Skip to contentSkip to conversation

Bind Mount Extensions 0.06

PATCH In review Herbert Poetzl 3 participants 6 messages Feb 22, 2005 – Feb 26, 2005

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 4/6 Bind Mount Extensions 0.06

;
; Bind Mount Extensions
;
; This part adds mnt_may_create() and mnt_may_unlink() checks ; and uses them in lookup_create(), sys_rmdir() and sys_unlink() ; it also adds a RDONLY check to generic_write_checks() and ; for pipe_writev()
;
; Copyright (C) 2003-2005 Herbert Pötzl <[email protected]> ;
; Changelog:
;
; 0.01 - broken out part from bme0.05
;
; this patch is free software; you can redistribute it and/or ; modify it under the terms of the GNU General Public License ; as published by the Free Software Foundation; either version 2 ; of the License, or (at your option) any later version. ;
; this patch is distributed in the hope that it will be useful, ; but WITHOUT ANY WARRANTY; without even the implied warranty of ; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the ; GNU General Public License for more details.
;

fs/namei.c +37 -7
diff -NurpP --minimal linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01/fs/namei.c linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01-co0.01/fs/namei.c
@@ -1168,6 +1168,24 @@ static inline int may_create(struct inod
 	return permission(dir,MAY_WRITE | MAY_EXEC, nd);
 }
 
+static inline int mnt_may_create(struct vfsmount *mnt, struct inode *dir, struct dentry *child) {
+       if (child->d_inode)
+               return -EEXIST;
+       if (IS_DEADDIR(dir))
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;
+       return 0;
+}
+
+static inline int mnt_may_unlink(struct vfsmount *mnt, struct inode *dir, struct dentry *child) {
+       if (!child->d_inode)
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;
+       return 0;
+}
+
 /* 
  * Special case: O_CREAT|O_EXCL implies O_NOFOLLOW for security
  * reasons.
@@ -1518,23 +1536,28 @@ do_link:
 struct dentry *lookup_create(struct nameidata *nd, int is_dir)
 {
 	struct dentry *dentry;
+	int error;
 
 	down(&nd->dentry->d_inode->i_sem);
-	dentry = ERR_PTR(-EEXIST);
+	error = -EEXIST;
 	if (nd->last_type != LAST_NORM)
-		goto fail;
+		goto out;
 	nd->flags &= ~LOOKUP_PARENT;
 	dentry = lookup_hash(&nd->last, nd->dentry);
 	if (IS_ERR(dentry))
+		goto ret;
+	error = mnt_may_create(nd->mnt, nd->dentry->d_inode, dentry);
+	if (error)
 		goto fail;
+	error = -ENOENT;
 	if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-		goto enoent;
+		goto fail;
+ret:
 	return dentry;
-enoent:
-	dput(dentry);
-	dentry = ERR_PTR(-ENOENT);
 fail:
-	return dentry;
+	dput(dentry);
+out:
+	return ERR_PTR(error);
 }
 
 int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
@@ -1763,7 +1786,11 @@ asmlinkage long sys_rmdir(const char __u
 	dentry = lookup_hash(&nd.last, nd.dentry);
 	error = PTR_ERR(dentry);
 	if (!IS_ERR(dentry)) {
+		error = mnt_may_unlink(nd.mnt, nd.dentry->d_inode, dentry);
+		if (error)
+			goto exit2;
 		error = vfs_rmdir(nd.dentry->d_inode, dentry);
+	exit2:
 		dput(dentry);
 	}
 	up(&nd.dentry->d_inode->i_sem);
@@ -1835,6 +1862,9 @@ asmlinkage long sys_unlink(const char __
 		/* Why not before? Because we want correct error value */
 		if (nd.last.name[nd.last.len])
 			goto slashes;
+		error = mnt_may_unlink(nd.mnt, nd.dentry->d_inode, dentry);
+		if (error)
+			goto exit2;
 		inode = dentry->d_inode;
 		if (inode)
 			atomic_inc(&inode->i_count);
fs/pipe.c +1 -1
diff -NurpP --minimal linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01/fs/pipe.c linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01-co0.01/fs/pipe.c
@@ -334,7 +334,7 @@ out:
 		wake_up_interruptible(PIPE_WAIT(*inode));
 		kill_fasync(PIPE_FASYNC_READERS(*inode), SIGIO, POLL_IN);
 	}
-	if (ret > 0)
+	if ((ret > 0) && !MNT_IS_RDONLY(filp->f_vfsmnt))
 		inode_update_time(inode, 1);	/* mtime and ctime */
 	return ret;
 }
mm/filemap.c +2 -0
diff -NurpP --minimal linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01/mm/filemap.c linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01-co0.01/mm/filemap.c
@@ -1810,6 +1810,8 @@ inline int generic_write_checks(struct f
         }
 
 	if (!isblk) {
+		if (MNT_IS_RDONLY(file->f_vfsmnt))
+			return -EROFS;
 		/* FIXME: this is for backwards compatibility with 2.4 */
 		if (file->f_flags & O_APPEND)
                         *pos = i_size_read(inode);
#1 lore raw

ty den 22.02.2005 Klokka 13:12 (+0100) skreiv Herbert Poetzl:

Quoted 14 lines

diff -NurpP --minimal linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01/fs/namei.c linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01-co0.01/fs/namei.c

fs/namei.c +7 -0
@@ -1168,6 +1168,24 @@ static inline int may_create(struct inod
 	return permission(dir,MAY_WRITE | MAY_EXEC, nd);
 }
 
+static inline int mnt_may_create(struct vfsmount *mnt, struct inode *dir, struct dentry *child) {
+       if (child->d_inode)
+               return -EEXIST;
+       if (IS_DEADDIR(dir))
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;

fs/namei.c static inline int may_create(struct inod · line 1187

Quoted 10 lines
fs/namei.c +10 -0
+       return 0;
+}
+
+static inline int mnt_may_unlink(struct vfsmount *mnt, struct inode *dir, struct dentry *child) {
+       if (!child->d_inode)
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;
+       return 0;
+}

Most of these checks are redundant, since they are already being done elsewhere in the code.
child->d_inode is, for instance checked in may_delete() and in may_create.
IS_DEADDIR is also checked in may_create.

Quoted 27 lines
fs/namei.c +9 -3
 /* 
  * Special case: O_CREAT|O_EXCL implies O_NOFOLLOW for security
  * reasons.
@@ -1518,23 +1536,28 @@ do_link:
 struct dentry *lookup_create(struct nameidata *nd, int is_dir)
 {
 	struct dentry *dentry;
+	int error;
 
 	down(&nd->dentry->d_inode->i_sem);
-	dentry = ERR_PTR(-EEXIST);
+	error = -EEXIST;
 	if (nd->last_type != LAST_NORM)
-		goto fail;
+		goto out;
 	nd->flags &= ~LOOKUP_PARENT;
 	dentry = lookup_hash(&nd->last, nd->dentry);
 	if (IS_ERR(dentry))
+		goto ret;
+	error = mnt_may_create(nd->mnt, nd->dentry->d_inode, dentry);
+	if (error)
 		goto fail;
+	error = -ENOENT;
 	if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-		goto enoent;
+		goto fail;
+ret:

fs/namei.c do_link: · line 1561

Quoted 10 lines
fs/namei.c +3 -4
 	return dentry;
-enoent:
-	dput(dentry);
-	dentry = ERR_PTR(-ENOENT);
 fail:
-	return dentry;
+	dput(dentry);
+out:
+	return ERR_PTR(error);
 }

What is the value of adding "error"? The current code is more efficient, and just as readable.

Cheers,
Trond

Signature · 1 line
Trond Myklebust <[email protected]>
#2 lore raw
Quoted 23 lines · Trond Myklebust

ty den 22.02.2005 Klokka 13:12 (+0100) skreiv Herbert Poetzl:

Quoted 21 lines

diff -NurpP --minimal linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01/fs/namei.c linux-2.6.11-rc4-bme0.06-bm0.01-at0.01-cc0.01-co0.01/fs/namei.c

fs/namei.c +14 -0
@@ -1168,6 +1168,24 @@ static inline int may_create(struct inod
 	return permission(dir,MAY_WRITE | MAY_EXEC, nd);
 }
 
+static inline int mnt_may_create(struct vfsmount *mnt, struct inode *dir, struct dentry *child) {
+       if (child->d_inode)
+               return -EEXIST;
+       if (IS_DEADDIR(dir))
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;
+       return 0;
+}
+
+static inline int mnt_may_unlink(struct vfsmount *mnt, struct inode *dir, struct dentry *child) {
+       if (!child->d_inode)
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
Quoted 8 lines

fs/namei.c static inline int may_create(struct inod · line 1179

Quoted 3 lines
fs/namei.c +3 -0
+               return -EROFS;
+       return 0;
+}

Most of these checks are redundant, since they are already being done elsewhere in the code.
child->d_inode is, for instance checked in may_delete() and in may_create.

well, but mnt_may_create() for example is called from lookup_create() which in turn is (for example) called from sys_mknod() which doesn't call may_delete() or may_create() ... did I miss something?

Quoted 1 line

IS_DEADDIR is also checked in may_create.

ditto ...

