vfs: cleanup of permission()
Subsystems (15)
Hi Andrew! Christoph! Al!
after thinking some time about the oracle words (sent in reply to previous BME submissions) we (Sam and I) came to the conclusion that it would be a good idea to remove the nameidata introduced in September 2003 from the inode permission() checks, so that vfs_permission() can take care of them ...
this is in two parts, the first one does the removal and the second one fixes up nfs and fuse by passing the relevant nd_flags via the mask
Note: this is just a suggestion, so please let
us know what you think
RFC 1/2 vfs: remove nameidata from *_permission()
remove struct nameidata from *_permission() and
related security functions.
---
fs/cifs/cifsfs.c +1 -1
@@ -220,7 +220,7 @@ cifs_statfs(struct super_block *sb, strulonger available? */ } -static int cifs_permission (struct inode * inode,int mask,struct nameidata *nd) +static int cifs_permission (struct inode * inode,int mask) {struct cifs_sb_info *cifs_sb;
fs/coda/dir.c +1 -1
@@ -151,7 +151,7 @@ exit: } -int coda_permission (struct inode *inode,int mask,struct nameidata *nd) +int coda_permission (struct inode *inode,int mask) {int error =0 ;
fs/coda/pioctl.c +2 -4
@@ -24,8 +24,7 @@#include <linux/coda_psdev.h> /* pioctl ops */ -static int coda_ioctl_permission (struct inode *inode,int mask, -struct nameidata *nd); +static int coda_ioctl_permission (struct inode *inode,int mask);static int coda_pioctl (struct inode * inode,struct file * filp,unsigned int cmd,unsigned long user_data);
@@ -42,8 +41,7 @@ struct file_operations coda_ioctl_operat };/* the coda pioctl inode ops */ -static int coda_ioctl_permission (struct inode *inode,int mask, -struct nameidata *nd) +static int coda_ioctl_permission (struct inode *inode,int mask) {return 0 ; }
fs/ext2/acl.c +1 -1
@@ -294,7 +294,7 @@ ext2_check_acl(struct inode *inode, int }int -ext2_permission (struct inode *inode,int mask,struct nameidata *nd) +ext2_permission (struct inode *inode,int mask) {return generic_permission (inode, mask, ext2_check_acl); }
fs/ext3/acl.c +1 -1
@@ -299,7 +299,7 @@ ext3_check_acl(struct inode *inode, int }int -ext3_permission (struct inode *inode,int mask,struct nameidata *nd) +ext3_permission (struct inode *inode,int mask) {return generic_permission (inode, mask, ext3_check_acl); }
fs/fuse/dir.c +1 -1
@@ -702,7 +702,7 @@ static int fuse_access(struct inode *ino* access request is sent. Execute permission is still checked * locally based on file mode. */ -static int fuse_permission (struct inode *inode,int mask,struct nameidata *nd) +static int fuse_permission (struct inode *inode,int mask) {struct fuse_conn *fc =get_fuse_conn (inode);
fs/hfs/inode.c +1 -2
@@ -519,8 +519,7 @@ void hfs_clear_inode(struct inode *inode } } -static int hfs_permission (struct inode *inode,int mask, -struct nameidata *nd) +static int hfs_permission (struct inode *inode,int mask) {if (S_ISREG (inode->i_mode) && mask &MAY_EXEC )return 0 ;
fs/hfsplus/inode.c +1 -1
@@ -237,7 +237,7 @@ static void hfsplus_set_perms(struct ino perms->dev =cpu_to_be32 (HFSPLUS_I (inode).dev); } -static int hfsplus_permission (struct inode *inode,int mask,struct nameidata *nd) +static int hfsplus_permission (struct inode *inode,int mask) {/* MAY_EXEC is also used for lookup, if no x bit is set allow lookup, * open_exec has the same test, so it's still not executable, if a x bit
fs/hostfs/hostfs_kern.c +1 -1
@@ -796,7 +796,7 @@ int hostfs_rename(struct inode *from_inoreturn (err); } -int hostfs_permission (struct inode *ino,int desired,struct nameidata *nd) +int hostfs_permission (struct inode *ino,int desired) {char *name;int r =0 , w =0 , x =0 , err;
