diff mbox series

[v5,10/10] io_uring: add support for IORING_OP_MKNODAT

Message ID 20210603051836.2614535-11-dkadashev@gmail.com (mailing list archive)
State New
Headers show
Series io_uring: add mkdir, [sym]linkat and mknodat support | expand

Commit Message

Dmitry Kadashev June 3, 2021, 5:18 a.m. UTC
IORING_OP_MKNODAT behaves like mknodat(2) and takes the same flags and
arguments.

Suggested-by: Christian Brauner <christian.brauner@ubuntu.com>
Link: https://lore.kernel.org/io-uring/20210514145259.wtl4xcsp52woi6ab@wittgenstein/
Signed-off-by: Dmitry Kadashev <dkadashev@gmail.com>
---
 fs/internal.h                 |  2 ++
 fs/io_uring.c                 | 56 +++++++++++++++++++++++++++++++++++
 fs/namei.c                    |  2 +-
 include/uapi/linux/io_uring.h |  2 ++
 4 files changed, 61 insertions(+), 1 deletion(-)

Comments

Pavel Begunkov June 22, 2021, 11:52 a.m. UTC | #1
On 6/3/21 6:18 AM, Dmitry Kadashev wrote:
> IORING_OP_MKNODAT behaves like mknodat(2) and takes the same flags and
> arguments.
> 
> Suggested-by: Christian Brauner <christian.brauner@ubuntu.com>
> Link: https://lore.kernel.org/io-uring/20210514145259.wtl4xcsp52woi6ab@wittgenstein/
> Signed-off-by: Dmitry Kadashev <dkadashev@gmail.com>
> ---
>  fs/internal.h                 |  2 ++
>  fs/io_uring.c                 | 56 +++++++++++++++++++++++++++++++++++
>  fs/namei.c                    |  2 +-
>  include/uapi/linux/io_uring.h |  2 ++
>  4 files changed, 61 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/internal.h b/fs/internal.h
> index 15a7d210cc67..c6fb9974006f 100644

[...]

>  static bool io_disarm_next(struct io_kiocb *req);
> @@ -3687,6 +3697,44 @@ static int io_linkat(struct io_kiocb *req, int issue_flags)
>  	io_req_complete(req, ret);
>  	return 0;
>  }
> +static int io_mknodat_prep(struct io_kiocb *req,
> +			    const struct io_uring_sqe *sqe)
> +{
> +	struct io_mknod *mkn = &req->mknod;
> +	const char __user *fname;
> +
> +	if (unlikely(req->flags & REQ_F_FIXED_FILE))
> +		return -EBADF;

IOPOLL won't support it, and the check is missing.
Probably for other opcodes as well.

if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
	return -EINVAL;

> +
> +	mkn->dfd = READ_ONCE(sqe->fd);
> +	mkn->mode = READ_ONCE(sqe->len);
> +	fname = u64_to_user_ptr(READ_ONCE(sqe->addr));
> +	mkn->dev = READ_ONCE(sqe->mknod_dev);
> +
> +	mkn->filename = getname(fname);
> +	if (IS_ERR(mkn->filename))
> +		return PTR_ERR(mkn->filename);
> +
> +	req->flags |= REQ_F_NEED_CLEANUP;
> +	return 0;
> +}
Dmitry Kadashev June 23, 2021, 6:26 a.m. UTC | #2
On Tue, Jun 22, 2021 at 6:52 PM Pavel Begunkov <asml.silence@gmail.com> wrote:
>
> On 6/3/21 6:18 AM, Dmitry Kadashev wrote:
> > IORING_OP_MKNODAT behaves like mknodat(2) and takes the same flags and
> > arguments.
> >
> > Suggested-by: Christian Brauner <christian.brauner@ubuntu.com>
> > Link: https://lore.kernel.org/io-uring/20210514145259.wtl4xcsp52woi6ab@wittgenstein/
> > Signed-off-by: Dmitry Kadashev <dkadashev@gmail.com>
> > ---
> >  fs/internal.h                 |  2 ++
> >  fs/io_uring.c                 | 56 +++++++++++++++++++++++++++++++++++
> >  fs/namei.c                    |  2 +-
> >  include/uapi/linux/io_uring.h |  2 ++
> >  4 files changed, 61 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/internal.h b/fs/internal.h
> > index 15a7d210cc67..c6fb9974006f 100644
>
> [...]
>
> >  static bool io_disarm_next(struct io_kiocb *req);
> > @@ -3687,6 +3697,44 @@ static int io_linkat(struct io_kiocb *req, int issue_flags)
> >       io_req_complete(req, ret);
> >       return 0;
> >  }
> > +static int io_mknodat_prep(struct io_kiocb *req,
> > +                         const struct io_uring_sqe *sqe)
> > +{
> > +     struct io_mknod *mkn = &req->mknod;
> > +     const char __user *fname;
> > +
> > +     if (unlikely(req->flags & REQ_F_FIXED_FILE))
> > +             return -EBADF;
>
> IOPOLL won't support it, and the check is missing.
> Probably for other opcodes as well.
>
> if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
>         return -EINVAL;

