linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Amir Goldstein <amir73il@gmail.com>
To: Miklos Szeredi <miklos@szeredi.hu>
Cc: Daniel Rosenberg <drosen@google.com>,
	Paul Lawrence <paullawrence@google.com>,
	Alessio Balsini <balsini@android.com>,
	fuse-devel@lists.sourceforge.net, linux-fsdevel@vger.kernel.org
Subject: [PATCH v13 03/10] fuse: Passthrough initialization and release
Date: Fri, 19 May 2023 15:56:58 +0300	[thread overview]
Message-ID: <20230519125705.598234-4-amir73il@gmail.com> (raw)
In-Reply-To: <20230519125705.598234-1-amir73il@gmail.com>

Implement the FUSE passthrough ioctl that associates a backing
(passthrough) file with the fuse_file.

The file descriptor passed to the ioctl by the FUSE daemon is used to
access the relative file pointer, that will be copied to the fuse_file
data structure to consolidate the link between the FUSE and the backing
file.

To enable the passthrough mode, FUSE daemon calls the
FUSE_DEV_IOC_PASSTHROUGH_OPEN ioctl and, if the call succeeds, receives
back an identifier that will be used at open/create response time in the
fuse_open_out field to associate the FUSE file to the backing file.

The value returned by the ioctl to user space can be:
  > 0: success, the value can be used as part of an open/create reply.
 <= 0: an error occurred.
 == 0: represents an error to preserve backward compatibility - the
fuse_open_out field that is used to pass the backing_id back to the
kernel uses the same bits that were previously as struct padding, and is
commonly zero-initialized (e.g., in the libfuse implementation).

Removing 0 from the correct values fixes the ambiguity between the case
in which 0 corresponds to a real backing_id, a missing implementation
of FUSE passthrough or a request for a normal FUSE file, simplifying the
user space implementation.

For the passthrough mode to be successfully activated, the backing file
filesystem must implement both read_iter and write_iter file operations.
This extra check avoids special pseudo files to be used for passhrough.
Passthrough comes with another limitation: no further filesystem stacking
is allowed for those FUSE filesystems using passthrough.

The FUSE daemon must call FUSE_DEV_IOC_PASSTHROUGH_CLOSE ioctl to drop
the reference on the backing file.  This can be done immediattely after
responding to open or at a later time.

In any case, the backing file will be kept open by the FUSE driver until
the last fuse_file that was setup to passthrough to that backing file is
closed AND the FUSE_DEV_IOC_PASSTHROUGH_CLOSE ioctl was called.

Signed-off-by: Alessio Balsini <balsini@android.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
---
 fs/fuse/fuse_i.h      |  1 +
 fs/fuse/inode.c       |  1 +
 fs/fuse/passthrough.c | 89 +++++++++++++++++++++++++++++++++++++++++--
 3 files changed, 87 insertions(+), 4 deletions(-)

diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h
index f52604534ff6..9ad1cc37a5c4 100644
--- a/fs/fuse/fuse_i.h
+++ b/fs/fuse/fuse_i.h
@@ -181,6 +181,7 @@ struct fuse_passthrough {
 
 	/** refcount */
 	refcount_t count;
+	struct rcu_head rcu;
 };
 
 /** FUSE specific file data */
diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index 271586fac008..4200fb13883e 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -1301,6 +1301,7 @@ static int fuse_passthrough_id_free(int id, void *p, void *data)
 
 	WARN_ON_ONCE(refcount_read(&passthrough->count) != 1);
 	fuse_passthrough_free(passthrough);
+
 	return 0;
 }
 
diff --git a/fs/fuse/passthrough.c b/fs/fuse/passthrough.c
index fc723e004de9..8d090ae252f2 100644
--- a/fs/fuse/passthrough.c
+++ b/fs/fuse/passthrough.c
@@ -3,6 +3,7 @@
 #include "fuse_i.h"
 
