On 1/7/25 00:25, Bernd Schubert wrote:
This adds support for fuse request completion through ring SQEs
(FUSE_URING_CMD_COMMIT_AND_FETCH handling). After committing
the ring entry it becomes available for new fuse requests.
Handling of requests through the ring (SQE/CQE handling)
is complete now.
Fuse request data are copied through the mmaped ring buffer,
there is no support for any zero copy yet.
Reviewed-by: Pavel Begunkov <asml.silence@xxxxxxxxx> # io_uring
Signed-off-by: Bernd Schubert <bschubert@xxxxxxx>
---
fs/fuse/dev_uring.c | 450 ++++++++++++++++++++++++++++++++++++++++++++++++++
fs/fuse/dev_uring_i.h | 12 ++
fs/fuse/fuse_i.h | 4 +
3 files changed, 466 insertions(+)
diff --git a/fs/fuse/dev_uring.c b/fs/fuse/dev_uring.c
index b44ba4033615e01041313c040035b6da6af0ee17..f44e66a7ea577390da87e9ac7d118a9416898c28 100644
--- a/fs/fuse/dev_uring.c
+++ b/fs/fuse/dev_uring.c
@@ -26,6 +26,19 @@ bool fuse_uring_enabled(void)
...
+
+/*
+ * Write data to the ring buffer and send the request to userspace,
+ * userspace will read it
+ * This is comparable with classical read(/dev/fuse)
+ */
+static int fuse_uring_send_next_to_ring(struct fuse_ring_ent *ring_ent,
+ unsigned int issue_flags)
+{
+ int err = 0;
+ struct fuse_ring_queue *queue = ring_ent->queue;
+
+ err = fuse_uring_prepare_send(ring_ent);
+ if (err)
+ goto err;
+
+ spin_lock(&queue->lock);
+ ring_ent->state = FRRS_USERSPACE;
+ list_move(&ring_ent->list, &queue->ent_in_userspace);
+ spin_unlock(&queue->lock);
+
+ io_uring_cmd_done(ring_ent->cmd, 0, 0, issue_flags);
+ ring_ent->cmd = NULL;
I haven't checked if it races with some reallocation, but
you might want to consider clearing it under the spin.
spin_lock(&queue->lock);
...
cmd = ring_ent->cmd;
ring_ent->cmd = NULL;
spin_unlock(&queue->lock);
io_uring_cmd_done(cmd);
Can be done on top if even needed.
--
Pavel Begunkov