Quoted 32 lines
Quoted 32 lines
fs/namei.c +9 -6
 /* 
  * Special case: O_CREAT|O_EXCL implies O_NOFOLLOW for security
  * reasons.
@@ -1518,23 +1536,28 @@ do_link:
 struct dentry *lookup_create(struct nameidata *nd, int is_dir)
 {
 	struct dentry *dentry;
+	int error;
 
 	down(&nd->dentry->d_inode->i_sem);
-	dentry = ERR_PTR(-EEXIST);
+	error = -EEXIST;
 	if (nd->last_type != LAST_NORM)
-		goto fail;
+		goto out;
 	nd->flags &= ~LOOKUP_PARENT;
 	dentry = lookup_hash(&nd->last, nd->dentry);
 	if (IS_ERR(dentry))
+		goto ret;
+	error = mnt_may_create(nd->mnt, nd->dentry->d_inode, dentry);
+	if (error)
 		goto fail;
+	error = -ENOENT;
 	if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-		goto enoent;
+		goto fail;
+ret:
 	return dentry;
-enoent:
-	dput(dentry);
-	dentry = ERR_PTR(-ENOENT);
 fail:
Quoted 8 lines

fs/namei.c do_link: · line 1561

Quoted 5 lines
fs/namei.c +3 -1
-	return dentry;
+	dput(dentry);
+out:
+	return ERR_PTR(error);
 }

What is the value of adding "error"?
The current code is more efficient, and just as readable.

hmm, a compile test resulted in almost identical code, but maybe I did miss some aspect of the efficiency ...

anyway, thanks for the feedback,
Herbert

Quoted 5 lines

Cheers,
Trond

Signature · 1 line
Trond Myklebust <[email protected]>
#3 lore raw
In reply to Trond Myklebust
Quoted 33 lines · Trond Myklebust

ty den 22.02.2005 Klokka 13:12 (+0100) skreiv Herbert Poetzl:

Quoted 31 lines

* Special case: O_CREAT|O_EXCL implies O_NOFOLLOW for security
* reasons.

fs/namei.c +9 -6
@@ -1518,23 +1536,28 @@ do_link:
 struct dentry *lookup_create(struct nameidata *nd, int is_dir)
 {
 	struct dentry *dentry;
+	int error;
 
 	down(&nd->dentry->d_inode->i_sem);
-	dentry = ERR_PTR(-EEXIST);
+	error = -EEXIST;
 	if (nd->last_type != LAST_NORM)
-		goto fail;
+		goto out;
 	nd->flags &= ~LOOKUP_PARENT;
 	dentry = lookup_hash(&nd->last, nd->dentry);
 	if (IS_ERR(dentry))
+		goto ret;
+	error = mnt_may_create(nd->mnt, nd->dentry->d_inode, dentry);
+	if (error)
 		goto fail;
+	error = -ENOENT;
 	if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-		goto enoent;
+		goto fail;
+ret:
 	return dentry;
-enoent:
-	dput(dentry);
-	dentry = ERR_PTR(-ENOENT);
 fail:
Quoted 8 lines

fs/namei.c do_link: · line 1561

Quoted 5 lines
fs/namei.c +3 -1
-	return dentry;
+	dput(dentry);
+out:
+	return ERR_PTR(error);
 }

What is the value of adding "error"? The current code is more efficient, and just as readable.

okay, had a deeper look at this and now I remember why I added 'error' in the first place (quite some time ago ;)

basically we need to check for RO in lookup_create() now to give the same (error) results than a 'normal' ro mounted filesystem, it is required to do the dentry lookup and check the dentry->d_inode to return EEXIST before returning EROFS ...

something like this would be the result:

        if (dentry->d_inode) {
                dput(dentry);
                dentry = ERR_PTR(-EEXIST);
                goto fail;
        }
        if (MNT_IS_RDONLY(nd->mnt)) {
                dput(dentry);
                dentry = ERR_PTR(-EROFS);
                goto fail;
        }

I'm currently looking into moving that check upwards so that it will happen _after_ the EEXIST one ...

best,
Herbert

Quoted 5 lines

Cheers,
Trond

Signature · 1 line
Trond Myklebust <[email protected]>
#4 lore raw
In reply to Herbert Poetzl

Hi,

fs/namei.c static inline int may_create(struct inod · line 1179

Quoted 7 lines · Herbert Poetzl

+static inline int mnt_may_unlink(struct vfsmount *mnt, struct inode *dir,
struct dentry *child) { + if (!child->d_inode)

fs/namei.c +5 -0
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;
+       return 0;
+}

The argument "dir" is not used. Please remove it and fix the callers.

Regards

Ingo Oeser

#5 lore raw
Quoted 12 lines · Ingo Oeser

Hi,

fs/namei.c static inline int may_create(struct inod · line 1179

Quoted 7 lines · Herbert Poetzl

+static inline int mnt_may_unlink(struct vfsmount *mnt, struct inode *dir,
struct dentry *child) { + if (!child->d_inode)

fs/namei.c +5 -0
+               return -ENOENT;
+       if (MNT_IS_RDONLY(mnt))
+               return -EROFS;
+       return 0;
+}

The argument "dir" is not used. Please remove it and fix the callers.

hmm, yes, just saw that the mnt_may*() functions are not according to the 'current' coding style for that section/file, so I'll rework that anyway ...

thanks a lot for the input,
Herbert

Quoted 9 lines

Regards

Ingo Oeser

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/