fs/hpfs/namei.c +1 -1
@@ -415,7 +415,7 @@ again:d_drop (dentry);spin_lock (&dentry->d_lock);if (atomic_read (&dentry->d_count) >1 || -permission (inode,MAY_WRITE ,NULL ) || +permission (inode,MAY_WRITE ) || !S_ISREG (inode->i_mode) ||get_write_access (inode)) {spin_unlock (&dentry->d_lock);
fs/jfs/acl.c +1 -1
@@ -140,7 +140,7 @@ static int jfs_check_acl(struct inode *ireturn -EAGAIN ; } -int jfs_permission (struct inode *inode,int mask,struct nameidata *nd) +int jfs_permission (struct inode *inode,int mask) {return generic_permission (inode, mask, jfs_check_acl); }
fs/jfs/jfs_acl.h +1 -1
@@ -20,7 +20,7 @@#ifdef CONFIG_JFS_POSIX_ACL -int jfs_permission (struct inode *,int ,struct nameidata *); +int jfs_permission (struct inode *,int );int jfs_init_acl (tid_t ,struct inode *,struct inode *);int jfs_setattr (struct dentry *,struct iattr *);
fs/namei.c +14 -10
@@ -225,7 +225,7 @@ int generic_permission(struct inode *inoreturn -EACCES ; } -int permission (struct inode *inode,int mask,struct nameidata *nd) +int permission (struct inode *inode,int mask) {int retval, submask;
@@ -250,13 +250,13 @@ int permission(struct inode *inode, int/* Ordinary permission routines do not understand MAY_APPEND. */ submask = mask & ~MAY_APPEND ;if (inode->i_op && inode->i_op->permission) - retval = inode->i_op->permission (inode, submask, nd); + retval = inode->i_op->permission (inode, submask);else retval =generic_permission (inode, submask,NULL );if (retval)return retval; -return security_inode_permission (inode, mask, nd); +return security_inode_permission (inode, mask); }/**
@@ -271,7 +271,8 @@ int permission(struct inode *inode, int*/ int vfs_permission (struct nameidata *nd,int mask) { -return permission (nd->dentry->d_inode, mask, nd); +/* do nd based stuff here */ +return permission (nd->dentry->d_inode, mask); }/**
@@ -288,7 +289,8 @@ int vfs_permission(struct nameidata *nd,*/ int file_permission (struct file *file,int mask) { -return permission (file->f_dentry->d_inode, mask,NULL ); +/* do file based stuff here */ +return permission (file->f_dentry->d_inode, mask); }/*
@@ -425,7 +427,7 @@ static int exec_permission_lite(struct ireturn -EACCES ; ok: -return security_inode_permission (inode,MAY_EXEC , nd); +return security_inode_permission (inode,MAY_EXEC ); }/*
@@ -1219,7 +1221,8 @@ static struct dentry * __lookup_hash(strint err; inode = base->d_inode; - err =permission (inode,MAY_EXEC , nd); +/* should become vfs_permission() */ + err =permission (inode,MAY_EXEC ); dentry =ERR_PTR (err);if (err) goto out;
@@ -1354,7 +1357,7 @@ static int may_delete(struct inode *dir,BUG_ON (victim->d_parent->d_inode != dir); - error =permission (dir,MAY_WRITE |MAY_EXEC ,NULL ); + error =permission (dir,MAY_WRITE |MAY_EXEC );if (error)return error;if (IS_APPEND (dir))
@@ -1391,7 +1394,8 @@ static inline int may_create(struct inodreturn -EEXIST ;if (IS_DEADDIR (dir))return -ENOENT ; -return permission (dir,MAY_WRITE |MAY_EXEC , nd); +/* should become vfs_permission() */ +return permission (dir,MAY_WRITE |MAY_EXEC ); }/*
@@ -2313,7 +2317,7 @@ static int vfs_rename_dir(struct inode ** we'll need to flip '..'. */ if (new_dir != old_dir) { - error =permission (old_dentry->d_inode,MAY_WRITE ,NULL ); + error =permission (old_dentry->d_inode,MAY_WRITE );if (error)return error; }
fs/nfs/dir.c +1 -1
@@ -1635,7 +1635,7 @@ out:return -EACCES ; } -int nfs_permission (struct inode *inode,int mask,struct nameidata *nd) +int nfs_permission (struct inode *inode,int mask) {struct rpc_cred *cred;int res =0 ;
fs/nfsd/nfsfh.c +1 -1
