Add a new notification type FUSE_NOTIFY_LOCK. This notification can be
sent by file server to signifiy that a previous locking request has
completed and actual caller should be woken up.
As of now we don't support blocking variant of posix locks and daemon
returns -EOPNOTSUPP. Reason being that it can lead to deadlocks.
Virtqueue size is limited and it is possible we fill virtqueue with
all the requests of fcntl(F_SETLKW) and wait for reply. And later a
subsequent unlock request can't make progress because virtqueue is full.
And that means F_SETLKW can't make progress and we are deadlocked.
This problem is not limited to posix locks only. I think blocking remote
flock and open file description locks should face the same issue. Right
now fuse does not support open file description locks, so its not
a problem. But fuse/virtiofs does support remote flock and they can use
same mechanism too.
Use notification queue to solve this problem. After submitting lock
request device will send a reply asking requester to wait. Once lock is
available, requester will get a notification saying lock is available.
That way we don't keep the request virtueue busy while we are waiting for
lock and further unlock requests can make progress.
When we get a reply in response to lock request, we need a way to know
if we need to wait for notification or not. I have overloaded the
fuse_out_header->error field. If value is ->error is 1, that's a signal
to caller to wait for lock notification. Overloading ->error in this way
is not the best way to do it. But I am running out of ideas.
Signed-off-by: Vivek Goyal <vgoyal@xxxxxxxxxx>
Signed-off-by: Ioannis Angelakopoulos <iangelak@xxxxxxxxxx>
---
fs/fuse/virtio_fs.c | 73 +++++++++++++++++++++++++++++++++++++++
include/uapi/linux/fuse.h | 11 +++++-
2 files changed, 83 insertions(+), 1 deletion(-)
diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c
index 8d33879d62fb..1634ea2d0555 100644
--- a/fs/fuse/virtio_fs.c
+++ b/fs/fuse/virtio_fs.c
@@ -48,6 +48,7 @@ struct virtio_fs_vq {
struct virtqueue *vq; /* protected by ->lock */
struct work_struct done_work;
struct list_head queued_reqs;
+ struct list_head wait_reqs; /* Requests waiting for notification */
struct list_head end_reqs; /* End these requests */
struct virtio_fs_notify_node *notify_nodes;
struct list_head notify_reqs; /* List for queuing notify requests */
@@ -575,13 +576,72 @@ static int virtio_fs_enqueue_all_notify(struct virtio_fs_vq *fsvq)
return 0;
}
+static int notify_complete_waiting_req(struct virtio_fs *vfs,
+ struct fuse_notify_lock_out *out_args)
+{
+ /* TODO: Handle multiqueue */
+ struct virtio_fs_vq *fsvq = &vfs->vqs[vfs->first_reqq_idx];
+ struct fuse_req *req, *next;
+ bool found = false;
+
+ /* Find waiting request with the unique number and end it */
+ spin_lock(&fsvq->lock);
+ list_for_each_entry_safe(req, next, &fsvq->wait_reqs, list) {
+ if (req->in.h.unique == out_args->unique) {
+ list_del_init(&req->list);
+ clear_bit(FR_SENT, &req->flags);
+ /* Transfer error code from notify */
+ req->out.h.error = out_args->error;
+ found = true;
+ break;
+ }
+ }
+ spin_unlock(&fsvq->lock);
+
+ /*
+ * TODO: It is possible that some re-ordering happens in notify
+ * comes before request is complete. Deal with it.
+ */
+ if (found) {
+ end_req_dec_in_flight(req, fsvq);
+ } else
+ pr_debug("virtio-fs: Did not find waiting request with unique=0x%llx\n",
+ out_args->unique);
+
+ return 0;
+}
+
+static int virtio_fs_handle_notify(struct virtio_fs *vfs,
+ struct virtio_fs_notify *notify)
+{
+ int ret = 0;
+ struct fuse_out_header *oh = ¬ify->out_hdr;
+ struct fuse_notify_lock_out *lo;
+
+ /*
+ * For notifications, oh.unique is 0 and oh->error contains code
+ * for which notification as arrived.
