_table *entry;
 	int err;
 
-	err = sysctl_check_dups(dir, header->ctl_table);
-	if (err)
-		return err;
-
 	dir->header.nreg++;
 	header->parent = dir;
 	err = insert_links(header);
 	if (err)
 		goto fail_links;
-	list_add_tail(&header->ctl_entry, &header->parent->list);
+	for (entry = header->ctl_table; entry->procname; entry++) {
+		err = insert_entry(header, entry);
+		if (err)
+			goto fail;
+	}
 	return 0;
+fail:
+	erase_header(header);
+	put_links(header);
 fail_links:
 	header->parent = NULL;
 	drop_sysctl_table(&dir->header);
@@ -241,19 +307,14 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
 	return entry;
 }
 
-static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
-						  struct list_head *tmp)
+static struct ctl_node *first_usable_entry(struct rb_node *node)
 {
-	struct ctl_table_header *head;
-
-	for (tmp = tmp->next; tmp != &dir->list; tmp = tmp->next) {
-		head = list_entry(tmp, struct ctl_table_header, ctl_entry);
+	struct ctl_node *ctl_node;
 
-		if (!head->ctl_table->procname ||
-		    !use_table(head))
-			continue;
-
-		return head;
+	for (;node; node = rb_next(node)) {
+		ctl_node = rb_entry(node, struct ctl_node, node);
+		if (use_table(ctl_node->header))
+			return ctl_node;
 	}
 	return NULL;
 }
@@ -261,14 +322,17 @@ static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
 static void first_entry(struct ctl_dir *dir,
 	struct ctl_table_header **phead, struct ctl_table **pentry)
 {
-	struct ctl_table_header *head;
+	struct ctl_table_header *head = NULL;
 	struct ctl_table *entry = NULL;
+	struct ctl_node *ctl_node;
 
 	spin_lock(&sysctl_lock);
-	head = next_usable_entry(dir, &dir->list);
+	ctl_node = first_usable_entry(rb_first(&dir->root));
 	spin_unlock(&sysctl_lock);
-	if (head)
-		entry = head->ctl_table;
+	if (ctl_node) {
+		head = ctl_node->header;
+		entry = &head->ctl_table[ctl_node - head->node];
+	}
 	*phead = head;
 	*pentry = entry;
 }
@@ -277,15 +341,17 @@ static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentr
 {
 	struct ctl_table_header *head = *phead;
 	struct ctl_table *entry = *pentry;
+	struct ctl_node *ctl_node = &head->node[entry - head->ctl_table];
 
-	entry++;
-	if (!entry->procname) {
-		spin_lock(&sysctl_lock);
-		unuse_table(head);
-		head = next_usable_entry(head->parent, &head->ctl_entry);
-		spin_unlock(&sysctl_lock);
-		if (head)
-			entry = head->ctl_table;
+	spin_lock(&sysctl_lock);
+	unuse_table(head);
+
+	ctl_node = first_usable_entry(rb_next(&ctl_node->node));
+	spin_unlock(&sysctl_lock);
+	head = NULL;
+	if (ctl_node) {
+		head = ctl_node->header;
+		entry = &head->ctl_table[ctl_node - head->node];
 	}
 	*phead = head;
 	*pentry = entry;
@@ -777,21 +843,23 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 {
 	struct ctl_table *table;
 	struct ctl_dir *new;
+	struct ctl_node *node;
 	char *new_name;
 
-	new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 +
-		      namelen + 1, GFP_KERNEL);
+	new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) +
+		      sizeof(struct ctl_table)*2 +  namelen + 1,
+		      GFP_KERNEL);
 	if (!new)
 		return NULL;
 
-	table = (struct ctl_table *)(new + 1);
+	node = (struct ctl_node *)(new + 1);
+	table = (struct ctl_table *)(node + 1);
 	new_name = (char *)(table + 2);
 	memcpy(new_name, name, namelen);
 	new_name[namelen] = '\0';
-	INIT_LIST_HEAD(&new->list);
 	table[0].procname = new_name;
 	table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
-	init_header(&new->header, set->dir.header.root, set, table);
+	init_header(&new->header, set->dir.header.root, set, node, table);
 
 	return new;
 }
@@ -892,40 +960,6 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	return ret;
 }
 
-static int sysctl_check_table_dups(struct ctl_dir *dir, struct ctl_table *old,
-	struct ctl_table *table)
-{
-	struct ctl_table *entry, *test;
-	int error = 0;
-
-	for (entry = old; entry->procname; entry++) {
-		for (test = table; test->procname; test++) {
-			if (strcmp(entry->procname, test->procname) == 0) {
-				printk(KERN_ERR "sysctl duplicate entry: ");
-				sysctl_print_dir(dir);
-				printk(KERN_CONT "/%s\n", test->procname);
-				error = -EEXIST;
-			}
-		}
-	}
-	return error;
-}
-
-static int sysctl_check_dups(struct ctl_dir *dir, struct ctl_table *table)
-{
-	struct ctl_table_header *head;
-	int error = 0;
-
-	list_for_each_entry(head, &dir->list, ctl_entry) {
-		if (head->unregistering)
-			continue;
-		if (head->parent != dir)
-			continue;
-		error = sysctl_check_table_dups(dir, head->ctl_table, table);
-	}
-	return error;
-}
-
 static int sysctl_err(const char *path, struct ctl_table *table, char *fmt, ...)
 {
 	struct va_format vaf;
@@ -977,6 +1011,7 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 {
 	struct ctl_table *link_table, *entry, *link;
 	struct ctl_table_header *links;
+	struct ctl_node *node;
 	char *link_name;
 	int nr_entries, name_bytes;
 
@@ -988,6 +1023,7 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 	}
 
 	links = kzalloc(sizeof(struct ctl_table_header) +
+			sizeof(struct ctl_node)*nr_entries +
 			sizeof(struct ctl_table)*(nr_entries + 1) +
 			name_bytes,
 			GFP_KERNEL);
@@ -995,7 +1031,8 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 	if (!links)
 		return NULL;
 
-	link_table = (struct ctl_table *)(links + 1);
+	node = (struct ctl_node *)(links + 1);
+	link_table = (struct ctl_table *)(node + nr_entries);
 	link_name = (char *)&link_table[nr_entries + 1];
 
 	for (link = link_table, entry = table; entry->procname; link++, entry++) {
@@ -1006,7 +1043,7 @@ static struct ctl_table_header *new_links(struct ctl_dir *dir, struct ctl_table
 		link->data = link_root;
 		link_name += len;
 	}
-	init_header(links, dir->header.root, dir->header.set, link_table);
+	init_header(links, dir->header.root, dir->header.set, node, link_table);
 	links->nreg = nr_entries;
 
 	return links;
@@ -1132,12 +1169,20 @@ struct ctl_table_header *__register_sysctl_table(
 	struct ctl_table_header *header;
 	const char *name, *nextname;
 	struct ctl_dir *dir;
+	struct ctl_table *entry;
+	struct ctl_node *node;
+	int nr_entries = 0;
+
+	for (entry = table; entry->procname; entry++)
+		nr_entries++;
 
-	header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
+	header = kzalloc(sizeof(struct ctl_table_header) +
+			 sizeof(struct ctl_node)*nr_entries, GFP_KERNEL);
 	if (!header)
 		return NULL;
 
-	init_header(header, root, set, table);
+	node = (struct ctl_node *)(header + 1);
+	init_header(header, root, set, node, table);
 	if (sysctl_check_table(path, table))
 		goto fail;
 
@@ -1489,13 +1534,12 @@ void setup_sysctl_set(struct ctl_table_set *set,
 {
 	memset(set, sizeof(*set), 0);
 	set->is_seen = is_seen;
-	INIT_LIST_HEAD(&set->dir.list);
-	init_header(&set->dir.header, root, set, root_table);
+	init_header(&set->dir.header, root, set, NULL, root_table);
 }
 
 void retire_sysctl_set(struct ctl_table_set *set)
 {
-	WARN_ON(!list_empty(&set->dir.list));
+	WARN_ON(!RB_EMPTY_ROOT(&set->dir.root));
 }
 
 int __init proc_sys_init(void)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 36dec75..35c50ed 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -932,6 +932,7 @@ enum
 #include <linux/list.h>
 #include <linux/rcupdate.h>
 #include <linux/wait.h>
+#include <linux/rbtree.h>
 
 /* For the /proc/sys support */
 struct ctl_table;
@@ -1023,6 +1024,11 @@ struct ctl_table
 	void *extra2;
 };
 
+struct ctl_node {
+	struct rb_node node;
+	struct ctl_table_header *header;
+};
+
 /* struct ctl_table_header is used to maintain dynamic lists of
    struct ctl_table trees. */
 struct ctl_table_header
@@ -1030,7 +1036,6 @@ struct ctl_table_header
 	union {
 		struct {
 			struct ctl_table *ctl_table;
-			struct list_head ctl_entry;
 			int used;
 			int count;
 			int nreg;
@@ -1042,12 +1047,13 @@ struct ctl_table_header
 	struct ctl_table_root *root;
 	struct ctl_table_set *set;
 	struct ctl_dir *parent;
+	struct ctl_node *node;
 };
 
 struct ctl_dir {
 	/* Header must be at the start of ctl_dir */
 	struct ctl_table_header header;
-	struct list_head list;
+	struct rb_root root;
 };
 
 struct ctl_table_set {
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (56 preceding siblings ...)
  2012-01-27  4:52 ` [PATCH 28/29] sysctl: Index sysctl directories with rbtrees Eric W. Biederman
@ 2012-01-27  4:52 ` Eric W. Biederman
  2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
  58 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-27  4:52 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Eric W. Biederman

The plan is to convert all callers of register_sysctl_table
and register_sysctl_paths to register_sysctl.  The interface
to register_sysctl is enough nicer this should make the callers
a bit more readable.  Additionally after the conversion the
230 lines of backwards compatibility can be removed.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c  |   17 +++++++++++++++++
 include/linux/sysctl.h |    1 +
 2 files changed, 18 insertions(+), 0 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 05c393a..8dc7f0e 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -1228,6 +1228,23 @@ fail:
 	return NULL;
 }
 