@@ -56,7 +56,7 @@ static int nfsd_acceptable(void *expv, s/* make sure parents give x permission to user */ int err; parent =dget_parent (tdentry); - err =permission (parent->d_inode,MAY_EXEC ,NULL ); + err =permission (parent->d_inode,MAY_EXEC );if (err <0 ) {dput (parent);break ;
fs/nfsd/vfs.c +2 -2
@@ -1817,12 +1817,12 @@ nfsd_permission(struct svc_export *exp, inode->i_uid == current->fsuid)return 0 ; - err =permission (inode, acc & (MAY_READ |MAY_WRITE |MAY_EXEC ),NULL ); + err =permission (inode, acc & (MAY_READ |MAY_WRITE |MAY_EXEC ));/* Allow read access to binaries even when mode 111 */ if (err == -EACCES &&S_ISREG (inode->i_mode) && acc == (MAY_READ |MAY_OWNER_OVERRIDE )) - err =permission (inode,MAY_EXEC ,NULL ); + err =permission (inode,MAY_EXEC );return err?nfserrno (err) :0 ; }
fs/proc/base.c +2 -2
@@ -579,14 +579,14 @@ static int proc_check_root(struct inodereturn proc_check_chroot (root, vfsmnt); } -static int proc_permission (struct inode *inode,int mask,struct nameidata *nd) +static int proc_permission (struct inode *inode,int mask) {if (generic_permission (inode, mask,NULL ) !=0 )return -EACCES ;return proc_check_root (inode); } -static int proc_task_permission (struct inode *inode,int mask,struct nameidata *nd) +static int proc_task_permission (struct inode *inode,int mask) {struct dentry *root;struct vfsmount *vfsmnt;
fs/reiserfs/xattr.c +1 -1
@@ -1343,7 +1343,7 @@ static int reiserfs_check_acl(struct inoreturn error; } -int reiserfs_permission (struct inode *inode,int mask,struct nameidata *nd) +int reiserfs_permission (struct inode *inode,int mask) {/* * We don't do permission checks on the internal objects.
fs/smbfs/file.c +1 -1
@@ -387,7 +387,7 @@ smb_file_release(struct inode *inode, st* privileges, so we need our own check for this. */ static int -smb_file_permission (struct inode *inode,int mask,struct nameidata *nd) +smb_file_permission (struct inode *inode,int mask) {int mode = inode->i_mode;int error =0 ;
fs/xattr.c +1 -1
@@ -58,7 +58,7 @@ xattr_permission(struct inode *inode, coreturn -EPERM ; } -return permission (inode, mask,NULL ); +return permission (inode, mask); }int
fs/xfs/linux-2.6/xfs_iops.c +1 -2
@@ -614,8 +614,7 @@ linvfs_put_link( STATIC intlinvfs_permission (struct inode *inode, -int mode, -struct nameidata *nd) +int mode) {vnode_t *vp =LINVFS_GET_VP (inode);int error;
include/linux/fs.h +3 -2
@@ -1040,7 +1040,7 @@ struct inode_operations {void * (*follow_link) (struct dentry *,struct nameidata *);void (*put_link) (struct dentry *,struct nameidata *,void *);void (*truncate) (struct inode *); -int (*permission) (struct inode *,int ,struct nameidata *); +int (*permission) (struct inode *,int );int (*setattr) (struct dentry *,struct iattr *);int (*getattr) (struct vfsmount *mnt,struct dentry *,struct kstat *);int (*setxattr) (struct dentry *,const char *,const void *,size_t ,int );
@@ -1465,7 +1465,8 @@ extern int do_remount_sb(struct super_bl void *data, int force);extern sector_t bmap (struct inode *,sector_t );extern int notify_change (struct dentry *,struct iattr *); -extern int permission (struct inode *,int ,struct nameidata *); +extern int permission (struct inode *,int ); +// extern int vfs_permission(struct nameidata *, int); extern int generic_permission (struct inode *,int ,int (*check_acl)(struct inode *,int ));
include/linux/nfs_fs.h +1 -1
