VFS: pass the flag setting by fcntl() to vfs
Several flags such as O_NONBLOCK are set using the fcntl() system call.
Currently the flags are validated in setfl() by the vfs using the
"check_flags" operation, but the eventual setting is not passed to the
vfs.
We have an application that uses the vfs/fuse and the flag O_NONBLOCK
is critical for the application.
This patch extends the "check_flags" function so that the flags are
passed to the vfs layer for setting in addition to validating. More
specifically:
o add additional parameters to "check_flags", one for the file pointer,
and one for setting. Also change the "flags" parameter from "int" to
"unsigned int" to make it consistent with "filp->f_flags".
o in setfl() pass the flags to vfs for setting once the flags are
set correctly in the kernel.
o use the "check_flags" operation in ioctl_fionbio() also for
ioctl(fd, FIONOIO, &on).
o make necessary adjustments to several files in fs/nfs (NFS is the
only module using "check_flags").
Signed-off-by: Enke Chen <[email protected]>
Version: 4.6.0_rc6_next_20160503
Documentation/filesystems/Locking | 2 +-
Documentation/filesystems/vfs.txt | 2 +-
fs/fcntl.c | 10 +++++++++-
fs/ioctl.c | 9 +++++++++
fs/nfs/dir.c | 2 +-
fs/nfs/file.c | 7 +++++--
fs/nfs/internal.h | 2 +-
fs/nfs/nfs4file.c | 2 +-
include/linux/fs.h | 2 +-
9 files changed, 29 insertions(+), 9 deletions(-)
diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index 75eea7c..c1cf807 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -452,7 +452,7 @@ prototypes:
loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long,
unsigned long, unsigned long, unsigned long);
- int (*check_flags)(int);
+ int (*check_flags)(unsigned int, struct file *, int);
int (*flock) (struct file *, int, struct file_lock *);
ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
size_t, unsigned int);
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index c61a223..193ee19 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -822,7 +822,7 @@ struct file_operations {
int (*lock) (struct file *, int, struct file_lock *);
ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
- int (*check_flags)(int);
+ int (*check_flags)(unsigned int, struct file *, int);
int (*flock) (struct file *, int, struct file_lock *);
ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
diff --git a/fs/fcntl.c b/fs/fcntl.c
index 350a2c8..5582dc8 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -32,6 +32,7 @@
static int setfl(int fd, struct file * filp, unsigned long arg)
{
struct inode * inode = file_inode(filp);
+ unsigned int flags;
int error = 0;
/*
@@ -59,7 +60,7 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
}
if (filp->f_op->check_flags)
- error = filp->f_op->check_flags(arg);
+ error = filp->f_op->check_flags(arg, filp, 0);
if (error)
return error;
@@ -75,8 +76,15 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
}
spin_lock(&filp->f_lock);
filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK);
+ flags = filp->f_flags;
spin_unlock(&filp->f_lock);
+ /*
+ * Pass the flags to VFS for setting.
+ */
+ if (filp->f_op->check_flags)
+ error = filp->f_op->check_flags(flags, filp, 1);
+
out:
return error;
}
diff --git a/fs/ioctl.c b/fs/ioctl.c
index 116a333..de06d93 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -496,6 +496,7 @@ static int file_ioctl(struct file *filp, unsigned int cmd,
static int ioctl_fionbio(struct file *filp, int __user *argp)
{
unsigned int flag;
+ unsigned int setfl;
int on, error;
error = get_user(on, argp);
@@ -512,7 +513,15 @@ static int ioctl_fionbio(struct file *filp, int __user *argp)
filp->f_flags |= flag;
else
filp->f_flags &= ~flag;
+ setfl = filp->f_flags;
spin_unlock(&filp->f_lock);
+
+ /*
+ * Do the same as in fcntl().
+ */
+ if (filp->f_op->check_flags)
+ error = filp->f_op->check_flags(setfl, filp, 1);
+
return error;
}
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index aaf7bd0..e16de49 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1495,7 +1495,7 @@ int nfs_atomic_open(struct inode *dir, struct dentry *dentry,
dfprintk(VFS, "NFS: atomic_open(%s/%lu), %pd\n",
dir->i_sb->s_id, dir->i_ino, dentry);
- err = nfs_check_flags(open_flags);
+ err = nfs_check_flags(open_flags, file, 0);
if (err)
return err;
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 717a8d6..d049929 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -48,8 +48,11 @@ static const struct vm_operations_struct nfs_file_vm_ops;
# define IS_SWAPFILE(inode) (0)
#endif
-int nfs_check_flags(int flags)
+int nfs_check_flags(unsigned int flags, struct file *filp, int setting)
{
+ if (setting)
+ return 0;
+
if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
return -EINVAL;
@@ -68,7 +71,7 @@ nfs_file_open(struct inode *inode, struct file *filp)
dprintk("NFS: open file(%pD2)\n", filp);
nfs_inc_stats(inode, NFSIOS_VFSOPEN);
- res = nfs_check_flags(filp->f_flags);
+ res = nfs_check_flags(filp->f_flags, filp, 0);
if (res)
return res;
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index f1d1d2c..0916d99 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -368,7 +368,7 @@ ssize_t nfs_file_write(struct kiocb *, struct iov_iter *);
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_check_flags(unsigned int, struct file *, int);
/* inode.c */
extern struct workqueue_struct *nfsiod_workqueue;
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index d039051..7ae0d15 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -44,7 +44,7 @@ nfs4_file_open(struct inode *inode, struct file *filp)
dprintk("NFS: open file(%pd2)\n", dentry);
- err = nfs_check_flags(openflags);
+ err = nfs_check_flags(openflags, filp, 0);
if (err)
return err;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index d61cd08..fdc9cc9 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1717,7 +1717,7 @@ struct file_operations {
int (*lock) (struct file *, int, struct file_lock *);
ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
- int (*check_flags)(int);
+ int (*check_flags)(unsigned int, struct file *, int);
int (*flock) (struct file *, int, struct file_lock *);
ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
File systems don't really have a business getting this call, and fuse
file systems even less so.