This change is based on some other similar opcodes (unlinkat, renameat) that
were added a while ago. Those lack the check as well. I guess I'll just prepare
a patch that adds the checks to all of them. Thanks, Pavel.

Jens, separately it's my understanding that you do not want the MKNODAT opcode
at all, should I drop this from the subsequent series?
Pavel Begunkov June 23, 2021, 11:58 a.m. UTC | #3
On 6/23/21 7:26 AM, Dmitry Kadashev wrote:
> On Tue, Jun 22, 2021 at 6:52 PM Pavel Begunkov <asml.silence@gmail.com> wrote:
>>
>> On 6/3/21 6:18 AM, Dmitry Kadashev wrote:
>>> IORING_OP_MKNODAT behaves like mknodat(2) and takes the same flags and
>>> arguments.
>>>
>>> Suggested-by: Christian Brauner <christian.brauner@ubuntu.com>
>>> Link: https://lore.kernel.org/io-uring/20210514145259.wtl4xcsp52woi6ab@wittgenstein/
>>> Signed-off-by: Dmitry Kadashev <dkadashev@gmail.com>
>>> ---
>>>  fs/internal.h                 |  2 ++
>>>  fs/io_uring.c                 | 56 +++++++++++++++++++++++++++++++++++
>>>  fs/namei.c                    |  2 +-
>>>  include/uapi/linux/io_uring.h |  2 ++
>>>  4 files changed, 61 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/fs/internal.h b/fs/internal.h
>>> index 15a7d210cc67..c6fb9974006f 100644
>>
>> [...]
>>
>>>  static bool io_disarm_next(struct io_kiocb *req);
>>> @@ -3687,6 +3697,44 @@ static int io_linkat(struct io_kiocb *req, int issue_flags)
>>>       io_req_complete(req, ret);
>>>       return 0;
>>>  }
>>> +static int io_mknodat_prep(struct io_kiocb *req,
>>> +                         const struct io_uring_sqe *sqe)
>>> +{
>>> +     struct io_mknod *mkn = &req->mknod;
>>> +     const char __user *fname;
>>> +
>>> +     if (unlikely(req->flags & REQ_F_FIXED_FILE))
>>> +             return -EBADF;
>>
>> IOPOLL won't support it, and the check is missing.
>> Probably for other opcodes as well.
>>
>> if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
>>         return -EINVAL;
> 
> This change is based on some other similar opcodes (unlinkat, renameat) that
> were added a while ago. Those lack the check as well. I guess I'll just prepare
> a patch that adds the checks to all of them. Thanks, Pavel.
> 
> Jens, separately it's my understanding that you do not want the MKNODAT opcode
> at all, should I drop this from the subsequent series?

