autofs4: allow autofs to work outside the initial PID namespace
authorSukadev Bhattiprolu <sukadev@us.ibm.com>
Thu, 23 Jan 2014 23:54:57 +0000 (15:54 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 24 Jan 2014 00:36:58 +0000 (16:36 -0800)
Enable autofs4 to work in a "container".  oz_pgrp is converted from
pid_t to struct pid and this is stored at mount time based on the
"pgrp=" option or if the option is missing then the current pgrp.

The "pgrp=" option is interpreted in the PID namespace of the current
process.  This option is flawed in that it doesn't carry the namespace
information, so it should be deprecated.  AFAICS the autofs daemon
always sends the current pgrp, which is the default anyway.

The oz_pgrp is also set from the AUTOFS_DEV_IOCTL_SETPIPEFD_CMD ioctl.
This ioctl sets oz_pgrp to the current pgrp.  It is not allowed to
change the pid namespace.

oz_pgrp is used mainly to determine whether the process traversing the
autofs mount tree is the autofs daemon itself or not.  This function now
compares the pid pointers instead of the pid_t values.

One other use of oz_pgrp is in autofs4_show_options.  There is shows the
virtual pid number (i.e.  the one that is valid inside the PID namespace
of the calling process)

For debugging printk convert oz_pgrp to the value in the initial pid
namespace.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Acked-by: Ian Kent <raven@themaw.net>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/autofs4/autofs_i.h
fs/autofs4/dev-ioctl.c
fs/autofs4/inode.c

index 4218e26..acf3205 100644 (file)
@@ -104,7 +104,7 @@ struct autofs_sb_info {
        u32 magic;
        int pipefd;
        struct file *pipe;
-       pid_t oz_pgrp;
+       struct pid *oz_pgrp;
        int catatonic;
        int version;
        int sub_version;
@@ -140,7 +140,7 @@ static inline struct autofs_info *autofs4_dentry_ino(struct dentry *dentry)
    filesystem without "magic".) */
 
 static inline int autofs4_oz_mode(struct autofs_sb_info *sbi) {
-       return sbi->catatonic || task_pgrp_nr(current) == sbi->oz_pgrp;
+       return sbi->catatonic || task_pgrp(current) == sbi->oz_pgrp;
 }
 
 /* Does a dentry have some pending activity? */
index 1818ce7..3182c0e 100644 (file)
@@ -346,6 +346,7 @@ static int autofs_dev_ioctl_setpipefd(struct file *fp,
 {
        int pipefd;
        int err = 0;
+       struct pid *new_pid = NULL;
 
        if (param->setpipefd.pipefd == -1)
                return -EINVAL;
@@ -357,7 +358,17 @@ static int autofs_dev_ioctl_setpipefd(struct file *fp,
                mutex_unlock(&sbi->wq_mutex);
                return -EBUSY;
        } else {
-               struct file *pipe = fget(pipefd);
+               struct file *pipe;
+
+               new_pid = get_task_pid(current, PIDTYPE_PGID);
+
+               if (ns_of_pid(new_pid) != ns_of_pid(sbi->oz_pgrp)) {
+                       AUTOFS_WARN("Not allowed to change PID namespace");
+                       err = -EINVAL;
+                       goto out;
+               }
+
+               pipe = fget(pipefd);
                if (!pipe) {
                        err = -EBADF;
                        goto out;
@@ -367,12 +378,13 @@ static int autofs_dev_ioctl_setpipefd(struct file *fp,
                        fput(pipe);
                        goto out;
                }
-               sbi->oz_pgrp = task_pgrp_nr(current);
+               swap(sbi->oz_pgrp, new_pid);
                sbi->pipefd = pipefd;
                sbi->pipe = pipe;
                sbi->catatonic = 0;
        }
 out:
+       put_pid(new_pid);
        mutex_unlock(&sbi->wq_mutex);
        return err;
 }
index 3b9cc9b..a3de082 100644 (file)
@@ -56,8 +56,11 @@ void autofs4_kill_sb(struct super_block *sb)
         * just call kill_anon_super when we are called from
         * deactivate_super.
         */
-       if (sbi) /* Free wait queues, close pipe */
+       if (sbi) {
+               /* Free wait queues, close pipe */
                autofs4_catatonic_mode(sbi);
+               put_pid(sbi->oz_pgrp);
+       }
 
