Skip to contentSkip to conversation

proc: stop using BKL

PATCH In review Alexey Dobriyan 2 participants 3 messages Oct 24, 2008 Proc filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 proc: stop using BKL

[added to proc.git]

commit e137447fbf30784a86b79707ef0d76de8c6bcf8d

Author: Alexey Dobriyan <[email protected]>
Date:   Fri Oct 24 15:30:36 2008 +0400

proc: stop using BKL

There are four BKL users in proc: de_put(), proc_lookup_de(), proc_readdir_de(), proc_root_readdir(),

1) de_put()
-----------
de_put() is classic atomic_dec_and_test() refcount wrapper -- no BKL needed. BKL doesn't matter to possible refcount leak as well.

2) proc_lookup_de()
-------------------
Walking PDE list is protected by proc_subdir_lock(), proc_get_inode() is potentially blocking, all callers of proc_lookup_de() eventually end up from ->lookup hooks which is protected by directory's ->i_mutex -- BKL doesn't protect anything.

3) proc_readdir_de()
--------------------
"." and ".." part doesn't need BKL, walking PDE list is under proc_subdir_lock, calling filldir callback is potentially blocking because it writes to luserspace. All proc_readdir_de() callers eventually come from ->readdir hook which is under directory's ->i_mutex -- BKL doesn't protect anything.

4) proc_root_readdir_de()
-------------------------
proc_root_readdir_de is ->readdir hook, see (3).

fs/proc/generic.c +1 -6
index 60a359b..988a807 100644
@@ -14,7 +14,6 @@
 #include <linux/stat.h>
 #include <linux/module.h>
 #include <linux/mount.h>
-#include <linux/smp_lock.h>
 #include <linux/init.h>
 #include <linux/idr.h>
 #include <linux/namei.h>
@@ -379,7 +378,6 @@ struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
 	struct inode *inode = NULL;
 	int error = -ENOENT;
 
-	lock_kernel();
 	spin_lock(&proc_subdir_lock);
 	for (de = de->subdir; de ; de = de->next) {
 		if (de->namelen != dentry->d_name.len)
@@ -397,7 +395,6 @@ struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
 	}
 	spin_unlock(&proc_subdir_lock);
 out_unlock:
-	unlock_kernel();
 
 	if (inode) {
 		dentry->d_op = &proc_dentry_operations;
@@ -432,8 +429,6 @@ int proc_readdir_de(struct proc_dir_entry *de, struct file *filp, void *dirent,
 	struct inode *inode = filp->f_path.dentry->d_inode;
 	int ret = 0;
 
-	lock_kernel();
-
 	ino = inode->i_ino;
 	i = filp->f_pos;
 	switch (i) {
@@ -487,7 +482,7 @@ int proc_readdir_de(struct proc_dir_entry *de, struct file *filp, void *dirent,
 			spin_unlock(&proc_subdir_lock);
 	}
 	ret = 1;
-out:	unlock_kernel();
+out:
 	return ret;	
 }
 
fs/proc/inode.c +0 -3
index 2543fd0..3e76bb9 100644
@@ -35,16 +35,13 @@ struct proc_dir_entry *de_get(struct proc_dir_entry *de)
  */
 void de_put(struct proc_dir_entry *de)
 {
-	lock_kernel();
 	if (!atomic_read(&de->count)) {
 		printk("de_put: entry %s already free!\n", de->name);
-		unlock_kernel();
 		return;
 	}
 
 	if (atomic_dec_and_test(&de->count))
 		free_proc_entry(de);
-	unlock_kernel();
 }
 
 /*
fs/proc/proc_net.c +0 -1
index 7bc296f..a7bd61e 100644
@@ -18,7 +18,6 @@
 #include <linux/sched.h>
 #include <linux/module.h>
 #include <linux/bitops.h>
-#include <linux/smp_lock.h>
 #include <linux/mount.h>
 #include <linux/nsproxy.h>
 #include <net/net_namespace.h>
fs/proc/root.c +1 -7
index 7761602..f6299a2 100644
@@ -16,7 +16,6 @@
 #include <linux/sched.h>
 #include <linux/module.h>
 #include <linux/bitops.h>
-#include <linux/smp_lock.h>
 #include <linux/mount.h>
 #include <linux/pid_namespace.h>
 