 #include <linux/file.h>
+#include <linux/idr.h>
 
 /*
  * Returns passthrough_fh id that can be passed with FOPEN_PASSTHROUGH
@@ -10,18 +11,98 @@
  */
 int fuse_passthrough_open(struct fuse_conn *fc, int backing_fd)
 {
-	return -EINVAL;
+	struct file *passthrough_filp;
+	struct inode *passthrough_inode;
+	struct super_block *passthrough_sb;
+	struct fuse_passthrough *passthrough;
+	int res;
+
+	if (!fc->passthrough)
+		return -EPERM;
+
+	passthrough_filp = fget(backing_fd);
+	if (!passthrough_filp)
+		return -EBADF;
+
+	res = -EOPNOTSUPP;
+	if (!passthrough_filp->f_op->read_iter ||
+	    !passthrough_filp->f_op->write_iter)
+		goto out_fput;
+
+	passthrough_inode = file_inode(passthrough_filp);
+	passthrough_sb = passthrough_inode->i_sb;
+	res = -ELOOP;
+	if (passthrough_sb->s_stack_depth >= FILESYSTEM_MAX_STACK_DEPTH)
+		goto out_fput;
+
+	passthrough = kmalloc(sizeof(struct fuse_passthrough), GFP_KERNEL);
+	res = -ENOMEM;
+	if (!passthrough)
+		goto out_fput;
+
+	passthrough->filp = passthrough_filp;
+	refcount_set(&passthrough->count, 1);
+
+	idr_preload(GFP_KERNEL);
+	spin_lock(&fc->lock);
+	res = idr_alloc_cyclic(&fc->passthrough_files_map, passthrough, 1, 0,
+			       GFP_ATOMIC);
+	spin_unlock(&fc->lock);
+	idr_preload_end();
+
+	if (res < 0)
+		fuse_passthrough_free(passthrough);
+
+	return res;
+
+out_fput:
+	fput(passthrough_filp);
+
+	return res;
 }
 
 int fuse_passthrough_close(struct fuse_conn *fc, int passthrough_fh)
 {
-	return -EINVAL;
+	struct fuse_passthrough *passthrough;
+
+	if (!fc->passthrough)
+		return -EPERM;
+
+	if (passthrough_fh <= 0)
+		return -EINVAL;
+
+	spin_lock(&fc->lock);
+	passthrough = idr_remove(&fc->passthrough_files_map, passthrough_fh);
+	spin_unlock(&fc->lock);
+
+	if (!passthrough)
+		return -ENOENT;
+
+	fuse_passthrough_put(passthrough);
+
+	return 0;
 }
 
 int fuse_passthrough_setup(struct fuse_conn *fc, struct fuse_file *ff,
 			   struct fuse_open_out *openarg)
 {
-	return -EINVAL;
+	int passthrough_fh = openarg->passthrough_fh;
+	struct fuse_passthrough *passthrough;
+
+	if (passthrough_fh <= 0)
+		return -EINVAL;
+
+	rcu_read_lock();
+	passthrough = idr_find(&fc->passthrough_files_map, passthrough_fh);
+	if (passthrough && !refcount_inc_not_zero(&passthrough->count))
+		passthrough = NULL;
+	rcu_read_unlock();
+	if (!passthrough)
+		return -ENOENT;
+
+	ff->passthrough = passthrough;
+
+	return 0;
 }
 
 void fuse_passthrough_put(struct fuse_passthrough *passthrough)
@@ -36,5 +117,5 @@ void fuse_passthrough_free(struct fuse_passthrough *passthrough)
 		fput(passthrough->filp);
 		passthrough->filp = NULL;
 	}
-	kfree(passthrough);
+	kfree_rcu(passthrough, rcu);
 }
-- 
2.34.1


  parent reply	other threads:[~2023-05-19 12:58 UTC|newest]