+ */
+ switch (oh->error) {
+ case FUSE_NOTIFY_LOCK:
+ lo = (struct fuse_notify_lock_out *) ¬ify->outarg;
+ notify_complete_waiting_req(vfs, lo);
+ break;
+ default:
+ pr_err("virtio-fs: Unexpected notification %d\n", oh->error);
+ }
+ return ret;
+}
+
static void virtio_fs_notify_done_work(struct work_struct *work)
{
struct virtio_fs_vq *fsvq = container_of(work, struct virtio_fs_vq,
done_work);
struct virtqueue *vq = fsvq->vq;
+ struct virtio_fs *vfs = vq->vdev->priv;
LIST_HEAD(reqs);
struct virtio_fs_notify_node *notifyn, *next;
+ struct fuse_out_header *oh;
spin_lock(&fsvq->lock);
do {
@@ -597,6 +657,10 @@ static void virtio_fs_notify_done_work(struct work_struct *work)
/* Process notify */
list_for_each_entry_safe(notifyn, next, &reqs, list) {
+ oh = ¬ifyn->notify.out_hdr;
+ WARN_ON(oh->unique);
+ /* Handle notification */
+ virtio_fs_handle_notify(vfs, ¬ifyn->notify);
spin_lock(&fsvq->lock);
dec_in_flight_req(fsvq);
list_del_init(¬ifyn->list);
@@ -696,6 +760,14 @@ static void virtio_fs_request_complete(struct fuse_req *req,
* TODO verify that server properly follows FUSE protocol
* (oh.uniq, oh.len)
*/
+ if (req->out.h.error == 1) {
+ /* Wait for notification to complete request */
+ spin_lock(&fsvq->lock);
+ list_add_tail(&req->list, &fsvq->wait_reqs);
+ spin_unlock(&fsvq->lock);
+ return;
+ }
+
args = req->args;
copy_args_from_argbuf(args, req);
@@ -788,6 +860,7 @@ static int virtio_fs_init_vq(struct virtio_fs *fs, struct virtio_fs_vq *fsvq,
strncpy(fsvq->name, name, VQ_NAME_LEN);
spin_lock_init(&fsvq->lock);
INIT_LIST_HEAD(&fsvq->queued_reqs);
+ INIT_LIST_HEAD(&fsvq->wait_reqs);
INIT_LIST_HEAD(&fsvq->end_reqs);
INIT_LIST_HEAD(&fsvq->notify_reqs);
init_completion(&fsvq->in_flight_zero);
diff --git a/include/uapi/linux/fuse.h b/include/uapi/linux/fuse.h
index 36ed092227fa..46838551ea84 100644
--- a/include/uapi/linux/fuse.h
+++ b/include/uapi/linux/fuse.h
@@ -184,6 +184,8 @@
*
* 7.34
* - add FUSE_SYNCFS
+ * 7.35
+ * - add FUSE_NOTIFY_LOCK
*/
#ifndef _LINUX_FUSE_H
@@ -219,7 +221,7 @@
#define FUSE_KERNEL_VERSION 7
/** Minor version number of this interface */
-#define FUSE_KERNEL_MINOR_VERSION 34
+#define FUSE_KERNEL_MINOR_VERSION 35
/** The node ID of the root inode */
#define FUSE_ROOT_ID 1
@@ -529,6 +531,7 @@ enum fuse_notify_code {
FUSE_NOTIFY_STORE = 4,
FUSE_NOTIFY_RETRIEVE = 5,
FUSE_NOTIFY_DELETE = 6,
+ FUSE_NOTIFY_LOCK = 7,
FUSE_NOTIFY_CODE_MAX,
};
@@ -920,6 +923,12 @@ struct fuse_notify_retrieve_in {
uint64_t dummy4;
};
+struct fuse_notify_lock_out {
+ uint64_t unique;
+ int32_t error;
+ int32_t padding;
+};
+
/* Device ioctls: */
#define FUSE_DEV_IOC_MAGIC 229
#define FUSE_DEV_IOC_CLONE _IOR(FUSE_DEV_IOC_MAGIC, 0, uint32_t)
--
2.31.1