@@ -162,17 +161,12 @@ static int proc_root_readdir(struct file * filp,
 	unsigned int nr = filp->f_pos;
 	int ret;
 
-	lock_kernel();
-
 	if (nr < FIRST_PROCESS_ENTRY) {
 		int error = proc_readdir(filp, dirent, filldir);
-		if (error <= 0) {
-			unlock_kernel();
+		if (error <= 0)
 			return error;
-		}
 		filp->f_pos = FIRST_PROCESS_ENTRY;
 	}
-	unlock_kernel();
 
 	ret = proc_pid_readdir(filp, dirent, filldir);
 	return ret;
#1 lore raw
Quoted 8 lines · Alexey Dobriyan

3) proc_readdir_de()
--------------------
"." and ".." part doesn't need BKL, walking PDE list is under proc_subdir_lock, calling filldir callback is potentially blocking because it writes to luserspace. All proc_readdir_de() callers eventually come from ->readdir hook which is under directory's ->i_mutex -- BKL doesn't protect anything.

Quoted 5 lines

4) proc_root_readdir_de()
-------------------------
proc_root_readdir_de is ->readdir hook, see (3).

Once you stop taking BKL in readdir you also need to switch to generic_file_llseek for those directories. It would be good if you could switch the fallback proc_reg_llseek to no llseek at all in proc_reg_llseek while you're at it. I'll prepare a similar change for non-procfs fops instances.

#2 lore raw
Quoted 9 lines · Christoph Hellwig
Quoted 8 lines · Alexey Dobriyan

3) proc_readdir_de()
--------------------
"." and ".." part doesn't need BKL, walking PDE list is under proc_subdir_lock, calling filldir callback is potentially blocking because it writes to luserspace. All proc_readdir_de() callers eventually come from ->readdir hook which is under directory's ->i_mutex -- BKL doesn't protect anything.

Quoted 8 lines
Quoted 5 lines

4) proc_root_readdir_de()
-------------------------
proc_root_readdir_de is ->readdir hook, see (3).

Once you stop taking BKL in readdir you also need to switch to generic_file_llseek for those directories.

Hm, indeed. here is second version:

commit af22e6782098da2da2bd7e7428a7327f821cb157

Author: Alexey Dobriyan <[email protected]>
Date:   Fri Oct 24 23:32:05 2008 +0400

proc: stop using BKL

There are four BKL users in proc: de_put(), proc_lookup_de(), proc_readdir_de(), proc_root_readdir(),

1) de_put()
-----------
de_put() is classic atomic_dec_and_test() refcount wrapper -- no BKL needed. BKL doesn't matter to possible refcount leak as well.

2) proc_lookup_de()
-------------------
Walking PDE list is protected by proc_subdir_lock(), proc_get_inode() is potentially blocking, all callers of proc_lookup_de() eventually end up from ->lookup hooks which is protected by directory's ->i_mutex -- BKL doesn't protect anything.

3) proc_readdir_de()
--------------------
"." and ".." part doesn't need BKL, walking PDE list is under proc_subdir_lock, calling filldir callback is potentially blocking because it writes to luserspace. All proc_readdir_de() callers eventually come from ->readdir hook which is under directory's ->i_mutex -- BKL doesn't protect anything.

4) proc_root_readdir_de()
-------------------------
proc_root_readdir_de is ->readdir hook, see (3).