+/**
+ * register_sysctl - register a sysctl table
+ * @path: The path to the directory the sysctl table is in.
+ * @table: the table structure
+ *
+ * Register a sysctl table. @table should be a filled in ctl_table
+ * array. A completely 0 filled entry terminates the table.
+ *
+ * See __register_sysctl_table for more details.
+ */
+struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table)
+{
+	return __register_sysctl_table(&sysctl_table_root.default_set,
+					path, table);
+}
+EXPORT_SYMBOL(register_sysctl);
+
 static char *append_path(const char *path, char *pos, const char *name)
 {
 	int namelen;
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 35c50ed..c34b4c8 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -1090,6 +1090,7 @@ struct ctl_table_header *__register_sysctl_table(
 struct ctl_table_header *__register_sysctl_paths(
 	struct ctl_table_set *set,
 	const struct ctl_path *path, struct ctl_table *table);
+struct ctl_table_header *register_sysctl(const char *path, struct ctl_table *table);
 struct ctl_table_header *register_sysctl_table(struct ctl_table * table);
 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
 						struct ctl_table *table);
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* Re: [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry.
  2012-01-27  4:51 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
@ 2012-01-29 15:49   ` Lucian Adrian Grijincu
  0 siblings, 0 replies; 76+ messages in thread
From: Lucian Adrian Grijincu @ 2012-01-29 15:49 UTC (permalink / raw)
  To: Eric W. Biederman; +Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=UTF-8, Size: 6670 bytes --]

On Fri, Jan 27, 2012 at 6:51 AM, Eric W. Biederman
<ebiederm@xmission.com> wrote:
> Replace the helpers that proc_sys_lookup uses with helpers that work
> in terms of an entire sysctl directory.  This is worse for sysctl_lock
> hold times but it is much better for code clarity and the code cleanups
> to come.
>
> find_in_table is no longer needed so it is removed.
>
> find_entry a general helper to find entries in a directory is added.
>
> lookup_entry is a simple wrapper around find_entry that takes the
> sysctl_lock increases the use count if an entry is found and drops
> the sysctl_lock.
>
> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
> ---
>  fs/proc/proc_sysctl.c |  102 ++++++++++++++++++++++++++++++++++++------------
>  1 files changed, 76 insertions(+), 26 deletions(-)
>
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index 88d1b06..3b63f29 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -49,6 +49,55 @@ static struct ctl_table_root sysctl_table_root = {
>
>  static DEFINE_SPINLOCK(sysctl_lock);
>
> +static int namecmp(const char *name1, int len1, const char *name2, int len2)
> +{
> +       int minlen;
> +       int cmp;
> +
> +       minlen = len1;
> +       if (minlen > len2)
> +               minlen = len2;
> +
> +       cmp = memcmp(name1, name2, minlen);
> +       if (cmp == 0)
> +               cmp = len1 - len2;
> +       return cmp;
> +}
> +


I would add a note that this needs sysctl_lock as this requirement
remains valid in the final version of the code too:

/* called under sysctl_lock */

> +static struct ctl_table *find_entry(struct ctl_table_header **phead,
> +       struct ctl_table_set *set,
> +       struct ctl_table_header *dir_head, struct ctl_table *dir,
> +       const char *name, int namelen)
> +{
> +       struct ctl_table_header *head;
> +       struct ctl_table *entry;
> +
> +       if (dir_head->set == set) {
> +               for (entry = dir; entry->procname; entry++) {
> +                       const char *procname = entry->procname;
> +                       if (namecmp(procname, strlen(procname), name, namelen) == 0) {
> +                               *phead = dir_head;
> +                               return entry;
> +                       }
> +               }
> +       }
> +
> +       list_for_each_entry(head, &set->list, ctl_entry) {
> +               if (head->unregistering)
> +                       continue;
> +               if (head->attached_to != dir)
> +                       continue;
> +               for (entry = head->attached_by; entry->procname; entry++) {
> +                       const char *procname = entry->procname;
> +                       if (namecmp(procname, strlen(procname), name, namelen) == 0) {
> +                               *phead = head;
> +                               return entry;
> +                       }
> +               }
> +       }
> +       return NULL;
> +}
> +
>  static void init_header(struct ctl_table_header *head,
>        struct ctl_table_root *root, struct ctl_table_set *set,
>        struct ctl_table *table)
> @@ -168,6 +217,32 @@ lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
>        return &set->list;
>  }
>
> +static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
> +                                     struct ctl_table_header *dir_head,
> +                                     struct ctl_table *dir,
> +                                     const char *name, int namelen)
> +{
> +       struct ctl_table_header *head;
> +       struct ctl_table *entry;
> +       struct ctl_table_root *root;
> +       struct ctl_table_set *set;
> +
> +       spin_lock(&sysctl_lock);
> +       root = &sysctl_table_root;
> +       do {
> +               set = lookup_header_set(root, current->nsproxy);
> +               entry = find_entry(&head, set, dir_head, dir, name, namelen);
> +               if (entry && use_table(head))
> +                       *phead = head;
> +               else
> +                       entry = NULL;
> +               root = list_entry(root->root_list.next,
> +                                 struct ctl_table_root, root_list);
> +       } while (!entry && root != &sysctl_table_root);
> +       spin_unlock(&sysctl_lock);
> +       return entry;
> +}
> +
>  static struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
>                                                struct ctl_table_header *prev)
>  {
> @@ -284,21 +359,6 @@ out:
>        return inode;
>  }
>
> -static struct ctl_table *find_in_table(struct ctl_table *p, struct qstr *name)
> -{
> -       for ( ; p->procname; p++) {
> -               if (strlen(p->procname) != name->len)
> -                       continue;
> -
> -               if (memcmp(p->procname, name->name, name->len) != 0)
> -                       continue;
> -
> -               /* I have a match */
> -               return p;
> -       }
> -       return NULL;
> -}
> -
>  static struct ctl_table_header *grab_header(struct inode *inode)
>  {
>        struct ctl_table_header *head = PROC_I(inode)->sysctl;
> @@ -328,17 +388,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>
>        table = table ? table->child : &head->ctl_table[1];
>
> -       p = find_in_table(table, name);
> -       if (!p) {
> -               for (h = sysctl_head_next(NULL); h; h = sysctl_head_next(h)) {
> -                       if (h->attached_to != table)
> -                               continue;
> -                       p = find_in_table(h->attached_by, name);
> -                       if (p)
> -                               break;
> -               }
> -       }
> -
> +       p = lookup_entry(&h, head, table, name->name, name->len);
>        if (!p)
>                goto out;
>
> --
> 1.7.2.5
>



-- 
 .
..: Lucian
ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±þG«éÿŠ{ayº\x1dʇڙë,j\a­¢f£¢·hšïêÿ‘êçz_è®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿÿ¾\a«þG«éÿ¢¸?™¨è­Ú&£ø§~á¶iO•æ¬z·švØ^\x14\x04\x1a¶^[m§ÿÿÃ\fÿ¶ìÿ¢¸?–I¥

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set
  2012-01-27  4:51 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
@ 2012-01-29 17:19   ` Lucian Adrian Grijincu
  0 siblings, 0 replies; 76+ messages in thread
From: Lucian Adrian Grijincu @ 2012-01-29 17:19 UTC (permalink / raw)
  To: Eric W. Biederman; +Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

On Fri, Jan 27, 2012 at 6:51 AM, Eric W. Biederman
<ebiederm@xmission.com> wrote:
> Add a ctl_table_root pointer to ctl_table set so it is easy to

"ctl_table set" -> "ctl_table_set"

> go from a ctl_table_set to a ctl_table_root.
>
> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
> ---
>  fs/proc/proc_sysctl.c  |    3 +++
>  include/linux/sysctl.h |    3 +++
>  net/sysctl_net.c       |    5 ++---
>  3 files changed, 8 insertions(+), 3 deletions(-)
>
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index d9c3ae6..65c13dd 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -45,6 +45,7 @@ static struct ctl_table_header root_table_header = {
>  static struct ctl_table_root sysctl_table_root = {
>        .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
>        .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
> +       .default_set.root = &sysctl_table_root,
>  };
>
>  static DEFINE_SPINLOCK(sysctl_lock);
> @@ -1348,9 +1349,11 @@ void unregister_sysctl_table(struct ctl_table_header * header)
>  EXPORT_SYMBOL(unregister_sysctl_table);
>
>  void setup_sysctl_set(struct ctl_table_set *p,
> +       struct ctl_table_root *root,
>        int (*is_seen)(struct ctl_table_set *))
>  {
>        INIT_LIST_HEAD(&p->list);
> +       p->root = root;
>        p->is_seen = is_seen;
>  }
>
> diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
> index e40b8f6..e73ba33 100644
> --- a/include/linux/sysctl.h
> +++ b/include/linux/sysctl.h
> @@ -1047,6 +1047,7 @@ struct ctl_table_header
>
>  struct ctl_table_set {
>        struct list_head list;
> +       struct ctl_table_root *root;
>        int (*is_seen)(struct ctl_table_set *);
>  };
>
> @@ -1069,6 +1070,7 @@ struct ctl_path {
>  void proc_sys_poll_notify(struct ctl_table_poll *poll);
>
>  extern void setup_sysctl_set(struct ctl_table_set *p,
> +       struct ctl_table_root *root,
>        int (*is_seen)(struct ctl_table_set *));
>  extern void retire_sysctl_set(struct ctl_table_set *set);
>
> @@ -1103,6 +1105,7 @@ static inline void unregister_sysctl_table(struct ctl_table_header * table)
>  }
>
>  static inline void setup_sysctl_set(struct ctl_table_set *p,
> +       struct ctl_table_root *root,
>        int (*is_seen)(struct ctl_table_set *))
>  {
>  }
> diff --git a/net/sysctl_net.c b/net/sysctl_net.c
> index 07c6b87..e998c64 100644
> --- a/net/sysctl_net.c
> +++ b/net/sysctl_net.c
> @@ -74,8 +74,7 @@ static struct ctl_table_root net_sysctl_ro_root = {
>
>  static int __net_init sysctl_net_init(struct net *net)
>  {
> -       setup_sysctl_set(&net->sysctls,
> -                        is_seen);
> +       setup_sysctl_set(&net->sysctls, &net_sysctl_root, is_seen);
>        return 0;
>  }
>
> @@ -95,7 +94,7 @@ static __init int net_sysctl_init(void)
>        ret = register_pernet_subsys(&sysctl_pernet_ops);
>        if (ret)
>                goto out;
> -       setup_sysctl_set(&net_sysctl_ro_root.default_set, NULL);
> +       setup_sysctl_set(&net_sysctl_ro_root.default_set, &net_sysctl_ro_root, NULL);
>        register_sysctl_root(&net_sysctl_ro_root);
>        register_sysctl_root(&net_sysctl_root);
>  out:
> --
> 1.7.2.5
>



-- 
 .
..: Lucian

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 18/29] sysctl: Normalize the root_table data structure.
  2012-01-27  4:51 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
@ 2012-01-29 17:36   ` Lucian Adrian Grijincu
  2012-01-30  0:22     ` Eric W. Biederman
  0 siblings, 1 reply; 76+ messages in thread
From: Lucian Adrian Grijincu @ 2012-01-29 17:36 UTC (permalink / raw)
  To: Eric W. Biederman; +Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

On Fri, Jan 27, 2012 at 6:51 AM, Eric W. Biederman
<ebiederm@xmission.com> wrote:
> Every other directory has a .child member and we look at the .child
> for our entries.  Do the same for the root_table.
>
> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
> ---
>  fs/proc/proc_sysctl.c |   15 +++++++++++----
>  1 files changed, 11 insertions(+), 4 deletions(-)
>
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index 7e96a26..88d1b06 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -25,7 +25,14 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
>        wake_up_interruptible(&poll->wait);
>  }
>
> -static struct ctl_table root_table[1];
> +static struct ctl_table root_table[] = {
> +       {
> +               .procname = "",
> +               .mode = S_IRUGO|S_IXUGO,

Why not: .mode = S_IFDIR|S_IRUGO|S_IXUGO ?