As explained in the email, currently the validation of the flags does reach
the vfs, but the setting does not. We have an application that uses the FUSE
to implement a user-space socket, and the flags are needed.
Best regards,
-- Enke
On 5/8/16 3:40 AM, Christoph Hellwig wrote:
> File systems don't really have a business getting this call, and fuse
> file systems even less so.
>
On Sun, May 08, 2016 at 05:50:32PM -0700, Enke Chen wrote:
> As explained in the email, currently the validation of the flags does reach
> the vfs, but the setting does not.
s/vfs/fs/, and yes, that's intentional.
> We have an application that uses the FUSE
> to implement a user-space socket, and the flags are needed.
And that's not a supported use case.
Hi, Christoph:
On 5/9/16 12:02 AM, Christoph Hellwig wrote:
> On Sun, May 08, 2016 at 05:50:32PM -0700, Enke Chen wrote:
>> As explained in the email, currently the validation of the flags does reach
>> the vfs, but the setting does not.
>
> s/vfs/fs/, and yes, that's intentional.
>
>> We have an application that uses the FUSE
>> to implement a user-space socket, and the flags are needed.
>
> And that's not a supported use case.
>
Understood, and that is the reason for the simple patch :-)
User-space networking stacks exist for several reasons, e.g., for migrating
from micro-kernel based systems, or for avoiding large changes to the networking
stack in the kernel.
What is the concern with the patch?
Regards,
-- Enke
Hi, Folks:
Could you accept this patch submitted on May 5, 2016?
Thanks. -- Enke
---
VFS: pass the flag setting by fcntl() to vfs
Several flags such as O_NONBLOCK are set using the fcntl() system call.
Currently the flags are validated in setfl() by the vfs using the
"check_flags" operation, but the eventual setting is not passed to the
vfs.
We have an application that uses the vfs/fuse and the flag O_NONBLOCK
is critical for the application.
This patch extends the "check_flags" function so that the flags are
passed to the vfs layer for setting in addition to validating. More
specifically:
o add additional parameters to "check_flags", one for the file pointer,
and one for setting. Also change the "flags" parameter from "int" to
"unsigned int" to make it consistent with "filp->f_flags".
o in setfl() pass the flags to vfs for setting once the flags are
set correctly in the kernel.
o use the "check_flags" operation in ioctl_fionbio() also for
ioctl(fd, FIONOIO, &on).
o make necessary adjustments to several files in fs/nfs (NFS is the
only module using "check_flags").
Signed-off-by: Enke Chen <[email protected]>
Version: 4.6.0_rc6_next_20160503
Documentation/filesystems/Locking | 2 +-
Documentation/filesystems/vfs.txt | 2 +-
fs/fcntl.c | 10 +++++++++-
fs/ioctl.c | 9 +++++++++
fs/nfs/dir.c | 2 +-
fs/nfs/file.c | 7 +++++--
fs/nfs/internal.h | 2 +-
fs/nfs/nfs4file.c | 2 +-
include/linux/fs.h | 2 +-
9 files changed, 29 insertions(+), 9 deletions(-)
diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking
index 75eea7c..c1cf807 100644
--- a/Documentation/filesystems/Locking
+++ b/Documentation/filesystems/Locking
@@ -452,7 +452,7 @@ prototypes:
loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long,
unsigned long, unsigned long, unsigned long);
- int (*check_flags)(int);
+ int (*check_flags)(unsigned int, struct file *, int);
int (*flock) (struct file *, int, struct file_lock *);
ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
size_t, unsigned int);
diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index c61a223..193ee19 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -822,7 +822,7 @@ struct file_operations {
int (*lock) (struct file *, int, struct file_lock *);
ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
- int (*check_flags)(int);
+ int (*check_flags)(unsigned int, struct file *, int);
int (*flock) (struct file *, int, struct file_lock *);
ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
diff --git a/fs/fcntl.c b/fs/fcntl.c
index 350a2c8..5582dc8 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -32,6 +32,7 @@
static int setfl(int fd, struct file * filp, unsigned long arg)
{
struct inode * inode = file_inode(filp);
+ unsigned int flags;
int error = 0;
/*
@@ -59,7 +60,7 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
}
if (filp->f_op->check_flags)
- error = filp->f_op->check_flags(arg);
+ error = filp->f_op->check_flags(arg, filp, 0);
if (error)
return error;
@@ -75,8 +76,15 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
}
spin_lock(&filp->f_lock);
filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK);
+ flags = filp->f_flags;
spin_unlock(&filp->f_lock);
+ /*
+ * Pass the flags to VFS for setting.