@@ -296,7 +296,7 @@ extern struct inode *nfs_fhget(struct suextern int nfs_refresh_inode (struct inode *,struct nfs_fattr *);extern int nfs_post_op_update_inode (struct inode *inode,struct nfs_fattr *fattr);extern int nfs_getattr (struct vfsmount *,struct dentry *,struct kstat *); -extern int nfs_permission (struct inode *,int ,struct nameidata *); +extern int nfs_permission (struct inode *,int );extern int nfs_access_get_cached (struct inode *,struct rpc_cred *,struct nfs_access_entry *);extern void nfs_access_add_cache (struct inode *,struct nfs_access_entry *);extern int nfs_open (struct inode *,struct file *);
include/linux/security.h +4 -7
@@ -346,7 +346,6 @@ struct swap_info_struct;* called when the actual read/write operations are performed. * @inode contains the inode structure to check. * @mask contains the permission mask. -* @nd contains the nameidata (may be NULL). * Return 0 if permission is granted. * @inode_setattr: * Check permission before setting file attributes. Note that the kernel
@@ -1157,7 +1156,7 @@ struct security_operations { struct inode *new_dir, struct dentry *new_dentry);int (*inode_readlink) (struct dentry *dentry);int (*inode_follow_link) (struct dentry *dentry,struct nameidata *nd); -int (*inode_permission) (struct inode *inode,int mask,struct nameidata *nd); +int (*inode_permission) (struct inode *inode,int mask);int (*inode_setattr) (struct dentry *dentry,struct iattr *attr);int (*inode_getattr) (struct vfsmount *mnt,struct dentry *dentry);void (*inode_delete) (struct inode *inode);
@@ -1606,12 +1605,11 @@ static inline int security_inode_follow_return security_ops->inode_follow_link (dentry, nd); } -static inline int security_inode_permission (struct inode *inode,int mask, -struct nameidata *nd) +static inline int security_inode_permission (struct inode *inode,int mask) {if (unlikely (IS_PRIVATE (inode)))return 0 ; -return security_ops->inode_permission (inode, mask, nd); +return security_ops->inode_permission (inode, mask); } static inline int security_inode_setattr (struct dentry *dentry,
@@ -2270,8 +2268,7 @@ static inline int security_inode_follow_return 0 ; } -static inline int security_inode_permission (struct inode *inode,int mask, -struct nameidata *nd) +static inline int security_inode_permission (struct inode *inode,int mask) {return 0 ; }
ipc/mqueue.c +1 -1
@@ -639,7 +639,7 @@ static int oflag2acc[O_ACCMODE] = { MAY_return ERR_PTR (-EINVAL ); } -if (permission (dentry->d_inode, oflag2acc[oflag &O_ACCMODE ],NULL )) { +if (permission (dentry->d_inode, oflag2acc[oflag &O_ACCMODE ])) {dput (dentry);mntput (mqueue_mnt);return ERR_PTR (-EACCES );
security/dummy.c +1 -1
@@ -324,7 +324,7 @@ static int dummy_inode_follow_link (strureturn 0 ; } -static int dummy_inode_permission (struct inode *inode,int mask,struct nameidata *nd) +static int dummy_inode_permission (struct inode *inode,int mask) {return 0 ; }
security/seclvl.c +1 -1
@@ -422,7 +422,7 @@ static void seclvl_bd_release(struct ino* seclvl 1, we only deny writes to *mounted* block devices. */ static int -seclvl_inode_permission (struct inode *inode,int mask,struct nameidata *nd) +seclvl_inode_permission (struct inode *inode,int mask) {if (current->pid !=1 &&S_ISBLK (inode->i_mode) && (mask &MAY_WRITE )) {switch (seclvl) {
security/selinux/hooks.c +2 -3
@@ -2052,12 +2052,11 @@ static int selinux_inode_follow_link(strreturn dentry_has_perm (current,NULL , dentry,FILE__READ ); } -static int selinux_inode_permission (struct inode *inode,int mask, -struct nameidata *nd) +static int selinux_inode_permission (struct inode *inode,int mask) {int rc; - rc = secondary_ops->inode_permission (inode, mask, nd); + rc = secondary_ops->inode_permission (inode, mask);if (rc)return rc;
RFC 2/2 vfs: fixup nfs and fuse by passing nd_flags via mask
In reply to Herbert Poetzlfixup nfs and fuse nameidata related checks by
passing the relevant flags via the mask.
---
fs/fuse/dir.c +2 -1
@@ -731,8 +731,9 @@ static int fuse_permission(struct inodeif ((mask &MAY_EXEC ) && !S_ISDIR (mode) && !(mode &S_IXUGO ))return -EACCES ; -if (nd && (nd->flags &LOOKUP_ACCESS )) +if (mask &LOOKUP_ACCESS )return fuse_access (inode, mask); +return 0 ; } }
fs/namei.c +4 -1
@@ -271,8 +271,11 @@ int permission(struct inode *inode, int*/ int vfs_permission (struct nameidata *nd,int mask) { +int nd_flags = nd->flags & + (LOOKUP_ACCESS |LOOKUP_CREATE |LOOKUP_OPEN ); +/* do nd based stuff here */ -return permission (nd->dentry->d_inode, mask); +return permission (nd->dentry->d_inode, mask | nd_flags); }/**
fs/nfs/dir.c +2 -3
@@ -1643,7 +1643,7 @@ int nfs_permission(struct inode *inode,if (mask ==0 ) goto out;/* Is this sys_access() ? */ -if (nd !=NULL && (nd->flags &LOOKUP_ACCESS )) +if (mask &LOOKUP_ACCESS ) goto force_lookup;switch (inode->i_mode &S_IFMT ) {
@@ -1652,8 +1652,7 @@ int nfs_permission(struct inode *inode,case S_IFREG :/* NFSv4 has atomic_open... */ if (nfs_server_capable (inode,NFS_CAP_ATOMIC_OPEN ) - && nd !=NULL - && (nd->flags &LOOKUP_OPEN )) + && (mask &LOOKUP_OPEN )) goto out;break ;case S_IFDIR :
Quoted 9 lines · Herbert Poetzl
Hi Andrew! Christoph! Al!
after thinking some time about the oracle words (sent in reply to previous BME submissions) we (Sam and I) came to the conclusion that it would be a good idea to remove the nameidata introduced in September 2003 from the inode
permission()checks, so thatvfs_permission()can take care of them ...
Why? There may be perfectly legitimate reasons for the filesystem to request information about the path. I can think of server failover situations in NFSv4 where the client may need to look up the filehandle for the file on the new server before it can service the ACCESS call.
Quoted 6 lines
this is in two parts, the first one does the removal and the second one fixes up nfs and fuse by passing the relevant
nd_flagsvia the maskNote: this is just a suggestion, so please let
us know what you think
Firstly, the fact that the lookup intent flags happen not to collide with MAY_* is a complete fluke, not a design. The numerical values of either set of flags could change tomorrow for all you know.
Secondly, an intent is _not_ a permissions mask by any stretch of the imagination.
IOW: at the very least make that intent flag a separate parameter.
Cheers,
Trond
Quoted 10 lines · Herbert Poetzl
Hi Andrew! Christoph! Al!
after thinking some time about the oracle words (sent in reply to previous BME submissions) we (Sam and I) came to the conclusion that it would be a good idea to remove the nameidata introduced in September 2003 from the inode
permission()checks, so thatvfs_permission()can take care of them ...
Could you please provide a link to that 'previous BME submissions'? Thanks.
Also, since you are modifying LSM interfaces, why not discuss it on the LSM mailing list?
And finally, please don't remove nameidata. Modules out there depend on it and we at Sophos are about to release a new product which needs it as well. The plan was to announce the whole thing parallel with the release, but after spotting your post I was prompted to react ahead of the schedule. However, I am very busy at the moment so the actual announcment with full details will have to wait for a week or two.
Quoted 3 lines
And finally, please don't remove nameidata. Modules out there depend on it
are those modules about to merged into the kernel? The current intent infrastructure isn't fulfilling what it should do well, and from what I've seen on the discussions it sounds that the best way forward is to undo the current implementation and then roll out one which caters to the needs of the existing users better.
As external module, you have little say so far simply because your usage isn't visible. I'd urge you to quickly submit your code so that the things you need from this are better visible to the people who are thinking and working on the redesign.
Quoted 2 lines
and we at Sophos are about to release a new product which needs it as well.
I assume we're talking about an open source product, or at least kernel component, here?
Arjan van de Ven <[email protected]> wrote on 01/03/2006 12:37:51:
Quoted 1 line
Quoted 1 line
And finally, please don't remove nameidata. Modules out there depend
on it
Quoted 2 lines
are those modules about to merged into the kernel? The current intent
See third paragraph of my reply.