        DPRINTK("shutting down");
        kill_litter_super(sb);
@@ -80,7 +83,7 @@ static int autofs4_show_options(struct seq_file *m, struct dentry *root)
        if (!gid_eq(root_inode->i_gid, GLOBAL_ROOT_GID))
                seq_printf(m, ",gid=%u",
                        from_kgid_munged(&init_user_ns, root_inode->i_gid));
-       seq_printf(m, ",pgrp=%d", sbi->oz_pgrp);
+       seq_printf(m, ",pgrp=%d", pid_vnr(sbi->oz_pgrp));
        seq_printf(m, ",timeout=%lu", sbi->exp_timeout/HZ);
        seq_printf(m, ",minproto=%d", sbi->min_proto);
        seq_printf(m, ",maxproto=%d", sbi->max_proto);
@@ -124,7 +127,8 @@ static const match_table_t tokens = {
 };
 
 static int parse_options(char *options, int *pipefd, kuid_t *uid, kgid_t *gid,
-               pid_t *pgrp, unsigned int *type, int *minproto, int *maxproto)
+                        int *pgrp, bool *pgrp_set, unsigned int *type,
+                        int *minproto, int *maxproto)
 {
        char *p;
        substring_t args[MAX_OPT_ARGS];
@@ -132,7 +136,6 @@ static int parse_options(char *options, int *pipefd, kuid_t *uid, kgid_t *gid,
 
        *uid = current_uid();
        *gid = current_gid();
-       *pgrp = task_pgrp_nr(current);
 
        *minproto = AUTOFS_MIN_PROTO_VERSION;
        *maxproto = AUTOFS_MAX_PROTO_VERSION;
@@ -171,6 +174,7 @@ static int parse_options(char *options, int *pipefd, kuid_t *uid, kgid_t *gid,
                        if (match_int(args, &option))
                                return 1;
                        *pgrp = option;
+                       *pgrp_set = true;
                        break;
                case Opt_minproto:
                        if (match_int(args, &option))
@@ -206,6 +210,8 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
        int pipefd;
        struct autofs_sb_info *sbi;
        struct autofs_info *ino;
+       int pgrp;
+       bool pgrp_set = false;
 
        sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
        if (!sbi)
@@ -218,7 +224,7 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
        sbi->pipe = NULL;
        sbi->catatonic = 1;
        sbi->exp_timeout = 0;
-       sbi->oz_pgrp = task_pgrp_nr(current);
+       sbi->oz_pgrp = NULL;
        sbi->sb = s;
        sbi->version = 0;
        sbi->sub_version = 0;
@@ -255,12 +261,23 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
 
        /* Can this call block? */
        if (parse_options(data, &pipefd, &root_inode->i_uid, &root_inode->i_gid,
-                               &sbi->oz_pgrp, &sbi->type, &sbi->min_proto,
-                               &sbi->max_proto)) {
+                         &pgrp, &pgrp_set, &sbi->type, &sbi->min_proto,
+                         &sbi->max_proto)) {
                printk("autofs: called with bogus options\n");
                goto fail_dput;
        }
 
+       if (pgrp_set) {
+               sbi->oz_pgrp = find_get_pid(pgrp);
+               if (!sbi->oz_pgrp) {
+                       pr_warn("autofs: could not find process group %d\n",
+                               pgrp);
+                       goto fail_dput;
+               }
+       } else {
+               sbi->oz_pgrp = get_task_pid(current, PIDTYPE_PGID);
+       }
+
        if (autofs_type_trigger(sbi->type))
                __managed_dentry_set_managed(root);
 
@@ -284,9 +301,9 @@ int autofs4_fill_super(struct super_block *s, void *data, int silent)
                sbi->version = sbi->max_proto;
        sbi->sub_version = AUTOFS_PROTO_SUBVERSION;
 
-       DPRINTK("pipe fd = %d, pgrp = %u", pipefd, sbi->oz_pgrp);
+       DPRINTK("pipe fd = %d, pgrp = %u", pipefd, pid_nr(sbi->oz_pgrp));
        pipe = fget(pipefd);
-       
+
        if (!pipe) {
                printk("autofs: could not open pipe file descriptor\n");
                goto fail_dput;
@@ -316,6 +333,7 @@ fail_dput:
 fail_ino:
        kfree(ino);
 fail_free:
+       put_pid(sbi->oz_pgrp);
        kfree(sbi);
        s->s_fs_info = NULL;
 fail_unlock: