procfs: switch /proc/self away from proc_dir_entry

Just have it pinned in dcache all along and let procfs ->kill_sb()
drop it before kill_anon_super().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

diff --git a/fs/proc/base.c b/fs/proc/base.c
index 69078c7..593e7c5 100644 (file)
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2794,7 +2794,7 @@ retry:
        return iter;
 }
 
-#define TGID_OFFSET (FIRST_PROCESS_ENTRY)
+#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 1)
 
 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
        struct tgid_iter iter)
@@ -2817,13 +2817,21 @@ int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
        struct tgid_iter iter;
        struct pid_namespace *ns;
        filldir_t __filldir;
+       loff_t pos = filp->f_pos;
 
-       if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
+       if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
                goto out;
 
-       ns = filp->f_dentry->d_sb->s_fs_info;
+       if (pos == TGID_OFFSET - 1) {
+               if (proc_fill_cache(filp, dirent, filldir, "self", 4,
+                                       NULL, NULL, NULL) < 0)
+                       goto out;
+               iter.tgid = 0;
+       } else {
+               iter.tgid = pos - TGID_OFFSET;
+       }
        iter.task = NULL;
-       iter.tgid = filp->f_pos - TGID_OFFSET;
+       ns = filp->f_dentry->d_sb->s_fs_info;
        for (iter = next_tgid(ns, iter);
             iter.task;
             iter.tgid += 1, iter = next_tgid(ns, iter)) {

diff --git a/fs/proc/inode.c b/fs/proc/inode.c
index 869116c..908e974 100644 (file)
--- a/fs/proc/inode.c
+++ b/fs/proc/inode.c
@@ -506,5 +506,5 @@ int proc_fill_super(struct super_block *s)
                return -ENOMEM;
        }
 
-       return 0;
+       return proc_setup_self(s);
 }

diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 85ff3a4..9c93a53 100644 (file)
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -205,3 +205,4 @@ int proc_setattr(struct dentry *dentry, struct iattr *attr);
 extern const struct inode_operations proc_ns_dir_inode_operations;
 extern const struct file_operations proc_ns_dir_operations;
 
+extern int proc_setup_self(struct super_block *);

diff --git a/fs/proc/root.c b/fs/proc/root.c
index c6e9fac..20834b3 100644 (file)
--- a/fs/proc/root.c
+++ b/fs/proc/root.c
@@ -137,6 +137,8 @@ static void proc_kill_sb(struct super_block *sb)
        struct pid_namespace *ns;
 
        ns = (struct pid_namespace *)sb->s_fs_info;
+       if (ns->proc_self)
+               dput(ns->proc_self);
        kill_anon_super(sb);
        put_pid_ns(ns);
 }

diff --git a/fs/proc/self.c b/fs/proc/self.c
index d8a0252..21940d8 100644 (file)
--- a/fs/proc/self.c
+++ b/fs/proc/self.c
@@ -1,6 +1,7 @@
-#include <linux/proc_fs.h>
 #include <linux/sched.h>
 #include <linux/namei.h>
+#include <linux/pid_namespace.h>
+#include "internal.h"
 
 /*
  * /proc/self:
@@ -48,12 +49,43 @@ static const struct inode_operations proc_self_inode_operations = {
        .put_link       = proc_self_put_link,
 };
 
-void __init proc_self_init(void)
+static unsigned self_inum;
+
+int proc_setup_self(struct super_block *s)
 {
-       struct proc_dir_entry *proc_self_symlink;
-       umode_t mode;
+       struct inode *root_inode = s->s_root->d_inode;
+       struct pid_namespace *ns = s->s_fs_info;
+       struct dentry *self;
+       
+       mutex_lock(&root_inode->i_mutex);
+       self = d_alloc_name(s->s_root, "self");
+       if (self) {
+               struct inode *inode = new_inode_pseudo(s);
+               if (inode) {
+                       inode->i_ino = self_inum;
+                       inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
+                       inode->i_mode = S_IFLNK | S_IRWXUGO;
+                       inode->i_uid = GLOBAL_ROOT_UID;
+                       inode->i_gid = GLOBAL_ROOT_GID;
+                       inode->i_op = &proc_self_inode_operations;
+                       d_add(self, inode);
+               } else {
+                       dput(self);
+                       self = ERR_PTR(-ENOMEM);
+               }
+       } else {
+               self = ERR_PTR(-ENOMEM);
+       }
+       mutex_unlock(&root_inode->i_mutex);
+       if (IS_ERR(self)) {
+               pr_err("proc_fill_super: can't allocate /proc/self\n");
+               return PTR_ERR(self);
+       }
+       ns->proc_self = self;
+       return 0;
+}
 
-       mode = S_IFLNK | S_IRWXUGO;
-       proc_self_symlink = proc_create("self", mode, NULL, NULL );
-       proc_self_symlink->proc_iops = &proc_self_inode_operations;
+void __init proc_self_init(void)
+{
+       proc_alloc_inum(&self_inum);
 }

diff --git a/include/linux/pid_namespace.h b/include/linux/pid_namespace.h
index 215e5e3..5524f8c 100644 (file)
--- a/include/linux/pid_namespace.h
+++ b/include/linux/pid_namespace.h
@@ -28,6 +28,7 @@ struct pid_namespace {
        struct pid_namespace *parent;
 #ifdef CONFIG_PROC_FS
        struct vfsmount *proc_mnt;
+       struct dentry *proc_self;
 #endif
 #ifdef CONFIG_BSD_PROCESS_ACCT
        struct bsd_acct_struct *bacct;