Since readdir hooks doesn't use BKL anymore, switch to generic_file_llseek, since it also takes directory's i_mutex.

fs/proc/generic.c +2 -6
index 60a359b..db7fa5c 100644
@@ -14,7 +14,6 @@
 #include <linux/stat.h>
 #include <linux/module.h>
 #include <linux/mount.h>
-#include <linux/smp_lock.h>
 #include <linux/init.h>
 #include <linux/idr.h>
 #include <linux/namei.h>
@@ -379,7 +378,6 @@ struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
 	struct inode *inode = NULL;
 	int error = -ENOENT;
 
-	lock_kernel();
 	spin_lock(&proc_subdir_lock);
 	for (de = de->subdir; de ; de = de->next) {
 		if (de->namelen != dentry->d_name.len)
@@ -397,7 +395,6 @@ struct dentry *proc_lookup_de(struct proc_dir_entry *de, struct inode *dir,
 	}
 	spin_unlock(&proc_subdir_lock);
 out_unlock:
-	unlock_kernel();
 
 	if (inode) {
 		dentry->d_op = &proc_dentry_operations;
@@ -432,8 +429,6 @@ int proc_readdir_de(struct proc_dir_entry *de, struct file *filp, void *dirent,
 	struct inode *inode = filp->f_path.dentry->d_inode;
 	int ret = 0;
 
-	lock_kernel();
-
 	ino = inode->i_ino;
 	i = filp->f_pos;
 	switch (i) {
@@ -487,7 +482,7 @@ int proc_readdir_de(struct proc_dir_entry *de, struct file *filp, void *dirent,
 			spin_unlock(&proc_subdir_lock);
 	}
 	ret = 1;
-out:	unlock_kernel();
+out:
 	return ret;	
 }
 
@@ -504,6 +499,7 @@ int proc_readdir(struct file *filp, void *dirent, filldir_t filldir)
  * the /proc directory.
  */
 static const struct file_operations proc_dir_operations = {
+	.llseek			= generic_file_llseek,
 	.read			= generic_read_dir,
 	.readdir		= proc_readdir,
 };
fs/proc/inode.c +0 -3
index 2543fd0..3e76bb9 100644
@@ -35,16 +35,13 @@ struct proc_dir_entry *de_get(struct proc_dir_entry *de)
  */
 void de_put(struct proc_dir_entry *de)
 {
-	lock_kernel();
 	if (!atomic_read(&de->count)) {
 		printk("de_put: entry %s already free!\n", de->name);
-		unlock_kernel();
 		return;
 	}
 
 	if (atomic_dec_and_test(&de->count))
 		free_proc_entry(de);
-	unlock_kernel();
 }
 
 /*
fs/proc/proc_net.c +1 -1
index 7bc296f..04d1270 100644
@@ -18,7 +18,6 @@
 #include <linux/sched.h>
 #include <linux/module.h>
 #include <linux/bitops.h>
-#include <linux/smp_lock.h>
 #include <linux/mount.h>
 #include <linux/nsproxy.h>
 #include <net/net_namespace.h>
@@ -172,6 +171,7 @@ static int proc_tgid_net_readdir(struct file *filp, void *dirent,
 }
 
 const struct file_operations proc_net_operations = {
+	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
 	.readdir	= proc_tgid_net_readdir,
 };
fs/proc/root.c +1 -7
index 7761602..f6299a2 100644
@@ -16,7 +16,6 @@
 #include <linux/sched.h>
 #include <linux/module.h>
 #include <linux/bitops.h>
-#include <linux/smp_lock.h>
 #include <linux/mount.h>
 #include <linux/pid_namespace.h>
 
@@ -162,17 +161,12 @@ static int proc_root_readdir(struct file * filp,
 	unsigned int nr = filp->f_pos;
 	int ret;
 
-	lock_kernel();
-
 	if (nr < FIRST_PROCESS_ENTRY) {
 		int error = proc_readdir(filp, dirent, filldir);
-		if (error <= 0) {
-			unlock_kernel();
+		if (error <= 0)
 			return error;
-		}
 		filp->f_pos = FIRST_PROCESS_ENTRY;
 	}
-	unlock_kernel();
 
 	ret = proc_pid_readdir(filp, dirent, filldir);
 	return ret;