I guess that comment was to my note that the patch with it
doesn't compile, but as it should be already fixed up...
Jens Axboe June 24, 2021, 2:36 a.m. UTC | #4
On 6/23/21 12:26 AM, Dmitry Kadashev wrote:
> On Tue, Jun 22, 2021 at 6:52 PM Pavel Begunkov <asml.silence@gmail.com> wrote:
>>
>> On 6/3/21 6:18 AM, Dmitry Kadashev wrote:
>>> IORING_OP_MKNODAT behaves like mknodat(2) and takes the same flags and
>>> arguments.
>>>
>>> Suggested-by: Christian Brauner <christian.brauner@ubuntu.com>
>>> Link: https://lore.kernel.org/io-uring/20210514145259.wtl4xcsp52woi6ab@wittgenstein/
>>> Signed-off-by: Dmitry Kadashev <dkadashev@gmail.com>
>>> ---
>>>  fs/internal.h                 |  2 ++
>>>  fs/io_uring.c                 | 56 +++++++++++++++++++++++++++++++++++
>>>  fs/namei.c                    |  2 +-
>>>  include/uapi/linux/io_uring.h |  2 ++
>>>  4 files changed, 61 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/fs/internal.h b/fs/internal.h
>>> index 15a7d210cc67..c6fb9974006f 100644
>>
>> [...]
>>
>>>  static bool io_disarm_next(struct io_kiocb *req);
>>> @@ -3687,6 +3697,44 @@ static int io_linkat(struct io_kiocb *req, int issue_flags)
>>>       io_req_complete(req, ret);
>>>       return 0;
>>>  }
>>> +static int io_mknodat_prep(struct io_kiocb *req,
>>> +                         const struct io_uring_sqe *sqe)
>>> +{
>>> +     struct io_mknod *mkn = &req->mknod;
>>> +     const char __user *fname;
>>> +
>>> +     if (unlikely(req->flags & REQ_F_FIXED_FILE))
>>> +             return -EBADF;
>>
>> IOPOLL won't support it, and the check is missing.
>> Probably for other opcodes as well.
>>
>> if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
>>         return -EINVAL;
> 
> This change is based on some other similar opcodes (unlinkat, renameat) that
> were added a while ago. Those lack the check as well. I guess I'll just prepare
> a patch that adds the checks to all of them. Thanks, Pavel.
> 
> Jens, separately it's my understanding that you do not want the MKNODAT opcode
> at all, should I drop this from the subsequent series?

Right, just drop that one for now. Would be great if you could resend
the series with the suggested fixes folded in. Might as well just do
a clean sweep.
diff mbox series

Patch

diff --git a/fs/internal.h b/fs/internal.h
index 15a7d210cc67..c6fb9974006f 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -81,6 +81,8 @@  int do_mkdirat(int dfd, struct filename *name, umode_t mode);
 int do_symlinkat(struct filename *from, int newdfd, struct filename *to);
 int do_linkat(int olddfd, struct filename *old, int newdfd,
 			struct filename *new, int flags);
+int do_mknodat(int dfd, struct filename *name, umode_t mode,
+		unsigned int dev);
 
 /*
  * namespace.c
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 31e1aa7dd90b..475632374af8 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -688,6 +688,14 @@  struct io_hardlink {
 	int				flags;
 };
 
+struct io_mknod {
+	struct file			*file;
+	int				dfd;
+	umode_t				mode;
+	struct filename		*filename;
+	unsigned int		dev;
+};
+
 struct io_completion {
 	struct file			*file;
 	struct list_head		list;
@@ -835,6 +843,7 @@  struct io_kiocb {
 		struct io_mkdir		mkdir;
 		struct io_symlink	symlink;
 		struct io_hardlink	hardlink;
+		struct io_mknod		mknod;
 		/* use only after cleaning per-op data, see io_clean_op() */
 		struct io_completion	compl;
 	};
@@ -1050,6 +1059,7 @@  static const struct io_op_def io_op_defs[] = {
 	[IORING_OP_MKDIRAT] = {},
 	[IORING_OP_SYMLINKAT] = {},
 	[IORING_OP_LINKAT] = {},
+	[IORING_OP_MKNODAT] = {},
 };
 