Quoted 4 lines
infrastructure isn't fulfilling what it should do well, and from what I've seen on the discussions it sounds that the best way forward is to undo the current implementation and then roll out one which caters to the needs of the existing users better.
That I don't know so I can't comment at the moment. I haven't seen anything on linux-security-modules recently?
Quoted 4 lines
As external module, you have little say so far simply because your usage isn't visible. I'd urge you to quickly submit your code so that the things you need from this are better visible to the people who are thinking and working on the redesign.
I know all that, but it is a complicated matter to discuss. That's why I was planning to make a comprehensive announcement which would discuss most of the hot topics. Ideally yes, I would like to merge, but it won't happen now. The first thing I would like to do is establish common ground with other security vendors so that we could approach the problem together. Personaly, I am not sure whether insisting that everything should be a part of kernel is a right thing to do even though I think I understand all the up- and down-sides of both policies.
Having said all this above, I am afraid that there will be no other choice but to start working on the announcement asap. :)
Quoted 5 lines
Quoted 2 lines
and we at Sophos are about to release a new product which needs it as well.
I assume we're talking about an open source product, or at least kernel component, here?
Of course. Kernel component might be known to some as Talpa and it is released under GPL.
Quoted 7 lines · [email protected] wrote:
Quoted 4 lines
As external module, you have little say so far simply because your usage isn't visible. I'd urge you to quickly submit your code so that the things you need from this are better visible to the people who are thinking and working on the redesign.
I know all that, but it is a complicated matter to discuss. That's why I was planning to make a comprehensive announcement which would discuss most of the hot topics. Ideally yes, I would like to merge, but it won't happen now. The first thing I would like to do is establish common ground with other security vendors so that we could approach the problem together. Personaly, I am not sure whether insisting that everything should be a part of kernel is a right thing to do even though I think I understand all the up- and down-sides of both policies.
Having said all this above, I am afraid that there will be no other choice but to start working on the announcement asap. :)
Perhaps you can put up the work in progress on a git repository somewhere?
Sam.
Quoted 11 lines · On Wed, Mar 01, 2006 at 12:28:25PM +0000, [email protected] wrote:
Quoted 10 lines · Herbert Poetzl
Hi Andrew! Christoph! Al!
after thinking some time about the oracle words (sent in reply to previous BME submissions) we (Sam and I) came to the conclusion that it would be a good idea to remove the nameidata introduced in September 2003 from the inode
permission()checks, so thatvfs_permission()can take care of them ...
Quoted 1 line
Could you please provide a link to that 'previous BME submissions'?
here you go: http://lkml.org/lkml/2006/1/21/19
Quoted 1 line
Thanks.
you're welcome!
Quoted 2 lines
Also, since you are modifying LSM interfaces, why not discuss it on the LSM mailing list?
no problem with that, will cc the lsm folks next time (feel free to bounce the messages)
for now, here is a link to this thread:
http://lkml.org/lkml/2006/2/28/4
Quoted 6 lines
And finally, please don't remove nameidata. Modules out there depend on it and we at Sophos are about to release a new product which needs it as well. The plan was to announce the whole thing parallel with the release, but after spotting your post I was prompted to react ahead of the schedule. However, I am very busy at the moment so the actual announcment with full details will have to wait for a week or two.
thing is, permission() does inode based checks and the nameidata is not even provided in most cases, so you cannot rely on that information anyway
it would probably be better to have some kind of vfs_permission, which uses dentry/vfsmnt for decisions on the vfs layer, this would also allow to cover most of the cases where nameidata is not available (for example the filep based stuff)
best,
Herbert
Quoted 5 lines
-
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/
Quoted 3 lines · [email protected] wrote:
Also, since you are modifying LSM interfaces, why not discuss it on the LSM mailing list?
And finally, please don't remove nameidata. Modules out there depend on it and we at Sophos are about to release a new product which needs it as well. The plan was to announce the whole thing parallel with the release, but after spotting your post I was prompted to react ahead of the schedule. However, I am very busy at the moment so the actual announcment with full details will have to wait for a week or two.
You are treating the per-FS security hooks as if they were VFS security hooks. This was an easy mistake to make, as the appearance of a (struct nameidata*) sure makes it look like a VFS call.
However, most functions in the kernel don't pass anything in that nameidata slot. Some (eg, syscalls that work on open FDs) can't, either. So the fact that it does not guarantee VFS context information in all situations means permission() is not a VFS function.
ie, we don't disagree with what you're trying to do, but if you want path information then you should be working at the VFS layer, not the FS layer.