You change it later and add IFDIR in patch 22/29 because of this
change (from 22/29):
-	if (!table->child) {
+	if (!S_ISDIR(table->mode)) {

but if might as well be done here.


> +               .child = &root_table[1],
> +       },
> +       { }
> +};
>  static struct ctl_table_root sysctl_table_root;
>  static struct ctl_table_header root_table_header = {
>        {{.count = 1,
> @@ -319,7 +326,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>                goto out;
>        }

Somewhere above we have
  struct ctl_table *table = PROC_I(inode)->sysctl_entry;
and sysctl_entry can take two values:
   - NULL: fs/proc/inode.c:proc_alloc_inode()
   - a non-NULL table in: proc_sys_make_inode()

The only inode that can be passed to proc_sys_lookup that can have
table=NULL is the root inode.
In that case head will be &root_table_header.


So head->ctl_table[1] == root_table_header.ctl_table[1] == root_table[1].


>
> -       table = table ? table->child : head->ctl_table;
> +       table = table ? table->child : &head->ctl_table[1];

I think this could be improved to something like this:
/* table == NULL only for the procfs root directory */
table = table ? table->child : &root_table[1];


It's not that important, as this code will go away in a few patches.

>
>        p = find_in_table(table, name);
>        if (!p) {
> @@ -510,7 +517,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
>                goto out;
>        }
>
> -       table = table ? table->child : head->ctl_table;
> +       table = table ? table->child : &head->ctl_table[1];

Similar.

>        ret = 0;
>        /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
> @@ -966,7 +973,7 @@ struct ctl_table_header *__register_sysctl_table(
>        spin_lock(&sysctl_lock);
>        header->set = lookup_header_set(root, namespaces);
>        header->attached_by = header->ctl_table;
> -       header->attached_to = root_table;
> +       header->attached_to = &root_table[1];
>        header->parent = &root_table_header;
>        set = header->set;
>        root = header->root;
> --
> 1.7.2.5
>



-- 
 .
..: Lucian

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories
  2012-01-27  4:51 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
@ 2012-01-29 19:31   ` Lucian Adrian Grijincu
  2012-01-31  4:45     ` Eric W. Biederman
  0 siblings, 1 reply; 76+ messages in thread
From: Lucian Adrian Grijincu @ 2012-01-29 19:31 UTC (permalink / raw)
  To: Eric W. Biederman; +Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

On Fri, Jan 27, 2012 at 6:51 AM, Eric W. Biederman
<ebiederm@xmission.com> wrote:
> Simplify the code and the sysctl semantics by autogenerating
> sysctl directories when a sysctl table is registered that needs
> the directories and autodeleting the directories when there are
> no more sysctl tables registered that need them.
>
> Autogenerating directories keeps sysctl tables from depending
> on each other, removing all of the arcane register/unregister
> ordering constraints and makes it impossible to get the order
> wrong when reigsering and unregistering sysctl tables.
>
> Autogenerating directories yields one unique entity that dentries
> can point to, retaining the current effective use of the dcache.
>
> Add struct ctl_dir as the type of these new autogenerated
> directories.
>
> The attached_by and attached_to fields in ctl_table_header are
> removed as they are no longer needed.
>
> The child field in ctl_table is no longer needed by the core of
> the sysctl code.  ctl_table.child can be removed once all of the
> existing users have been updated.
>
> Benchmark before:
>    make-dummies 0 999 -> 0.7s
>    rmmod dummy        -> 0.07s
>    make-dummies 0 9999 -> 1m10s
>    rmmod dummy         -> 0.4s
>
> Benchmark after:
>    make-dummies 0 999 -> 0.44s
>    rmmod dummy        -> 0.065s
>    make-dummies 0 9999 -> 1m36s
>    rmmod dummy         -> 0.4s
>
> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
> ---
>  fs/proc/proc_sysctl.c  |  342 ++++++++++++++++++++----------------------------
>  include/linux/sysctl.h |   10 +-
>  2 files changed, 150 insertions(+), 202 deletions(-)
>
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index 65c13dd..3c0767d 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -28,28 +28,31 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
>  static struct ctl_table root_table[] = {
>        {
>                .procname = "",
> -               .mode = S_IRUGO|S_IXUGO,
> -               .child = &root_table[1],
> +               .mode = S_IFDIR|S_IRUGO|S_IXUGO,
>        },
>        { }
>  };
>  static struct ctl_table_root sysctl_table_root;
> -static struct ctl_table_header root_table_header = {
> -       {{.count = 1,
> -         .nreg = 1,
> -         .ctl_table = root_table,
> -         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
> -       .root = &sysctl_table_root,
> -       .set = &sysctl_table_root.default_set,
> +static struct ctl_dir sysctl_root_dir = {
> +       .header = {
> +               {{.count = 1,
> +                 .nreg = 1,
> +                 .ctl_table = root_table,
> +                 .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
> +               .root = &sysctl_table_root,
> +               .set = &sysctl_table_root.default_set,
> +       },
>  };
>  static struct ctl_table_root sysctl_table_root = {
>        .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
> -       .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
> +       .default_set.list = LIST_HEAD_INIT(sysctl_root_dir.header.ctl_entry),
>        .default_set.root = &sysctl_table_root,
>  };
>
>  static DEFINE_SPINLOCK(sysctl_lock);
>
> +static void drop_sysctl_table(struct ctl_table_header *header);
> +
>  static int namecmp(const char *name1, int len1, const char *name2, int len2)
>  {
>        int minlen;
> @@ -66,29 +69,18 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
>  }
>
>  static struct ctl_table *find_entry(struct ctl_table_header **phead,
> -       struct ctl_table_set *set,
> -       struct ctl_table_header *dir_head, struct ctl_table *dir,
> +       struct ctl_table_set *set, struct ctl_dir *dir,
>        const char *name, int namelen)
>  {
>        struct ctl_table_header *head;
>        struct ctl_table *entry;
>
> -       if (dir_head->set == set) {
> -               for (entry = dir; entry->procname; entry++) {
> -                       const char *procname = entry->procname;
> -                       if (namecmp(procname, strlen(procname), name, namelen) == 0) {
> -                               *phead = dir_head;
> -                               return entry;
> -                       }
> -               }
> -       }
> -
>        list_for_each_entry(head, &set->list, ctl_entry) {
>                if (head->unregistering)
>                        continue;
> -               if (head->attached_to != dir)
> +               if (head->parent != dir)
>                        continue;
> -               for (entry = head->attached_by; entry->procname; entry++) {
> +               for (entry = head->ctl_table; entry->procname; entry++) {
>                        const char *procname = entry->procname;
>                        if (namecmp(procname, strlen(procname), name, namelen) == 0) {
>                                *phead = head;
> @@ -103,6 +95,7 @@ static void init_header(struct ctl_table_header *head,
>        struct ctl_table_root *root, struct ctl_table_set *set,
>        struct ctl_table *table)
>  {
> +       head->ctl_table = table;
>        head->ctl_table_arg = table;
>        INIT_LIST_HEAD(&head->ctl_entry);
>        head->used = 0;
> @@ -119,9 +112,10 @@ static void erase_header(struct ctl_table_header *head)
>        list_del_init(&head->ctl_entry);
>  }
>
> -static void insert_header(struct ctl_table_header *header)
> +static void insert_header(struct ctl_dir *dir, struct ctl_table_header *header)
>  {
> -       header->parent->count++;
> +       header->parent = dir;
> +       header->parent->header.nreg++;
>        list_add_tail(&header->ctl_entry, &header->set->list);
>  }
>
> @@ -219,8 +213,7 @@ lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
>  }
>
>  static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
> -                                     struct ctl_table_header *dir_head,
> -                                     struct ctl_table *dir,
> +                                     struct ctl_dir *dir,
>                                      const char *name, int namelen)
>  {
>        struct ctl_table_header *head;
> @@ -232,7 +225,7 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
>        root = &sysctl_table_root;
>        do {
>                set = lookup_header_set(root, current->nsproxy);
> -               entry = find_entry(&head, set, dir_head, dir, name, namelen);
> +               entry = find_entry(&head, set, dir, name, namelen);
>                if (entry && use_table(head))
>                        *phead = head;
>                else
> @@ -244,7 +237,7 @@ static struct ctl_table *lookup_entry(struct ctl_table_header **phead,
>        return entry;
>  }
>
> -static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
> +static struct ctl_table_header *next_usable_entry(struct ctl_dir *dir,
>        struct ctl_table_root *root, struct list_head *tmp)
>  {
>        struct nsproxy *namespaces = current->nsproxy;
> @@ -256,8 +249,8 @@ static struct ctl_table_header *next_usable_entry(struct ctl_table *dir,
>                head = list_entry(tmp, struct ctl_table_header, ctl_entry);
>                root = head->root;
>
> -               if (head->attached_to != dir ||
> -                   !head->attached_by->procname ||
> +               if (head->parent != dir ||
> +                   !head->ctl_table->procname ||
>                    !use_table(head))
>                        goto next;
>
> @@ -281,47 +274,35 @@ out:
>        return NULL;
>  }
>
> -static void first_entry(
> -       struct ctl_table_header *dir_head, struct ctl_table *dir,
> +static void first_entry(struct ctl_dir *dir,
>        struct ctl_table_header **phead, struct ctl_table **pentry)
>  {
> -       struct ctl_table_header *head = dir_head;
> -       struct ctl_table *entry = dir;
> +       struct ctl_table_header *head;
> +       struct ctl_table *entry = NULL;
>
>        spin_lock(&sysctl_lock);
> -       if (entry->procname) {
> -               use_table(head);
> -       } else {
> -               head = next_usable_entry(dir, &sysctl_table_root,
> -                                        &sysctl_table_root.default_set.list);
> -               if (head)
> -                       entry = head->attached_by;
> -       }
> +       head = next_usable_entry(dir, &sysctl_table_root,
> +                                &sysctl_table_root.default_set.list);
>        spin_unlock(&sysctl_lock);
> +       if (head)
> +               entry = head->ctl_table;
>        *phead = head;
>        *pentry = entry;
>  }
>
> -static void next_entry(struct ctl_table *dir,
> -       struct ctl_table_header **phead, struct ctl_table **pentry)
> +static void next_entry(struct ctl_table_header **phead, struct ctl_table **pentry)
>  {
>        struct ctl_table_header *head = *phead;
>        struct ctl_table *entry = *pentry;
>
>        entry++;
>        if (!entry->procname) {
> -               struct ctl_table_root *root = head->root;
> -               struct list_head *tmp = &head->ctl_entry;
> -               if (head->attached_to != dir) {
> -                       root = &sysctl_table_root;
> -                       tmp = &sysctl_table_root.default_set.list;
> -               }
>                spin_lock(&sysctl_lock);
>                unuse_table(head);
> -               head = next_usable_entry(dir, root, tmp);
> +               head = next_usable_entry(head->parent, head->root, &head->ctl_entry);
>                spin_unlock(&sysctl_lock);
>                if (head)
> -                       entry = head->attached_by;
> +                       entry = head->ctl_table;
>        }
>        *phead = head;
>        *pentry = entry;
> @@ -381,7 +362,7 @@ static struct inode *proc_sys_make_inode(struct super_block *sb,
>
>        inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
>        inode->i_mode = table->mode;
> -       if (!table->child) {
> +       if (!S_ISDIR(table->mode)) {
>                inode->i_mode |= S_IFREG;
>                inode->i_op = &proc_sys_inode_operations;
>                inode->i_fop = &proc_sys_file_operations;
> @@ -398,7 +379,7 @@ static struct ctl_table_header *grab_header(struct inode *inode)
>  {
>        struct ctl_table_header *head = PROC_I(inode)->sysctl;
>        if (!head)
> -               head = &root_table_header;
> +               head = &sysctl_root_dir.header;
>        return sysctl_head_grab(head);
>  }
>
> @@ -406,24 +387,19 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>                                        struct nameidata *nd)
>  {
>        struct ctl_table_header *head = grab_header(dir);
> -       struct ctl_table *table = PROC_I(dir)->sysctl_entry;
>        struct ctl_table_header *h = NULL;
>        struct qstr *name = &dentry->d_name;
>        struct ctl_table *p;
>        struct inode *inode;
>        struct dentry *err = ERR_PTR(-ENOENT);
> +       struct ctl_dir *ctl_dir;
>
>        if (IS_ERR(head))
>                return ERR_CAST(head);
>
> -       if (table && !table->child) {
> -               WARN_ON(1);
> -               goto out;
> -       }
> +       ctl_dir = container_of(head, struct ctl_dir, header);
>
> -       table = table ? table->child : &head->ctl_table[1];
> -
> -       p = lookup_entry(&h, head, table, name->name, name->len);
> +       p = lookup_entry(&h, ctl_dir, name->name, name->len);
>        if (!p)
>                goto out;
>
> @@ -586,21 +562,16 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
>        struct dentry *dentry = filp->f_path.dentry;
>        struct inode *inode = dentry->d_inode;
>        struct ctl_table_header *head = grab_header(inode);
> -       struct ctl_table *table = PROC_I(inode)->sysctl_entry;
>        struct ctl_table_header *h = NULL;
>        struct ctl_table *entry;
> +       struct ctl_dir *ctl_dir;
>        unsigned long pos;
>        int ret = -EINVAL;
>
>        if (IS_ERR(head))
>                return PTR_ERR(head);
>
> -       if (table && !table->child) {
> -               WARN_ON(1);
> -               goto out;
> -       }
> -
> -       table = table ? table->child : &head->ctl_table[1];
> +       ctl_dir = container_of(head, struct ctl_dir, header);
>
>        ret = 0;
>        /* Avoid a switch here: arm builds fail with missing __cmpdi2 */
> @@ -618,7 +589,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
>        }
>        pos = 2;
>
> -       for (first_entry(head, table, &h, &entry); h; next_entry(table, &h, &entry)) {
> +       for (first_entry(ctl_dir, &h, &entry); h; next_entry(&h, &entry)) {
>                ret = scan(h, entry, &pos, filp, dirent, filldir);
>                if (ret) {
>                        sysctl_head_finish(h);
> @@ -779,52 +750,86 @@ static const struct dentry_operations proc_sys_dentry_operations = {
>        .d_compare      = proc_sys_compare,
>  };
>
> -static struct ctl_table *is_branch_in(struct ctl_table *branch,
> -                                     struct ctl_table *table)
> +static struct ctl_dir *find_subdir(struct ctl_table_set *set, struct ctl_dir *dir,
> +       const char *name, int namelen)
>  {
> -       struct ctl_table *p;
> -       const char *s = branch->procname;
> +       struct ctl_table_header *head;
> +       struct ctl_table *entry;
>
> -       /* branch should have named subdirectory as its first element */
> -       if (!s || !branch->child)
> -               return NULL;
> +       entry = find_entry(&head, set, dir, name, namelen);
> +       if (!entry)
> +               return ERR_PTR(-ENOENT);
> +       if (S_ISDIR(entry->mode))
> +               return container_of(head, struct ctl_dir, header);
> +       return ERR_PTR(-ENOTDIR);
> +}


