[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Jeff Layton]

GFS2 and NFS have setlease routines that always just return -EINVAL.
Turn that into a generic routine that can live in fs/libfs.c.

Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: <linux-nfs@vger.kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: <cluster-devel@redhat.com>
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
---
 fs/gfs2/file.c     | 22 +---------------------
 fs/libfs.c         | 16 ++++++++++++++++
 fs/nfs/file.c      | 13 +------------
 fs/nfs/internal.h  |  1 -
 fs/nfs/nfs4file.c  |  2 +-
 include/linux/fs.h |  1 +
 6 files changed, 20 insertions(+), 35 deletions(-)

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 26b3f952e6b1..2c02478a86b0 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -913,26 +913,6 @@ out_uninit:
 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
 
 /**
- * gfs2_setlease - acquire/release a file lease
- * @file: the file pointer
- * @arg: lease type
- * @fl: file lock
- *
- * We don't currently have a way to enforce a lease across the whole
- * cluster; until we do, disable leases (by just returning -EINVAL),
- * unless the administrator has requested purely local locking.
- *
- * Locking: called under i_lock
- *
- * Returns: errno
- */
-
-static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl)
-{
-	return -EINVAL;
-}
-
-/**
  * gfs2_lock - acquire/release a posix lock on a file
  * @file: the file pointer
  * @cmd: either modify or retrieve lock state, possibly wait
@@ -1069,7 +1049,7 @@ const struct file_operations gfs2_file_fops = {
 	.flock		= gfs2_flock,
 	.splice_read	= generic_file_splice_read,
 	.splice_write	= iter_file_splice_write,
-	.setlease	= gfs2_setlease,
+	.setlease	= simple_nosetlease,
 	.fallocate	= gfs2_fallocate,
 };
 
diff --git a/fs/libfs.c b/fs/libfs.c
index 88e3e00e2eca..29012a303ef8 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -1075,3 +1075,19 @@ struct inode *alloc_anon_inode(struct super_block *s)
 	return inode;
 }
 EXPORT_SYMBOL(alloc_anon_inode);
+
+/**
+ * simple_nosetlease - generic helper for prohibiting leases
+ * @filp: file pointer
+ * @arg: type of lease to obtain
+ * @flp: new lease supplied for insertion
+ *
+ * Generic helper for filesystems that do not wish to allow leases to be set.
+ * All arguments are ignored and it just returns -EINVAL.
+ */
+int
+simple_nosetlease(struct file *filp, long arg, struct file_lock **flp)
+{
+	return -EINVAL;
+}
+EXPORT_SYMBOL(simple_nosetlease);
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 524dd80d1898..8c4048ecdad1 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -891,17 +891,6 @@ int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
 }
 EXPORT_SYMBOL_GPL(nfs_flock);
 
-/*
- * There is no protocol support for leases, so we have no way to implement
- * them correctly in the face of opens by other clients.
- */
-int nfs_setlease(struct file *file, long arg, struct file_lock **fl)
-{
-	dprintk("NFS: setlease(%pD2, arg=%ld)\n", file, arg);
-	return -EINVAL;
-}
-EXPORT_SYMBOL_GPL(nfs_setlease);
-
 const struct file_operations nfs_file_operations = {
 	.llseek		= nfs_file_llseek,
 	.read		= new_sync_read,
@@ -918,6 +907,6 @@ const struct file_operations nfs_file_operations = {
 	.splice_read	= nfs_file_splice_read,
 	.splice_write	= iter_file_splice_write,
 	.check_flags	= nfs_check_flags,
-	.setlease	= nfs_setlease,
+	.setlease	= simple_nosetlease,
 };
 EXPORT_SYMBOL_GPL(nfs_file_operations);
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index 9056622d2230..94d922ebb5ac 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -346,7 +346,6 @@ int nfs_file_release(struct inode *, struct file *);
 int nfs_lock(struct file *, int, struct file_lock *);
 int nfs_flock(struct file *, int, struct file_lock *);
 int nfs_check_flags(int);
-int nfs_setlease(struct file *, long, struct file_lock **);
 