Thread overview: 55+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-19 12:56 [PATCH v13 00/10] fuse: Add support for passthrough read/write Amir Goldstein
2023-05-19 12:56 ` [PATCH v13 01/10] fs: Generic function to convert iocb to rw flags Amir Goldstein
2023-05-19 12:56 ` [PATCH v13 02/10] fuse: Definitions and ioctl for passthrough Amir Goldstein
2023-05-19 15:12   ` Miklos Szeredi
2023-05-20 10:20     ` Amir Goldstein
2023-05-22 14:50       ` Miklos Szeredi
2023-05-24 10:00         ` Amir Goldstein
2023-05-19 12:56 ` Amir Goldstein [this message]
2023-05-19 12:56 ` [PATCH v13 04/10] fuse: Introduce synchronous read and write " Amir Goldstein
2023-05-19 12:57 ` [PATCH v13 05/10] fuse: Handle asynchronous read and write in passthrough Amir Goldstein
2023-05-22 15:20   ` Miklos Szeredi
2023-05-24 10:03     ` Amir Goldstein
2023-08-21 15:27       ` Amir Goldstein
2023-08-22 10:18         ` Amir Goldstein
2023-08-22 11:03           ` Miklos Szeredi
2023-08-22 13:22             ` Amir Goldstein
2023-08-22 14:06               ` Miklos Szeredi
2023-08-24 12:11             ` Amir Goldstein
2023-08-29 12:42               ` Miklos Szeredi
2023-05-19 12:57 ` [PATCH v13 06/10] fuse: Use daemon creds in passthrough mode Amir Goldstein
2023-05-19 12:57 ` [PATCH v13 07/10] fuse: Introduce passthrough for mmap Amir Goldstein
2023-05-19 12:57 ` [PATCH v13 08/10] fuse: update inode size/mtime after passthrough write Amir Goldstein
2023-09-25  6:41   ` [External] " Zhang Tianci
2023-09-25 10:43     ` Amir Goldstein
2023-09-26 15:31       ` Jens Axboe
2023-09-26 15:48         ` Amir Goldstein
2023-09-26 16:19           ` Jens Axboe
2023-09-26 16:56             ` Amir Goldstein
2023-05-19 12:57 ` [PATCH v13 09/10] fuse: invalidate atime after passthrough read/mmap Amir Goldstein
2023-05-19 12:57 ` [PATCH v13 10/10] fuse: setup a passthrough fd without a permanent backing id Amir Goldstein
2023-06-06 10:22   ` Fwd: " Miklos Szeredi
2023-06-06 11:00     ` [fuse-devel] " Amir Goldstein
2023-06-06 12:46       ` Miklos Szeredi
2023-05-20 10:28 ` [PATCH v13 00/10] fuse: Add support for passthrough read/write Amir Goldstein
2023-06-06  9:13 ` Amir Goldstein
2023-06-06  9:49   ` Miklos Szeredi
2023-06-06 11:19     ` Amir Goldstein
2023-06-06 13:06       ` Miklos Szeredi
2023-08-29 18:14         ` Amir Goldstein
2023-09-20 13:56           ` Amir Goldstein
2023-09-20 18:15             ` Miklos Szeredi
2023-09-21  7:33               ` Amir Goldstein
2023-09-21  8:21                 ` Miklos Szeredi
2023-09-21  9:17                   ` Amir Goldstein
2023-09-21  9:30                     ` Miklos Szeredi
2023-09-21 10:31                       ` Amir Goldstein
2023-09-21 11:50                         ` Miklos Szeredi
2023-09-22 12:45                           ` Amir Goldstein
2023-10-08 17:53                             ` Amir Goldstein
2023-10-10 14:31                               ` Miklos Szeredi
2023-10-10 15:14                                 ` Amir Goldstein
2023-10-14 16:01                                   ` Amir Goldstein
2023-10-16 10:30                                 ` Amir Goldstein
2023-10-16 13:21                                   ` Miklos Szeredi
2023-10-16 19:16                                   ` Bernd Schubert

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230519125705.598234-4-amir73il@gmail.com \
    --to=amir73il@gmail.com \
    --cc=balsini@android.com \
    --cc=drosen@google.com \
    --cc=fuse-devel@lists.sourceforge.net \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=paullawrence@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).