I find this easier to read:

        entry = find_entry(&head, dir, name, namelen);
        if (!entry)
                return ERR_PTR(-ENOENT);
        if (!S_ISDIR(entry->mode))
                return ERR_PTR(-ENOTDIR);
        return container_of(head, struct ctl_dir, header);




> +
> +static struct ctl_dir *new_dir(struct ctl_table_set *set,
> +       const char *name, int namelen)
> +{
> +       struct ctl_table *table;
> +       struct ctl_dir *new;
> +       char *new_name;
>
> -       /* ... and nothing else */
> -       if (branch[1].procname)
> +       new = kzalloc(sizeof(*new) + sizeof(struct ctl_table)*2 +
> +                     namelen + 1, GFP_KERNEL);
> +       if (!new)
>                return NULL;
>
> -       /* table should contain subdirectory with the same name */
> -       for (p = table; p->procname; p++) {
> -               if (!p->child)
> -                       continue;
> -               if (p->procname && strcmp(p->procname, s) == 0)
> -                       return p;
> -       }
> -       return NULL;
> +       table = (struct ctl_table *)(new + 1);
> +       new_name = (char *)(table + 2);
> +       memcpy(new_name, name, namelen);
> +       new_name[namelen] = '\0';
> +       table[0].procname = new_name;
> +       table[0].mode = S_IFDIR|S_IRUGO|S_IXUGO;
> +       init_header(&new->header, set->root, set, table);
> +
> +       return new;
>  }
>
> -/* see if attaching q to p would be an improvement */
> -static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
>  {
> -       struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
> -       struct ctl_table *next;
> -       int is_better = 0;
> -       int not_in_parent = !p->attached_by;
> -
> -       while ((next = is_branch_in(by, to)) != NULL) {
> -               if (by == q->attached_by)
> -                       is_better = 1;
> -               if (to == p->attached_by)
> -                       not_in_parent = 1;
> -               by = by->child;
> -               to = next->child;
> -       }
> +       struct ctl_dir *subdir, *new = NULL;
>
> -       if (is_better && not_in_parent) {
> -               q->attached_by = by;
> -               q->attached_to = to;
> -               q->parent = p;



This one's hard to understand :) I would add these comments:



/* get_subdir - find and take a reference of a given subdir
 *                 - if no entry with the same name found, create one
 * - set - the set in which to look for the subdir (after looking in
the parent dir's set)
 * - dir - the parent dir. NOTE: before calling this function
 *          make sure to have a increased the count on the parent dir's nreg
 *          as this function will drop a reference!
 */

> +static struct ctl_dir *get_subdir(struct ctl_table_set *set,
> +       struct ctl_dir *dir, const char *name, int namelen)
> +       spin_lock(&sysctl_lock);

           /* first look in the parent dir's set */

> +       subdir = find_subdir(dir->header.set, dir, name, namelen);
> +       if (!IS_ERR(subdir))
> +               goto found;
> +       if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)

                   /* then look in the requested set if it's not the
same as the parent's set */

> +               subdir = find_subdir(set, dir, name, namelen);
> +       if (!IS_ERR(subdir))
> +               goto found;
> +       if (PTR_ERR(subdir) != -ENOENT)
> +               goto failed;
> +
> +       spin_unlock(&sysctl_lock);

           /* not found, let's make a new dir and we'll add it later */
> +       new = new_dir(set, name, namelen);

> +       spin_lock(&sysctl_lock);
> +       subdir = ERR_PTR(-ENOMEM);
> +       if (!new)
> +               goto failed;
> +
        /* check in the requested set: maybe someone added it while we
	   temporarily gave up the sysctl_lock to create the new dir.  */

> +       subdir = find_subdir(set, dir, name, namelen);
> +       if (!IS_ERR(subdir))
> +               goto found;
> +       if (PTR_ERR(subdir) != -ENOENT)
> +               goto failed;
> +
        /* nope, no one created an entry with this name while we gave
up the spin lock.
           We can add the new dir. */

> +       insert_header(dir, &new->header);
> +       subdir = new;
> +found:
> +       subdir->header.nreg++;
> +failed:
> +       if (unlikely(IS_ERR(subdir))) {
> +               printk(KERN_ERR "sysctl could not get directory: %*.*s %ld\n",
> +                       namelen, namelen, name, PTR_ERR(subdir));
>        }

           /* drop a reference from the parent dir, as we dive into
the subdir */

> +       drop_sysctl_table(&dir->header);
> +       if (new)
> +               drop_sysctl_table(&new->header);
> +       spin_unlock(&sysctl_lock);
> +       return subdir;
>  }
>


Why do you only check the given @set after creating the new dir. Why
not the parent dir too?