 /* inode.c */
 extern struct workqueue_struct *nfsiod_workqueue;
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index a816f0627a6c..3e987ad9ae25 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -131,5 +131,5 @@ const struct file_operations nfs4_file_operations = {
 	.splice_read	= nfs_file_splice_read,
 	.splice_write	= iter_file_splice_write,
 	.check_flags	= nfs_check_flags,
-	.setlease	= nfs_setlease,
+	.setlease	= simple_nosetlease,
 };
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 458f733c96bd..435e3d9ec5cf 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2599,6 +2599,7 @@ extern int simple_write_end(struct file *file, struct address_space *mapping,
 			struct page *page, void *fsdata);
 extern int always_delete_dentry(const struct dentry *);
 extern struct inode *alloc_anon_inode(struct super_block *);
+extern int simple_nosetlease(struct file *, long, struct file_lock **);
 extern const struct dentry_operations simple_dentry_operations;
 
 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
-- 
1.9.3

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Trond Myklebust]

On Thu, Sep 4, 2014 at 8:38 AM, Jeff Layton <jlayton@primarydata.com> wrote:
> GFS2 and NFS have setlease routines that always just return -EINVAL.
> Turn that into a generic routine that can live in fs/libfs.c.
>
> Cc: Trond Myklebust <trond.myklebust@primarydata.com>
> Cc: <linux-nfs@vger.kernel.org>
> Cc: Steven Whitehouse <swhiteho@redhat.com>
> Cc: <cluster-devel@redhat.com>
> Signed-off-by: Jeff Layton <jlayton@primarydata.com>
> ---
>  fs/gfs2/file.c     | 22 +---------------------
>  fs/libfs.c         | 16 ++++++++++++++++
>  fs/nfs/file.c      | 13 +------------
>  fs/nfs/internal.h  |  1 -
>  fs/nfs/nfs4file.c  |  2 +-
>  include/linux/fs.h |  1 +
>  6 files changed, 20 insertions(+), 35 deletions(-)

Acked-by: Trond Myklebust <trond.myklebust@primarydata.com>

-- 
Trond Myklebust

Linux NFS client maintainer, PrimaryData

trond.myklebust at primarydata.com

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Jeff Layton]

On Thu, 4 Sep 2014 08:41:51 -0400
Trond Myklebust <trond.myklebust@primarydata.com> wrote:

> On Thu, Sep 4, 2014 at 8:38 AM, Jeff Layton <jlayton@primarydata.com> wrote:
> > GFS2 and NFS have setlease routines that always just return -EINVAL.
> > Turn that into a generic routine that can live in fs/libfs.c.
> >
> > Cc: Trond Myklebust <trond.myklebust@primarydata.com>
> > Cc: <linux-nfs@vger.kernel.org>
> > Cc: Steven Whitehouse <swhiteho@redhat.com>
> > Cc: <cluster-devel@redhat.com>
> > Signed-off-by: Jeff Layton <jlayton@primarydata.com>
> > ---
> >  fs/gfs2/file.c     | 22 +---------------------
> >  fs/libfs.c         | 16 ++++++++++++++++
> >  fs/nfs/file.c      | 13 +------------
> >  fs/nfs/internal.h  |  1 -
> >  fs/nfs/nfs4file.c  |  2 +-
> >  include/linux/fs.h |  1 +
> >  6 files changed, 20 insertions(+), 35 deletions(-)
> 
> Acked-by: Trond Myklebust <trond.myklebust@primarydata.com>
> 

Thanks. While spinning this up, I did have a momentary pause to wonder
if -ENOLCK would be a better return value here.

It would make it easier to distinguish this from from "oops, I passed
in bogus arguments". For now, I'll leave it as -EINVAL, but it's
something to consider...