+ */
+ if (filp->f_op->check_flags)
+ error = filp->f_op->check_flags(flags, filp, 1);
+
out:
return error;
}
diff --git a/fs/ioctl.c b/fs/ioctl.c
index 116a333..de06d93 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -496,6 +496,7 @@ static int file_ioctl(struct file *filp, unsigned int cmd,
static int ioctl_fionbio(struct file *filp, int __user *argp)
{
unsigned int flag;
+ unsigned int setfl;
int on, error;
error = get_user(on, argp);
@@ -512,7 +513,15 @@ static int ioctl_fionbio(struct file *filp, int __user *argp)
filp->f_flags |= flag;
else
filp->f_flags &= ~flag;
+ setfl = filp->f_flags;
spin_unlock(&filp->f_lock);
+
+ /*
+ * Do the same as in fcntl().
+ */
+ if (filp->f_op->check_flags)
+ error = filp->f_op->check_flags(setfl, filp, 1);
+
return error;
}
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index aaf7bd0..e16de49 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1495,7 +1495,7 @@ int nfs_atomic_open(struct inode *dir, struct dentry *dentry,
dfprintk(VFS, "NFS: atomic_open(%s/%lu), %pd\n",
dir->i_sb->s_id, dir->i_ino, dentry);
- err = nfs_check_flags(open_flags);
+ err = nfs_check_flags(open_flags, file, 0);
if (err)
return err;
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 717a8d6..d049929 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -48,8 +48,11 @@ static const struct vm_operations_struct nfs_file_vm_ops;
# define IS_SWAPFILE(inode) (0)
#endif
-int nfs_check_flags(int flags)
+int nfs_check_flags(unsigned int flags, struct file *filp, int setting)
{
+ if (setting)
+ return 0;
+
if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
return -EINVAL;
@@ -68,7 +71,7 @@ nfs_file_open(struct inode *inode, struct file *filp)
dprintk("NFS: open file(%pD2)\n", filp);
nfs_inc_stats(inode, NFSIOS_VFSOPEN);
- res = nfs_check_flags(filp->f_flags);
+ res = nfs_check_flags(filp->f_flags, filp, 0);
if (res)
return res;
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index f1d1d2c..0916d99 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -368,7 +368,7 @@ ssize_t nfs_file_write(struct kiocb *, struct iov_iter *);
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_check_flags(unsigned int, struct file *, int);
/* inode.c */
extern struct workqueue_struct *nfsiod_workqueue;
diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c
index d039051..7ae0d15 100644
--- a/fs/nfs/nfs4file.c
+++ b/fs/nfs/nfs4file.c
@@ -44,7 +44,7 @@ nfs4_file_open(struct inode *inode, struct file *filp)
dprintk("NFS: open file(%pd2)\n", dentry);
- err = nfs_check_flags(openflags);
+ err = nfs_check_flags(openflags, filp, 0);
if (err)
return err;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index d61cd08..fdc9cc9 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1717,7 +1717,7 @@ struct file_operations {
int (*lock) (struct file *, int, struct file_lock *);
ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
- int (*check_flags)(int);
+ int (*check_flags)(unsigned int, struct file *, int);
int (*flock) (struct file *, int, struct file_lock *);
ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
On Wed, May 25, 2016 at 12:03:23PM -0700, Enke Chen wrote:
> Hi, Folks:
>
> Could you accept this patch submitted on May 5, 2016?
Not until you've addressed the objections in the original thread.
And no, "we have an application that wants it" does not qualify.
NAKed-by: Al Viro <[email protected]>
Hi, Al:
I did reply to the original thread:
-------------------
> And that's not a supported use case.
>
Understood, and that is the reason for the simple patch :-)
User-space networking stacks exist for several reasons, e.g., for migrating
from micro-kernel based systems, or for avoiding large changes to the networking
stack in the kernel.
------------------
Let me clarify a bit more:
FUSE, being a file system, should be able to support the same flag setting as
in the kernel. The patch facilitates that by passing the flags to the vfs.
Does this help?
Thanks. -- Enke
On 5/25/16 12:52 PM, Al Viro wrote:
> On Wed, May 25, 2016 at 12:03:23PM -0700, Enke Chen wrote:
>> Hi, Folks:
>>
>> Could you accept this patch submitted on May 5, 2016?
>
> Not until you've addressed the objections in the original thread.
> And no, "we have an application that wants it" does not qualify.
>
> NAKed-by: Al Viro <[email protected]>
>