>  static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
> @@ -846,24 +851,14 @@ static int sysctl_check_table_dups(const char *path, struct ctl_table *old,
>  }
>
>  static int sysctl_check_dups(struct nsproxy *namespaces,
> -       struct ctl_table_header *header,
> +       struct ctl_dir *dir,
>        const char *path, struct ctl_table *table)
>  {
>        struct ctl_table_root *root;
>        struct ctl_table_set *set;
> -       struct ctl_table_header *dir_head, *head;
> -       struct ctl_table *dir_table;
> +       struct ctl_table_header *head;
>        int error = 0;
>
> -       /* No dups if we are the only member of our directory */
> -       if (header->attached_by != table)
> -               return 0;
> -
> -       dir_head = header->parent;
> -       dir_table = header->attached_to;
> -
> -       error = sysctl_check_table_dups(path, dir_table, table);
> -
>        root = &sysctl_table_root;
>        do {
>                set = lookup_header_set(root, namespaces);
> @@ -871,9 +866,9 @@ static int sysctl_check_dups(struct nsproxy *namespaces,
>                list_for_each_entry(head, &set->list, ctl_entry) {
>                        if (head->unregistering)
>                                continue;
> -                       if (head->attached_to != dir_table)
> +                       if (head->parent != dir)
>                                continue;
> -                       error = sysctl_check_table_dups(path, head->attached_by,
> +                       error = sysctl_check_table_dups(path, head->ctl_table,
>                                                        table);
>                }
>                root = list_entry(root->root_list.next,
> @@ -977,47 +972,25 @@ struct ctl_table_header *__register_sysctl_table(
>        const char *path, struct ctl_table *table)
>  {
>        struct ctl_table_header *header;
> -       struct ctl_table *new, **prevp;
>        const char *name, *nextname;
> -       unsigned int npath = 0;
>        struct ctl_table_set *set;
> -       size_t path_bytes = 0;
> -       char *new_name;
> -
> -       /* Count the path components */
> -       for (name = path; name; name = nextname) {
> -               int namelen;
> -               nextname = strchr(name, '/');
> -               if (nextname) {
> -                       namelen = nextname - name;
> -                       nextname++;
> -               } else {
> -                       namelen = strlen(name);
> -               }
> -               if (namelen == 0)
> -                       continue;
> -               path_bytes += namelen + 1;
> -               npath++;
> -       }
> +       struct ctl_dir *dir;
>
> -       /*
> -        * For each path component, allocate a 2-element ctl_table array.
> -        * The first array element will be filled with the sysctl entry
> -        * for this, the second will be the sentinel (procname == 0).
> -        *
> -        * We allocate everything in one go so that we don't have to
> -        * worry about freeing additional memory in unregister_sysctl_table.
> -        */
> -       header = kzalloc(sizeof(struct ctl_table_header) + path_bytes +
> -                        (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
> +       header = kzalloc(sizeof(struct ctl_table_header), GFP_KERNEL);
>        if (!header)
>                return NULL;
>
> -       new = (struct ctl_table *) (header + 1);
> -       new_name = (char *)(new + (2 * npath));
> +       init_header(header, root, NULL, table);
> +       if (sysctl_check_table(path, table))
> +               goto fail;
> +
> +       spin_lock(&sysctl_lock);
> +       header->set = set = lookup_header_set(root, namespaces);
> +       dir = &sysctl_root_dir;


/* get_subdir will drop a reference to this dir as it dives into the subdir */

> +       dir->header.nreg++;
> +       spin_unlock(&sysctl_lock);
>
> -       /* Now connect the dots */
> -       prevp = &header->ctl_table;
> +       /* Find the directory for the ctl_table */
>        for (name = path; name; name = nextname) {
>                int namelen;
>                nextname = strchr(name, '/');
> @@ -1029,51 +1002,21 @@ struct ctl_table_header *__register_sysctl_table(
>                }
>                if (namelen == 0)
>                        continue;
> -               memcpy(new_name, name, namelen);
> -               new_name[namelen] = '\0';
> -
> -               new->procname = new_name;
> -               new->mode     = 0555;
> -
> -               *prevp = new;
> -               prevp = &new->child;
>
> -               new += 2;
> -               new_name += namelen + 1;
> +               dir = get_subdir(set, dir, name, namelen);
> +               if (IS_ERR(dir))
> +                       goto fail;
>        }
> -       *prevp = table;
> -
> -       init_header(header, root, NULL, table);
> -       if (sysctl_check_table(path, table))
> -               goto fail;
> -
>        spin_lock(&sysctl_lock);
> -       header->set = lookup_header_set(root, namespaces);
> -       header->attached_by = header->ctl_table;
> -       header->attached_to = &root_table[1];
> -       header->parent = &root_table_header;
> -       set = header->set;
> -       root = header->root;
> -       for (;;) {
> -               struct ctl_table_header *p;
> -               list_for_each_entry(p, &set->list, ctl_entry) {
> -                       if (p->unregistering)
> -                               continue;
> -                       try_attach(p, header);
> -               }
> -               if (root == &sysctl_table_root)
> -                       break;
> -               root = list_entry(root->root_list.prev,
> -                                 struct ctl_table_root, root_list);
> -               set = lookup_header_set(root, namespaces);
> -       }
> -       if (sysctl_check_dups(namespaces, header, path, table))
> -               goto fail_locked;
> -       insert_header(header);
> +       if (sysctl_check_dups(namespaces, dir, path, table))
> +               goto fail_put_dir_locked;
> +       insert_header(dir, header);
> +       drop_sysctl_table(&dir->header);
>        spin_unlock(&sysctl_lock);
>
>        return header;
> -fail_locked:
> +fail_put_dir_locked:
> +       drop_sysctl_table(&dir->header);
>        spin_unlock(&sysctl_lock);
>  fail:
>        kfree(header);
> @@ -1299,16 +1242,17 @@ EXPORT_SYMBOL(register_sysctl_table);
>
>  static void drop_sysctl_table(struct ctl_table_header *header)
>  {
> +       struct ctl_dir *parent = header->parent;
> +
>        if (--header->nreg)
>                return;
>
>        start_unregistering(header);
> -       if (!--header->parent->count) {
> -               WARN_ON(1);
> -               kfree_rcu(header->parent, rcu);
> -       }
>        if (!--header->count)
>                kfree_rcu(header, rcu);
> +
> +       if (parent)
> +               drop_sysctl_table(&parent->header);
>  }
>
>  /**
> diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
> index e73ba33..3084b62 100644
> --- a/include/linux/sysctl.h
> +++ b/include/linux/sysctl.h
> @@ -938,6 +938,7 @@ struct ctl_table;
>  struct nsproxy;
>  struct ctl_table_root;
>  struct ctl_table_header;
> +struct ctl_dir;
>
>  typedef struct ctl_table ctl_table;
>
> @@ -1040,9 +1041,12 @@ struct ctl_table_header
>        struct ctl_table *ctl_table_arg;
>        struct ctl_table_root *root;
>        struct ctl_table_set *set;
> -       struct ctl_table *attached_by;
> -       struct ctl_table *attached_to;
> -       struct ctl_table_header *parent;
> +       struct ctl_dir *parent;
> +};
> +
> +struct ctl_dir {
> +       /* Header must be at the start of ctl_dir */
> +       struct ctl_table_header header;
>  };
>
>  struct ctl_table_set {
> --
> 1.7.2.5
>



-- 
 .
..: Lucian

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-27  4:52 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
@ 2012-01-30  0:01   ` Lucian Adrian Grijincu
  2012-01-31  3:10     ` Eric W. Biederman
  2012-01-30 17:51   ` Dave Jones
  1 sibling, 1 reply; 76+ messages in thread
From: Lucian Adrian Grijincu @ 2012-01-30  0:01 UTC (permalink / raw)
  To: Eric W. Biederman; +Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

Very nice way to handle netns specific files with links between sets!

You did a much better job than I did at dealing with them.

It took me a while to understand how the code works. I'll try to write
something for Documentation/ because the inner workings are a bit
intertwined.

A few comments bellow.

On Fri, Jan 27, 2012 at 6:52 AM, Eric W. Biederman
<ebiederm@xmission.com> wrote:
> Piecing together directories by looking first in one directory
> tree, than in another directory tree and finally in a third

than -> then

> directory tree makes it hard to verify that some directory
> entries are not multiply defined and makes it hard to create
> efficient implementations the sysctl filesystem.
>
> Replace the sysctl wide list of roots with autogenerated
> links from the core sysctl directory tree to the other
> sysctl directory trees.
>
> This simplifies sysctl directory reading and lookups as now
> only entries in a single sysctl directory tree need to be
> considered.
>
> Benchmark before:
>    make-dummies 0 999 -> 0.44s
>    rmmod dummy        -> 0.065s
>    make-dummies 0 9999 -> 1m36s
>    rmmod dummy         -> 0.4s
>
> Benchmark after:
>    make-dummies 0 999 -> 0.63s
>    rmmod dummy        -> 0.12s
>    make-dummies 0 9999 -> 2m35s
>    rmmod dummy         -> 18s
>
> The slowdown is caused by the lookups used in insert_headers

insert_headers -> insert_header

> and put_links to see if we need to add links or remove links.




>  void register_sysctl_root(struct ctl_table_root *root)
>  {
> -       spin_lock(&sysctl_lock);
> -       list_add_tail(&root->root_list, &sysctl_table_root.root_list);
> -       spin_unlock(&sysctl_lock);
>  }


This function is empty. Can be deleted (there are two callers in
net/sysctl_net.c.