-- 
Jeff Layton <jlayton@primarydata.com>

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Christoph Hellwig]

On Thu, Sep 04, 2014 at 08:38:27AM -0400, Jeff Layton wrote:
> GFS2 and NFS have setlease routines that always just return -EINVAL.
> Turn that into a generic routine that can live in fs/libfs.c.

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Trond Myklebust]

On Thu, Sep 4, 2014 at 8:49 AM, Jeff Layton <jeff.layton@primarydata.com> wrote:
> On Thu, 4 Sep 2014 08:41:51 -0400
> Trond Myklebust <trond.myklebust@primarydata.com> wrote:
>
>> On Thu, Sep 4, 2014 at 8:38 AM, Jeff Layton <jlayton@primarydata.com> wrote:
>> > GFS2 and NFS have setlease routines that always just return -EINVAL.
>> > Turn that into a generic routine that can live in fs/libfs.c.
>> >
>> > Cc: Trond Myklebust <trond.myklebust@primarydata.com>
>> > Cc: <linux-nfs@vger.kernel.org>
>> > Cc: Steven Whitehouse <swhiteho@redhat.com>
>> > Cc: <cluster-devel@redhat.com>
>> > Signed-off-by: Jeff Layton <jlayton@primarydata.com>
>> > ---
>> >  fs/gfs2/file.c     | 22 +---------------------
>> >  fs/libfs.c         | 16 ++++++++++++++++
>> >  fs/nfs/file.c      | 13 +------------
>> >  fs/nfs/internal.h  |  1 -
>> >  fs/nfs/nfs4file.c  |  2 +-
>> >  include/linux/fs.h |  1 +
>> >  6 files changed, 20 insertions(+), 35 deletions(-)
>>
>> Acked-by: Trond Myklebust <trond.myklebust@primarydata.com>
>>
>
> Thanks. While spinning this up, I did have a momentary pause to wonder
> if -ENOLCK would be a better return value here.
>
> It would make it easier to distinguish this from from "oops, I passed
> in bogus arguments". For now, I'll leave it as -EINVAL, but it's
> something to consider...
>

Actually, it looks as if when you compile with !CONFIG_FILE_LOCKING,
then fcntl_setlease() returns the value '0' (which would be
"success!"). The word "confusing" only begins to describe it all.

Cheers
  Trond

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Christoph Hellwig]

On Thu, Sep 04, 2014 at 02:25:35PM -0400, Trond Myklebust wrote:
> Actually, it looks as if when you compile with !CONFIG_FILE_LOCKING,
> then fcntl_setlease() returns the value '0' (which would be
> "success!"). The word "confusing" only begins to describe it all.

That's incorrect for sure, we should agree on a single sensible code
for:

 1) !CONFIG_FILE_LOCKING
 2) !lease_enable
 3) filesystem doesn't support leases.

[Cluster-devel] [PATCH v2 01/17] locks: consolidate "nolease" routines

[Jeff Layton]

On Thu, 4 Sep 2014 13:12:00 -0700
Christoph Hellwig <hch@infradead.org> wrote:

> On Thu, Sep 04, 2014 at 02:25:35PM -0400, Trond Myklebust wrote:
> > Actually, it looks as if when you compile with !CONFIG_FILE_LOCKING,
> > then fcntl_setlease() returns the value '0' (which would be
> > "success!"). The word "confusing" only begins to describe it all.
> 
> That's incorrect for sure, we should agree on a single sensible code
> for:
> 
>  1) !CONFIG_FILE_LOCKING
>  2) !lease_enable
>  3) filesystem doesn't support leases.
> 

Agreed. I think -ENOLCK is really better than -EINVAL.

I usually take -EINVAL to mean "you sent me something bogus". Whereas
-ENOLCK just says "locking doesn't work". -ENOLCK seems closer to the
situation in all 3 cases above.

That said, this is a user-visible change. The main userland consumer of
leases (AFAIK) is samba, so I'll take a peek at that code and run it by
them before merging anything.

-- 
Jeff Layton <jlayton@primarydata.com>