 static bool io_disarm_next(struct io_kiocb *req);
@@ -3687,6 +3697,44 @@  static int io_linkat(struct io_kiocb *req, int issue_flags)
 	io_req_complete(req, ret);
 	return 0;
 }
+static int io_mknodat_prep(struct io_kiocb *req,
+			    const struct io_uring_sqe *sqe)
+{
+	struct io_mknod *mkn = &req->mknod;
+	const char __user *fname;
+
+	if (unlikely(req->flags & REQ_F_FIXED_FILE))
+		return -EBADF;
+
+	mkn->dfd = READ_ONCE(sqe->fd);
+	mkn->mode = READ_ONCE(sqe->len);
+	fname = u64_to_user_ptr(READ_ONCE(sqe->addr));
+	mkn->dev = READ_ONCE(sqe->mknod_dev);
+
+	mkn->filename = getname(fname);
+	if (IS_ERR(mkn->filename))
+		return PTR_ERR(mkn->filename);
+
+	req->flags |= REQ_F_NEED_CLEANUP;
+	return 0;
+}
+
+static int io_mknodat(struct io_kiocb *req, int issue_flags)
+{
+	struct io_mknod *mkn = &req->mknod;
+	int ret;
+
+	if (issue_flags & IO_URING_F_NONBLOCK)
+		return -EAGAIN;
+
+	ret = do_mknodat(mkn->dfd, mkn->filename, mkn->mode, mkn->dev);
+
+	req->flags &= ~REQ_F_NEED_CLEANUP;
+	if (ret < 0)
+		req_set_fail(req);
+	io_req_complete(req, ret);
+	return 0;
+}
 
 static int io_shutdown_prep(struct io_kiocb *req,
 			    const struct io_uring_sqe *sqe)
@@ -6100,6 +6148,8 @@  static int io_req_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		return io_symlinkat_prep(req, sqe);
 	case IORING_OP_LINKAT:
 		return io_linkat_prep(req, sqe);
+	case IORING_OP_MKNODAT:
+		return io_mknodat_prep(req, sqe);
 	}
 
 	printk_once(KERN_WARNING "io_uring: unhandled opcode %d\n",
@@ -6252,6 +6302,9 @@  static void io_clean_op(struct io_kiocb *req)
 			putname(req->hardlink.oldpath);
 			putname(req->hardlink.newpath);
 			break;
+		case IORING_OP_MKNODAT:
+			putname(req->mknod.filename);
+			break;
 		}
 		req->flags &= ~REQ_F_NEED_CLEANUP;
 	}
@@ -6387,6 +6440,9 @@  static int io_issue_sqe(struct io_kiocb *req, unsigned int issue_flags)
 	case IORING_OP_LINKAT:
 		ret = io_linkat(req, issue_flags);
 		break;
+	case IORING_OP_MKNODAT:
+		ret = io_mknodat(req, issue_flags);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
diff --git a/fs/namei.c b/fs/namei.c
index b85e457c43b7..a4b1848b5dd1 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -3743,7 +3743,7 @@  static int may_mknod(umode_t mode)
 	}
 }
 
-static int do_mknodat(int dfd, struct filename *name, umode_t mode,
+int do_mknodat(int dfd, struct filename *name, umode_t mode,
 		unsigned int dev)
 {
 	struct user_namespace *mnt_userns;
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 510e64a0a9c3..824b37f53a28 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -45,6 +45,7 @@  struct io_uring_sqe {
 		__u32		rename_flags;
 		__u32		unlink_flags;
 		__u32		hardlink_flags;
+		__u32		mknod_dev;
 	};
 	__u64	user_data;	/* data to be passed back at completion time */
 	union {
@@ -141,6 +142,7 @@  enum {
 	IORING_OP_MKDIRAT,
 	IORING_OP_SYMLINKAT,
 	IORING_OP_LINKAT,
+	IORING_OP_MKNODAT,
 
 	/* this goes last, obviously */
 	IORING_OP_LAST,