> @@ -400,6 +373,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>        struct inode *inode;
>        struct dentry *err = ERR_PTR(-ENOENT);
>        struct ctl_dir *ctl_dir;
> +       int ret;
>
>        if (IS_ERR(head))
>                return ERR_CAST(head);
> @@ -410,6 +384,11 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>        if (!p)
>                goto out;
>


"sysctl_follow_link" implies that it will follow the link. I would
pull out the check whether the header is a link or not. This wouldn't
save much (a function call), but it would make the code easier to
read:

	/* Get out quickly if not a link */
        if (S_ISLNK(p->mode)) {
		ret = sysctl_follow_link(&h, &p, current->nsproxy);
		err = ERR_PTR(ret);
		if (ret)
		        goto out;
        }



> +       ret = sysctl_follow_link(&h, &p, current->nsproxy);
> +       err = ERR_PTR(ret);
> +       if (ret)
> +               goto out;
> +
>        err = ERR_PTR(-ENOMEM);
>        inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
>        if (h)
> @@ -547,6 +526,25 @@ static int proc_sys_fill_cache(struct file *filp, void *dirent,
>        return !!filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
>  }
>
> +static int proc_sys_link_fill_cache(struct file *filp, void *dirent,
> +                                   filldir_t filldir,
> +                                   struct ctl_table_header *head,
> +                                   struct ctl_table *table)
> +{
> +       int err, ret = 0;
> +       head = sysctl_head_grab(head);
> +
> +       /* It is not an error if we can not follow the link ignore it */
> +       err = sysctl_follow_link(&head, &table, current->nsproxy);

similar

> +       if (err)
> +               goto out;
> +
> +       ret = proc_sys_fill_cache(filp, dirent, filldir, head, table);
> +out:
> +       sysctl_head_finish(head);
> +       return ret;
> +}
> +




> -static struct ctl_dir *get_subdir(struct ctl_table_set *set,
> -       struct ctl_dir *dir, const char *name, int namelen)
> +static struct ctl_dir *get_subdir(struct ctl_dir *dir,
> +                                 const char *name, int namelen)
>  {
> +       struct ctl_table_set *set = dir->header.set;
>        struct ctl_dir *subdir, *new = NULL;
>
>        spin_lock(&sysctl_lock);
> -       subdir = find_subdir(dir->header.set, dir, name, namelen);
> -       if (!IS_ERR(subdir))
> -               goto found;
> -       if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)
> -               subdir = find_subdir(set, dir, name, namelen);
> +       subdir = find_subdir(dir, name, namelen);
>        if (!IS_ERR(subdir))
>                goto found;
>        if (PTR_ERR(subdir) != -ENOENT)
> @@ -817,13 +815,14 @@ static struct ctl_dir *get_subdir(struct ctl_table_set *set,
>        if (!new)
>                goto failed;
>
> -       subdir = find_subdir(set, dir, name, namelen);
> +       subdir = find_subdir(dir, name, namelen);
>        if (!IS_ERR(subdir))
>                goto found;
>        if (PTR_ERR(subdir) != -ENOENT)
>                goto failed;

I think you're returning the wrong error here. If we got to this point
then subdir == ERR_PTR(-ENOENT).
We want to create a new dir here even if one doesn't exist.
So if we have an error in insert_header() we don't return that error,
but ENOENT.

>
> -       insert_header(dir, &new->header);
> +       if (insert_header(dir, &new->header))
> +               goto failed;
>        subdir = new;

Something like this would fix it:

        if (subdir = insert_header(dir, &new->header))



>  found:
>        subdir->header.nreg++;
> @@ -841,6 +840,57 @@ failed:
>        return subdir;
>  }
>




-- 
 .
..: Lucian

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 18/29] sysctl: Normalize the root_table data structure.
  2012-01-29 17:36   ` Lucian Adrian Grijincu
@ 2012-01-30  0:22     ` Eric W. Biederman
  0 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-30  0:22 UTC (permalink / raw)
  To: Lucian Adrian Grijincu
  Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps


Lucian thanks for the review.

Lucian Adrian Grijincu <lucian.grijincu@gmail.com> writes:

> On Fri, Jan 27, 2012 at 6:51 AM, Eric W. Biederman
> <ebiederm@xmission.com> wrote:
>> Every other directory has a .child member and we look at the .child
>> for our entries.  Do the same for the root_table.
>>
>> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
>> ---
>>  fs/proc/proc_sysctl.c |   15 +++++++++++----
>>  1 files changed, 11 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
>> index 7e96a26..88d1b06 100644
>> --- a/fs/proc/proc_sysctl.c
>> +++ b/fs/proc/proc_sysctl.c
>> @@ -25,7 +25,14 @@ void proc_sys_poll_notify(struct ctl_table_poll *poll)
>>        wake_up_interruptible(&poll->wait);
>>  }
>>
>> -static struct ctl_table root_table[1];
>> +static struct ctl_table root_table[] = {
>> +       {
>> +               .procname = "",
>> +               .mode = S_IRUGO|S_IXUGO,
>
> Why not: .mode = S_IFDIR|S_IRUGO|S_IXUGO ?
>
> You change it later and add IFDIR in patch 22/29 because of this
> change (from 22/29):
> -	if (!table->child) {
> +	if (!S_ISDIR(table->mode)) {
>
> but if might as well be done here.

Actually doing not adding S_IFDIR here is important for keeping the
notion of changing one logical thing at a time.

I see normalizing this table early as deliberately emphasizing that we
started using S_IFDIR later.

>> +               .child = &root_table[1],
>> +       },
>> +       { }
>> +};
>>  static struct ctl_table_root sysctl_table_root;
>>  static struct ctl_table_header root_table_header = {
>>        {{.count = 1,
>> @@ -319,7 +326,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>>                goto out;
>>        }
>
> Somewhere above we have
>   struct ctl_table *table = PROC_I(inode)->sysctl_entry;
> and sysctl_entry can take two values:
>    - NULL: fs/proc/inode.c:proc_alloc_inode()
>    - a non-NULL table in: proc_sys_make_inode()
>
> The only inode that can be passed to proc_sys_lookup that can have
> table=NULL is the root inode.
> In that case head will be &root_table_header.
>
>
> So head->ctl_table[1] == root_table_header.ctl_table[1] == root_table[1].

Yes.

>> -       table = table ? table->child : head->ctl_table;
>> +       table = table ? table->child : &head->ctl_table[1];
>
> I think this could be improved to something like this:
> /* table == NULL only for the procfs root directory */
> table = table ? table->child : &root_table[1];
>
>
> It's not that important, as this code will go away in a few patches.

Good catch.  I had to ask myself after I read this why the code wasn't
pointing at root_table already.  The answer as it turns out was because
it was static in kernel/sysctl.c until my earlier patches and I simply
missed the opportunity.

>>        p = find_in_table(table, name);
>>        if (!p) {
>> @@ -510,7 +517,7 @@ static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
>>                goto out;
>>        }
>>
>> -       table = table ? table->child : head->ctl_table;
>> +       table = table ? table->child : &head->ctl_table[1];
>
> Similar.

Agreed.  At this point in the game since this code does go away I don't
think it is worth making the change now. But I will keep it in mind
and I you have definitely spotted an opportunity missed in this
patchset.

Eric

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-27  4:52 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
  2012-01-30  0:01   ` Lucian Adrian Grijincu
@ 2012-01-30 17:51   ` Dave Jones
  2012-01-31  2:11     ` Eric W. Biederman
  1 sibling, 1 reply; 76+ messages in thread
From: Dave Jones @ 2012-01-30 17:51 UTC (permalink / raw)
  To: Eric W. Biederman
  Cc: linux-kernel, linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps

On Thu, Jan 26, 2012 at 08:52:00PM -0800, Eric W. Biederman wrote:

 > +	memset(set, sizeof(*set), 0);

checkpatch should warn you about this mistake.

	Dave

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-30 17:51   ` Dave Jones
@ 2012-01-31  2:11     ` Eric W. Biederman
  2012-01-31  2:13       ` Joe Perches
  0 siblings, 1 reply; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-31  2:11 UTC (permalink / raw)
  To: Dave Jones
  Cc: linux-kernel, linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps

Dave Jones <davej@redhat.com> writes:

> On Thu, Jan 26, 2012 at 08:52:00PM -0800, Eric W. Biederman wrote:
>
>  > +	memset(set, sizeof(*set), 0);
>
> checkpatch should warn you about this mistake.

Is there a way to tie checkpatch into git.  Perhaps into git commit?

Having something that would catch this before I embarrassed myself
certainly sounds nice.  However my overall impression of checkpatch
is that it catches newbie mistakes so unless it is trivial to run
checkpatch I don't expect I will run it because I don't expect to
tell me anything useful.

Eric

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-31  2:11     ` Eric W. Biederman
@ 2012-01-31  2:13       ` Joe Perches
  0 siblings, 0 replies; 76+ messages in thread
From: Joe Perches @ 2012-01-31  2:13 UTC (permalink / raw)
  To: Eric W. Biederman
  Cc: Dave Jones, linux-kernel, linux-fsdevel, netdev,
	Lucian Adrian Grijincu, Damien Millescamps

On Mon, 2012-01-30 at 18:11 -0800, Eric W. Biederman wrote:
> Is there a way to tie checkpatch into git.  Perhaps into git commit?

$ git log --format=email -p -1 $commit | \
  ./scripts/checkpatch.pl -

or maybe a front-end script prior to git commit

$ git diff --stat -p <paths> | \
  ./scripts.checkpatch.pl -



^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets.
  2012-01-30  0:01   ` Lucian Adrian Grijincu
@ 2012-01-31  3:10     ` Eric W. Biederman
  0 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-31  3:10 UTC (permalink / raw)
  To: Lucian Adrian Grijincu
  Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

Lucian Adrian Grijincu <lucian.grijincu@gmail.com> writes:

> Very nice way to handle netns specific files with links between sets!
>
> You did a much better job than I did at dealing with them.
>
> It took me a while to understand how the code works. I'll try to write
> something for Documentation/ because the inner workings are a bit
> intertwined.
>
> A few comments bellow.

Thank you for your review.

I'm going to see how many of your comments I can incorporate into the
code.

Since no significant bugs have been seen my plan is to use incremental
patches on top of the existing code, to address the small issues that
have been seen.  That way I don't have to retest yet again.

>>  void register_sysctl_root(struct ctl_table_root *root)
>>  {
>> -       spin_lock(&sysctl_lock);
>> -       list_add_tail(&root->root_list, &sysctl_table_root.root_list);
>> -       spin_unlock(&sysctl_lock);
>>  }
>
>
> This function is empty. Can be deleted (there are two callers in
> net/sysctl_net.c.

Agreed.  There might be something useful by registering roots.  Perhaps
initialize the default set.  So I have kept the function for the short
term.  register_sysctl_root makes sense as an interface.  We just don't
have anything for register_sysctl_root to do right now.

>> @@ -400,6 +373,7 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>>        struct inode *inode;
>>        struct dentry *err = ERR_PTR(-ENOENT);
>>        struct ctl_dir *ctl_dir;
>> +       int ret;
>>
>>        if (IS_ERR(head))
>>                return ERR_CAST(head);
>> @@ -410,6 +384,11 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
>>        if (!p)
>>                goto out;
>>
>
>
> "sysctl_follow_link" implies that it will follow the link. I would
> pull out the check whether the header is a link or not. This wouldn't
> save much (a function call), but it would make the code easier to
> read:
>
> 	/* Get out quickly if not a link */
>         if (S_ISLNK(p->mode)) {
> 		ret = sysctl_follow_link(&h, &p, current->nsproxy);
> 		err = ERR_PTR(ret);
> 		if (ret)
> 		        goto out;
>         }

Good point, and obviousness is part of what I am aiming for.


>>  {
>> +       struct ctl_table_set *set = dir->header.set;
>>        struct ctl_dir *subdir, *new = NULL;
>>
>>        spin_lock(&sysctl_lock);
>> -       subdir = find_subdir(dir->header.set, dir, name, namelen);
>> -       if (!IS_ERR(subdir))
>> -               goto found;
>> -       if ((PTR_ERR(subdir) == -ENOENT) && set != dir->header.set)
>> -               subdir = find_subdir(set, dir, name, namelen);
>> +       subdir = find_subdir(dir, name, namelen);
>>        if (!IS_ERR(subdir))
>>                goto found;
>>        if (PTR_ERR(subdir) != -ENOENT)
>> @@ -817,13 +815,14 @@ static struct ctl_dir *get_subdir(struct ctl_table_set *set,
>>        if (!new)
>>                goto failed;
>>
>> -       subdir = find_subdir(set, dir, name, namelen);
>> +       subdir = find_subdir(dir, name, namelen);
>>        if (!IS_ERR(subdir))
>>                goto found;
>>        if (PTR_ERR(subdir) != -ENOENT)
>>                goto failed;
>
> I think you're returning the wrong error here. If we got to this point
> then subdir == ERR_PTR(-ENOENT).
> We want to create a new dir here even if one doesn't exist.
> So if we have an error in insert_header() we don't return that error,
> but ENOENT.
>
>>
>> -       insert_header(dir, &new->header);
>> +       if (insert_header(dir, &new->header))
>> +               goto failed;
>>        subdir = new;
>
> Something like this would fix it:
>
>         if (subdir = insert_header(dir, &new->header))
>
>
>
>>  found:
>>        subdir->header.nreg++;
>> @@ -841,6 +840,57 @@ failed:
>>        return subdir;
>>  }
>>

^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories
  2012-01-29 19:31   ` Lucian Adrian Grijincu
@ 2012-01-31  4:45     ` Eric W. Biederman
  0 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-01-31  4:45 UTC (permalink / raw)
  To: Lucian Adrian Grijincu
  Cc: linux-kernel, linux-fsdevel, netdev, Damien Millescamps

Lucian Adrian Grijincu <lucian.grijincu@gmail.com> writes:

>
> Why do you only check the given @set after creating the new dir. Why
> not the parent dir too?

That is a race and a minor bug.  Somehow I had convinced myself that
either it didn't matter or that it would be hard to deal with, but
that doesn't hold up.  Oops.  Thankfully the final version using links
does not have this race/bug. 

Eric


^ permalink raw reply	[flat|nested] 76+ messages in thread

* Re: [Review][PATCH][0/4] sysctl bitty fixes
  2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
                   ` (57 preceding siblings ...)
  2012-01-27  4:52 ` [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users Eric W. Biederman
@ 2012-02-02  3:27 ` Eric W. Biederman
  2012-02-02  3:28   ` [PATCH 1/4] sysctl: An easier to read version of find_subdir Eric W. Biederman
                     ` (3 more replies)
  58 siblings, 4 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-02-02  3:27 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Dan Carpenter


After digging through Lucian comments here are my incremental fixes
to address them.

This code is also available at:
kernel.org/pub/scm/linux/kernel/git/ebiederm/sysctl.git

Eric W. Biederman (4):
      sysctl: An easier to read version of find_subdir
      sysctl: Correct error return from get_subdir
      sysctl: Comments to make the code clearer.
      sysctl: Don't call sysctl_follow_link unless we are a link.

 fs/proc/proc_sysctl.c |   51 +++++++++++++++++++++++++++++++++---------------
 1 files changed, 35 insertions(+), 16 deletions(-)

^ permalink raw reply	[flat|nested] 76+ messages in thread

* [PATCH 1/4] sysctl: An easier to read version of find_subdir
  2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
@ 2012-02-02  3:28   ` Eric W. Biederman
  2012-02-02  3:29   ` [PATCH 2/4] sysctl: Correct error return from get_subdir Eric W. Biederman
                     ` (2 subsequent siblings)
  3 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-02-02  3:28 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Dan Carpenter


Suggested-by:  Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 27e265b..ebe8b30 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -833,9 +833,9 @@ static struct ctl_dir *find_subdir(struct ctl_dir *dir,
 	entry = find_entry(&head, dir, name, namelen);
 	if (!entry)
 		return ERR_PTR(-ENOENT);
-	if (S_ISDIR(entry->mode))
-		return container_of(head, struct ctl_dir, header);
-	return ERR_PTR(-ENOTDIR);
+	if (!S_ISDIR(entry->mode))
+		return ERR_PTR(-ENOTDIR);
+	return container_of(head, struct ctl_dir, header);
 }
 
 static struct ctl_dir *new_dir(struct ctl_table_set *set,
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 2/4] sysctl: Correct error return from get_subdir
  2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
  2012-02-02  3:28   ` [PATCH 1/4] sysctl: An easier to read version of find_subdir Eric W. Biederman
@ 2012-02-02  3:29   ` Eric W. Biederman
  2012-02-02  3:29   ` [PATCH 3/4] sysctl: Comments to make the code clearer Eric W. Biederman
  2012-02-02  3:30   ` [PATCH 4/4] sysctl: Don't call sysctl_follow_link unless we are a link Eric W. Biederman
  3 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-02-02  3:29 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Dan Carpenter


When insert_header fails ensure we return the proper error value
from get_subdir.  In practice nothing cares, but there is no
need to be sloppy.

Reported-by: Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index ebe8b30..722ec11 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -869,6 +869,7 @@ static struct ctl_dir *get_subdir(struct ctl_dir *dir,
 {
 	struct ctl_table_set *set = dir->header.set;
 	struct ctl_dir *subdir, *new = NULL;
+	int err;
 
 	spin_lock(&sysctl_lock);
 	subdir = find_subdir(dir, name, namelen);
@@ -890,7 +891,9 @@ static struct ctl_dir *get_subdir(struct ctl_dir *dir,
 	if (PTR_ERR(subdir) != -ENOENT)
 		goto failed;
 
-	if (insert_header(dir, &new->header))
+	err = insert_header(dir, &new->header);
+	subdir = ERR_PTR(err);
+	if (err)
 		goto failed;
 	subdir = new;
 found:
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 3/4] sysctl: Comments to make the code clearer.
  2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
  2012-02-02  3:28   ` [PATCH 1/4] sysctl: An easier to read version of find_subdir Eric W. Biederman
  2012-02-02  3:29   ` [PATCH 2/4] sysctl: Correct error return from get_subdir Eric W. Biederman
@ 2012-02-02  3:29   ` Eric W. Biederman
  2012-02-02 16:28     ` Ben Hutchings
  2012-02-02  3:30   ` [PATCH 4/4] sysctl: Don't call sysctl_follow_link unless we are a link Eric W. Biederman
  3 siblings, 1 reply; 76+ messages in thread
From: Eric W. Biederman @ 2012-02-02  3:29 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Dan Carpenter


Document get_subdir and that find_subdir alwasy takes a reference.

Suggested-by:  Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   16 ++++++++++++++++
 1 files changed, 16 insertions(+), 0 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index 722ec11..e5601dc 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -73,6 +73,7 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
 	return cmp;
 }
 
+/* Called under sysctl_lock */
 static struct ctl_table *find_entry(struct ctl_table_header **phead,
 	struct ctl_dir *dir, const char *name, int namelen)
 {
@@ -864,6 +865,18 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
 	return new;
 }
 
+/**
+ * get_subdir - find or create a subdir with the specified name.
+ * @dir:  Directory to create the subdirectory in
+ * @name: The name of the subdirectory to find or create
+ * @namelen: The length of name
+ *
+ * Takes a directory with an elevated reference count so we know that
+ * if we drop the lock the directory will not go away.  Upon success
+ * the reference is moved from @dir to the returned subdirectory.
+ * Upon error an error code is returned and the reference on @dir is
+ * simply dropped.
+ */
 static struct ctl_dir *get_subdir(struct ctl_dir *dir,
 				  const char *name, int namelen)
 {
@@ -885,12 +898,14 @@ static struct ctl_dir *get_subdir(struct ctl_dir *dir,
 	if (!new)
 		goto failed;
 
+	/* Was the subdir added while we dropped the lock? */
 	subdir = find_subdir(dir, name, namelen);
 	if (!IS_ERR(subdir))
 		goto found;
 	if (PTR_ERR(subdir) != -ENOENT)
 		goto failed;
 
+	/* Nope.  Use the our freshly made directory entry. */
 	err = insert_header(dir, &new->header);
 	subdir = ERR_PTR(err);
 	if (err)
@@ -1190,6 +1205,7 @@ struct ctl_table_header *__register_sysctl_table(
 
 	spin_lock(&sysctl_lock);
 	dir = &set->dir;
+	/* Reference moved down the diretory tree get_subdir */
 	dir->header.nreg++;
 	spin_unlock(&sysctl_lock);
 
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* [PATCH 4/4] sysctl: Don't call sysctl_follow_link unless we are a link.
  2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
                     ` (2 preceding siblings ...)
  2012-02-02  3:29   ` [PATCH 3/4] sysctl: Comments to make the code clearer Eric W. Biederman
@ 2012-02-02  3:30   ` Eric W. Biederman
  3 siblings, 0 replies; 76+ messages in thread
From: Eric W. Biederman @ 2012-02-02  3:30 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Dan Carpenter


There are no functional changes.  Just code motion to make it
clear that we don't follow a link between sysctl roots unless the
directory entry actually is a link.