Perhaps you could first come up with a patch to the LSM base that adds VFS hooks rather than FS hooks and make your new system use those hooks?
I think that it might be more obvious where such hooks should go,
after applying this patch.
What we're aiming for, on the permission() front, is that all system calls, ioctls, etc, call either vfs_permission() or file_permission(). Only those two functions should end up calling permission() directly.
Sam.
Quoted 11 lines · Trond Myklebust
Quoted 9 lines · Herbert Poetzl
Hi Andrew! Christoph! Al!
after thinking some time about the oracle words (sent in reply to previous BME submissions) we (Sam and I) came to the conclusion that it would be a good idea to remove the nameidata introduced in September 2003 from the inode
permission()checks, so thatvfs_permission()can take care of them ...
Quoted 5 lines
Why? There may be perfectly legitimate reasons for the filesystem to request information about the path. I can think of server failover situations in NFSv4 where the client may need to look up the filehandle for the file on the new server before it can service the ACCESS call.
the second part is actually a hack to help nfs and fuse to get the 'required' information until there is a proper interface (at the vfs not inode level) to pass relevant information (probably dentry/vfsmount/flags)
Quoted 7 lines
Quoted 6 lines
this is in two parts, the first one does the removal and the second one fixes up nfs and fuse by passing the relevant
nd_flagsvia the maskNote: this is just a suggestion, so please let
us know what you think
Quoted 6 lines
Firstly, the fact that the lookup intent flags happen not to collide with MAY_* is a complete fluke, not a design. The numerical values of either set of flags could change tomorrow for all you know.
Secondly, an intent is _not_ a permissions mask by any stretch of the imagination.
see above
Quoted 1 line
IOW: at the very least make that intent flag a separate parameter.
IMHO it would be good to remove them completely form the current permission() checks.
best,
Herbert
Quoted 8 lines
Cheers,
Trond-
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/
Quoted 4 lines · Herbert Poetzl
the second part is actually a hack to help nfs and fuse to get the 'required' information until there is a proper interface (at the vfs not inode level) to pass relevant information (probably dentry/vfsmount/flags)
The nameidata _IS_ the vfs structure for storing path context information. You seem to be suggesting we need yet another one. Why?
Quoted 14 lines
Quoted 7 lines
Quoted 6 lines
this is in two parts, the first one does the
removal and the second one fixes up nfs and fuse
by passing the relevantnd_flagsvia the maskNote: this is just a suggestion, so please let
us know what you thinkQuoted 6 lines
Firstly, the fact that the lookup intent flags happen not to collide with MAY_* is a complete fluke, not a design. The numerical values of either set of flags could change tomorrow for all you know.
Secondly, an intent is _not_ a permissions mask by any stretch of the imagination.
Quoted 6 lines
see above
Quoted 1 line
IOW: at the very least make that intent flag a separate parameter.
IMHO it would be good to remove them completely form the current
permission()checks.
Vetoed!
Redundant RPC calls have performance costs to the client, the server and the network. That intent information is there in order to allow the filesystem to figure out whether or not it needs to do the permissions check, or if that check is already being done by other operations.
Removing the intents are therefore not an option.
Cheers,
Trond
Quoted 6 lines · Trond Myklebust
Quoted 4 lines
the second part is actually a hack to help nfs and fuse to get the 'required' information until there is a proper interface (at the vfs not inode level) to pass relevant information (probably dentry/vfsmount/flags)
The nameidata _IS_ the vfs structure for storing path context information. You seem to be suggesting we need yet another one. Why?
Because you can't make a nameidata without a lookup, and file based operations don't do a lookup. However you still have the vfsmnt and inode hanging off the file struct.
Either that or we make a dummy nameidata structure for this situation, possibly a filehandle relative lookup as used by openat() et al.
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
Secondly, an intent is _not_ a permissions mask by any stretch of the imagination.
see above
Quoted 1 line
IOW: at the very least make that intent flag a separate parameter.
Quoted 8 lines
Quoted 2 lines
IMHO it would be good to remove them completely form the current
permission()checks.Vetoed!
Redundant RPC calls have performance costs to the client, the server and the network. That intent information is there in order to allow the filesystem to figure out whether or not it needs to do the permissions check, or if that check is already being done by other operations. Removing the intents are therefore not an option.
OK, so we either make it an extra parameter or 'properly' stack them into a single word. Do you have any preferences either way there?
Sam.
Quoted 11 lines · Sam Vilain
Quoted 6 lines · Trond Myklebust
Quoted 4 lines
the second part is actually a hack to help nfs and fuse to get the 'required' information until there is a proper interface (at the vfs not inode level) to pass relevant information (probably dentry/vfsmount/flags)
The nameidata _IS_ the vfs structure for storing path context information. You seem to be suggesting we need yet another one. Why?
Because you can't make a nameidata without a lookup, and file based operations don't do a lookup. However you still have the vfsmnt and inode hanging off the file struct.
Which is fine by NFS, at least. We don't care about performing permissions checks in those cases anyway, so a NULL nameidata is OK. The only cases where we currently need to take action in ->permission() are
1) path walks (checking MAY_EXEC on the directory)
2) open() (on NFSv2/v3 only)
3) access()
In the future we can/should probably eliminate (2) by making an atomic_open() for NFSv2/v3, so that we can correctly set the credential that was used for the open() permissions check in the struct file.
Quoted 16 lines
Either that or we make a dummy nameidata structure for this situation, possibly a filehandle relative lookup as used by
openat()et al.Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
Secondly, an intent is _not_ a permissions mask by any stretch of the imagination.
see above
Quoted 1 line
IOW: at the very least make that intent flag a separate parameter.
Quoted 8 lines
Quoted 2 lines
IMHO it would be good to remove them completely form the current
permission()checks.Vetoed!
Redundant RPC calls have performance costs to the client, the server and the network. That intent information is there in order to allow the filesystem to figure out whether or not it needs to do the permissions check, or if that check is already being done by other operations. Removing the intents are therefore not an option.
Quoted 2 lines
OK, so we either make it an extra parameter or 'properly' stack them into a single word. Do you have any preferences either way there?
An extra parameter should suffice for (1) and (2).
Cheers,
Trond
Quoted 7 lines · Trond Myklebust
Quoted 5 lines
after thinking some time about the oracle words
(sent in reply to previous BME submissions) we (Sam and I) came to the conclusion that it would be a good idea to remove the nameidata introduced in September 2003 from the inodepermission()
checks, so thatvfs_permission()can take care
of them ...Why? There may be perfectly legitimate reasons for the filesystem to request information about the path.
Part of what we're trying to achieve is clearly seperating VFS from FS operations. A lot of the problems with supporting the per-vfsmnt options stemmed from this confusion.
In fact this was largely developed from your feedback on the prior thread¹ :) we audited the uses, and you were right - some places really couldn't set that field properly. After some playing and discussion, we came to guess that this kind of cleanup was a good way to satisfy your concerns as well as Christoph's².
¹ http://xrl.us/j9et
² http://xrl.us/j9eo
The only place that the nameidata structure is currently used is in those two places, and given that it *does* appear to represent FS vs VFS brain damage to some extent, it's quite important we get this right before umpteen security modules spring up using this bad hook ... the API is broken (doesn't cover (struct file*)-based operations, only lookup (struct nameidata*), and parameter is optional).
Quoted 3 lines
I can think of server failover
situations in NFSv4 where the client may need to look up the filehandle for the file on the new server before it can service the ACCESS call.
The current code doesn't seem to use that right now, at least not via permission(). Either way, the information should be available - it just needs to hook in somewhere else.
Quoted 5 lines
Firstly, the fact that the lookup intent flags happen not to collide with MAY_* is a complete fluke, not a design. The numerical values of either set of flags could change tomorrow for all you know. Secondly, an intent is _not_ a permissions mask by any stretch of the imagination.
Yeah, I had this in there at one point as a safety check;
#if (_LOOKUP_MASK ^ _ACCESS_MASK) != (_LOOKUP_MASK | _ACCESS_MASK) #error splode
#endif
You are right, it should either be converted to two parameters or the combination made more formal.
Quoted 1 line
IOW: at the very least make that intent flag a separate parameter.
Yes, or rename it mask_and_intent perhaps if they were to be combined in the same word (which makes sense, with a grand total of 7 bits being used, and gcc not being in a position to spot that and optimise for us).
Quoted 3 lines
Cheers,
Trond
Thanks for your feedback!
Sam.