Suggested-by:  Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
 fs/proc/proc_sysctl.c |   24 ++++++++++++------------
 1 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
index e5601dc..a7708b7 100644
--- a/fs/proc/proc_sysctl.c
+++ b/fs/proc/proc_sysctl.c
@@ -451,10 +451,12 @@ static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
 	if (!p)
 		goto out;
 
-	ret = sysctl_follow_link(&h, &p, current->nsproxy);
-	err = ERR_PTR(ret);
-	if (ret)
-		goto out;
+	if (S_ISLNK(p->mode)) {
+		ret = sysctl_follow_link(&h, &p, current->nsproxy);
+		err = ERR_PTR(ret);
+		if (ret)
+			goto out;
+	}
 
 	err = ERR_PTR(-ENOMEM);
 	inode = proc_sys_make_inode(dir->i_sb, h ? h : head, p);
@@ -601,10 +603,12 @@ static int proc_sys_link_fill_cache(struct file *filp, void *dirent,
 	int err, ret = 0;
 	head = sysctl_head_grab(head);
 
-	/* It is not an error if we can not follow the link ignore it */
-	err = sysctl_follow_link(&head, &table, current->nsproxy);
-	if (err)
-		goto out;
+	if (S_ISLNK(table->mode)) {
+		/* It is not an error if we can not follow the link ignore it */
+		err = sysctl_follow_link(&head, &table, current->nsproxy);
+		if (err)
+			goto out;
+	}
 
 	ret = proc_sys_fill_cache(filp, dirent, filldir, head, table);
 out:
@@ -950,10 +954,6 @@ static int sysctl_follow_link(struct ctl_table_header **phead,
 	struct ctl_dir *dir;
 	int ret;
 
-	/* Get out quickly if not a link */
-	if (!S_ISLNK((*pentry)->mode))
-		return 0;
-
 	ret = 0;
 	spin_lock(&sysctl_lock);
 	root = (*pentry)->data;
-- 
1.7.2.5


^ permalink raw reply related	[flat|nested] 76+ messages in thread

* Re: [PATCH 3/4] sysctl: Comments to make the code clearer.
  2012-02-02  3:29   ` [PATCH 3/4] sysctl: Comments to make the code clearer Eric W. Biederman
@ 2012-02-02 16:28     ` Ben Hutchings
  0 siblings, 0 replies; 76+ messages in thread
From: Ben Hutchings @ 2012-02-02 16:28 UTC (permalink / raw)
  To: Eric W. Biederman
  Cc: linux-kernel, linux-fsdevel, netdev, Lucian Adrian Grijincu,
	Damien Millescamps, Dan Carpenter

On Wed, 2012-02-01 at 19:29 -0800, Eric W. Biederman wrote:
> Document get_subdir and that find_subdir alwasy takes a reference.
> 
> Suggested-by:  Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
> ---
>  fs/proc/proc_sysctl.c |   16 ++++++++++++++++
>  1 files changed, 16 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index 722ec11..e5601dc 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -73,6 +73,7 @@ static int namecmp(const char *name1, int len1, const char *name2, int len2)
>  	return cmp;
>  }
>  
> +/* Called under sysctl_lock */
>  static struct ctl_table *find_entry(struct ctl_table_header **phead,
>  	struct ctl_dir *dir, const char *name, int namelen)
>  {
> @@ -864,6 +865,18 @@ static struct ctl_dir *new_dir(struct ctl_table_set *set,
>  	return new;
>  }
>  
> +/**
> + * get_subdir - find or create a subdir with the specified name.
> + * @dir:  Directory to create the subdirectory in
> + * @name: The name of the subdirectory to find or create
> + * @namelen: The length of name
> + *
> + * Takes a directory with an elevated reference count so we know that
> + * if we drop the lock the directory will not go away.  Upon success
> + * the reference is moved from @dir to the returned subdirectory.
> + * Upon error an error code is returned and the reference on @dir is
> + * simply dropped.
> + */
>  static struct ctl_dir *get_subdir(struct ctl_dir *dir,
>  				  const char *name, int namelen)
[...]

Should static functions really have kernel-doc?  (Aside from static
inlines defined in a header.)  This can pollute the documentation
namespace.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.


^ permalink raw reply	[flat|nested] 76+ messages in thread

end of thread, other threads:[~2012-02-02 16:28 UTC | newest]

Thread overview: 76+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-01-27  4:45 [Review][PATCH][0/29] sysctl rewrite for speed and clarity Eric W. Biederman
2012-01-27  4:49 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
2012-01-27  4:49 ` [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling Eric W. Biederman
2012-01-27  4:49 ` [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table Eric W. Biederman
2012-01-27  4:49 ` [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c Eric W. Biederman
2012-01-27  4:49 ` [PATCH 05/29] sysctl: Make the directories have nlink == 1 Eric W. Biederman
2012-01-27  4:49 ` [PATCH 06/29] sysctl: Implement retire_sysctl_set Eric W. Biederman
2012-01-27  4:49 ` [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept Eric W. Biederman
2012-01-27  4:49 ` [PATCH 08/29] sysctl: Create local copies of directory names used in paths Eric W. Biederman
2012-01-27  4:49 ` [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path Eric W. Biederman
2012-01-27  4:49 ` [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths Eric W. Biederman
2012-01-27  4:49 ` [PATCH 11/29] sysctl: register only tables of sysctl files Eric W. Biederman
2012-01-27  4:49 ` [PATCH 12/29] sysctl: Improve the sysctl sanity checks Eric W. Biederman
2012-01-27  4:49 ` [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field Eric W. Biederman
2012-01-27  4:49 ` [PATCH 14/29] sysctl: A more obvious version of grab_header Eric W. Biederman
2012-01-27  4:49 ` [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables Eric W. Biederman
2012-01-27  4:49 ` [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths Eric W. Biederman
2012-01-27  4:49 ` [PATCH 17/29] sysctl: Factor out insert_header and erase_header Eric W. Biederman
2012-01-27  4:49 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
2012-01-27  4:50 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
2012-01-27  4:50 ` [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry Eric W. Biederman
2012-01-27  4:50 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
2012-01-27  4:50 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
2012-01-27  4:50 ` [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir Eric W. Biederman
2012-01-27  4:50 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
2012-01-27  4:50 ` [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy Eric W. Biederman
2012-01-27  4:50 ` [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header Eric W. Biederman
2012-01-27  4:50 ` [PATCH 27/29] sysctl: Make the header lists per directory Eric W. Biederman
2012-01-27  4:50 ` [PATCH 28/29] sysctl: Index sysctl directories with rbtrees Eric W. Biederman
2012-01-27  4:50 ` [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users Eric W. Biederman
2012-01-27  4:51 ` [PATCH 01/29] sysctl: remove impossible condition check Eric W. Biederman
2012-01-27  4:51 ` [PATCH 02/29] sysctl: Consolidate !CONFIG_SYSCTL handling Eric W. Biederman
2012-01-27  4:51 ` [PATCH 03/29] sysctl: Register the base sysctl table like any other sysctl table Eric W. Biederman
2012-01-27  4:51 ` [PATCH 04/29] sysctl: Move the implementation into fs/proc/proc_sysctl.c Eric W. Biederman
2012-01-27  4:51 ` [PATCH 05/29] sysctl: Make the directories have nlink == 1 Eric W. Biederman
2012-01-27  4:51 ` [PATCH 06/29] sysctl: Implement retire_sysctl_set Eric W. Biederman
2012-01-27  4:51 ` [PATCH 07/29] sysctl: Remove the unnecessary sysctl_set parent concept Eric W. Biederman
2012-01-27  4:51 ` [PATCH 08/29] sysctl: Create local copies of directory names used in paths Eric W. Biederman
2012-01-27  4:51 ` [PATCH 09/29] sysctl: Add support for register sysctl tables with a normal cstring path Eric W. Biederman
2012-01-27  4:51 ` [PATCH 10/29] sysctl: Add ctl_table chains into cstring paths Eric W. Biederman
2012-01-27  4:51 ` [PATCH 11/29] sysctl: register only tables of sysctl files Eric W. Biederman
2012-01-27  4:51 ` [PATCH 12/29] sysctl: Improve the sysctl sanity checks Eric W. Biederman
2012-01-27  4:51 ` [PATCH 13/29] sysctl: Remove the now unused ctl_table parent field Eric W. Biederman
2012-01-27  4:51 ` [PATCH 14/29] sysctl: A more obvious version of grab_header Eric W. Biederman
2012-01-27  4:51 ` [PATCH 15/29] sysctl: Initial support for auto-unregistering sysctl tables Eric W. Biederman
2012-01-27  4:51 ` [PATCH 16/29] sysctl: Factor out init_header from __register_sysctl_paths Eric W. Biederman
2012-01-27  4:51 ` [PATCH 17/29] sysctl: Factor out insert_header and erase_header Eric W. Biederman
2012-01-27  4:51 ` [PATCH 18/29] sysctl: Normalize the root_table data structure Eric W. Biederman
2012-01-29 17:36   ` Lucian Adrian Grijincu
2012-01-30  0:22     ` Eric W. Biederman
2012-01-27  4:51 ` [PATCH 19/29] sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry Eric W. Biederman
2012-01-29 15:49   ` Lucian Adrian Grijincu
2012-01-27  4:51 ` [PATCH 20/29] sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry Eric W. Biederman
2012-01-27  4:51 ` [PATCH 21/29] sysctl: Add a root pointer to ctl_table_set Eric W. Biederman
2012-01-29 17:19   ` Lucian Adrian Grijincu
2012-01-27  4:51 ` [PATCH 22/29] sysctl: Stop requiring explicit management of sysctl directories Eric W. Biederman
2012-01-29 19:31   ` Lucian Adrian Grijincu
2012-01-31  4:45     ` Eric W. Biederman
2012-01-27  4:51 ` [PATCH 23/29] sysctl: Add sysctl_print_dir and use it in get_subdir Eric W. Biederman
2012-01-27  4:52 ` [PATCH 24/29] sysctl: Replace root_list with links between sysctl_table_sets Eric W. Biederman
2012-01-30  0:01   ` Lucian Adrian Grijincu
2012-01-31  3:10     ` Eric W. Biederman
2012-01-30 17:51   ` Dave Jones
2012-01-31  2:11     ` Eric W. Biederman
2012-01-31  2:13       ` Joe Perches
2012-01-27  4:52 ` [PATCH 25/29] sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy Eric W. Biederman
2012-01-27  4:52 ` [PATCH 26/29] sysctl: Move sysctl_check_dups into insert_header Eric W. Biederman
2012-01-27  4:52 ` [PATCH 27/29] sysctl: Make the header lists per directory Eric W. Biederman
2012-01-27  4:52 ` [PATCH 28/29] sysctl: Index sysctl directories with rbtrees Eric W. Biederman
2012-01-27  4:52 ` [PATCH 29/29] sysctl: Add register_sysctl for normal sysctl users Eric W. Biederman
2012-02-02  3:27 ` [Review][PATCH][0/4] sysctl bitty fixes Eric W. Biederman
2012-02-02  3:28   ` [PATCH 1/4] sysctl: An easier to read version of find_subdir Eric W. Biederman
2012-02-02  3:29   ` [PATCH 2/4] sysctl: Correct error return from get_subdir Eric W. Biederman
2012-02-02  3:29   ` [PATCH 3/4] sysctl: Comments to make the code clearer Eric W. Biederman
2012-02-02 16:28     ` Ben Hutchings
2012-02-02  3:30   ` [PATCH 4/4] sysctl: Don't call sysctl_follow_link unless we are a link Eric W. Biederman

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).