Skip to contentSkip to conversation

io_uring: return void from io_put_kbuf_comp()

PATCH Superseded Ming Lei 2 participants 3 messages Mar 29, 2024 – Mar 30, 2024 io_uring

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Ming Lei

author
#0 lore raw

PATCH 1/1 io_uring: return void from io_put_kbuf_comp()

The two callers don't handle the return value of io_put_kbuf_comp(), so change its return type into void.

---
 io_uring/kbuf.h | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)
io_uring/kbuf.h +3 -9
index 1c7b654ee726..86931fa655ad 100644
@@ -119,18 +119,12 @@ static inline void __io_put_kbuf_list(struct io_kiocb *req,
 	}
 }
 
-static inline unsigned int io_put_kbuf_comp(struct io_kiocb *req)
+static inline void io_put_kbuf_comp(struct io_kiocb *req)
 {
-	unsigned int ret;
-
 	lockdep_assert_held(&req->ctx->completion_lock);
 
-	if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)))
-		return 0;
-
-	ret = IORING_CQE_F_BUFFER | (req->buf_index << IORING_CQE_BUFFER_SHIFT);
-	__io_put_kbuf_list(req, &req->ctx->io_buffers_comp);
-	return ret;
+	if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))
+		__io_put_kbuf_list(req, &req->ctx->io_buffers_comp);
 }
 
 static inline unsigned int io_put_kbuf(struct io_kiocb *req,
Signature · 1 line
2.41.0
maintainer
#1 lore raw
Quoted 2 lines · Ming Lei

The two callers don't handle the return value of io_put_kbuf_comp(), so change its return type into void.

We might want to consider changing the name of it too, it's a bit different in that it's just recyling/dropping this kbuf rather than posting a completion on behalf of it.

Maybe io_kbuf_drop() would be better. Would distuingish it from the normal use cases of "drop this kbuf and return the cflags representation of it, as I'm posting a completionw ith it".

Quoted 19 lines
---
 io_uring/kbuf.h | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)
io_uring/kbuf.h +1 -3
index 1c7b654ee726..86931fa655ad 100644
@@ -119,18 +119,12 @@ static inline void __io_put_kbuf_list(struct io_kiocb *req,
 	}
 }
 
-static inline unsigned int io_put_kbuf_comp(struct io_kiocb *req)
+static inline void io_put_kbuf_comp(struct io_kiocb *req)
 {
-	unsigned int ret;
-
 	lockdep_assert_held(&req->ctx->completion_lock);

io_uring/kbuf.h static inline void __io_put_kbuf_list(struct io_kiocb *req, · line 128

Quoted 10 lines
io_uring/kbuf.h +2 -6
 
-	if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)))
-		return 0;
-
-	ret = IORING_CQE_F_BUFFER | (req->buf_index << IORING_CQE_BUFFER_SHIFT);
-	__io_put_kbuf_list(req, &req->ctx->io_buffers_comp);
-	return ret;
+	if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))
+		__io_put_kbuf_list(req, &req->ctx->io_buffers_comp);
 }

If you post a v2 with the above suggestion, let's please just keep the flags checking how it is. It's consistent with what we do elsewhere.

Signature · 1 line
Jens Axboe
author
#2 lore raw
Quoted 32 lines · Jens Axboe
Quoted 2 lines · Ming Lei

The two callers don't handle the return value of io_put_kbuf_comp(), so change its return type into void.

We might want to consider changing the name of it too, it's a bit different in that it's just recyling/dropping this kbuf rather than posting a completion on behalf of it.

Maybe io_kbuf_drop() would be better. Would distuingish it from the normal use cases of "drop this kbuf and return the cflags representation of it, as I'm posting a completionw ith it".

Quoted 20 lines
---
 io_uring/kbuf.h | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)
io_uring/kbuf.h +1 -3
index 1c7b654ee726..86931fa655ad 100644
@@ -119,18 +119,12 @@ static inline void __io_put_kbuf_list(struct io_kiocb *req,
 	}
 }
 
-static inline unsigned int io_put_kbuf_comp(struct io_kiocb *req)
+static inline void io_put_kbuf_comp(struct io_kiocb *req)
 {
-	unsigned int ret;
-
 	lockdep_assert_held(&req->ctx->completion_lock);
 
Quoted 12 lines

io_uring/kbuf.h static inline void __io_put_kbuf_list(struct io_kiocb *req, · line 128

Quoted 9 lines
io_uring/kbuf.h +2 -6
-	if (!(req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)))
-		return 0;
-
-	ret = IORING_CQE_F_BUFFER | (req->buf_index << IORING_CQE_BUFFER_SHIFT);
-	__io_put_kbuf_list(req, &req->ctx->io_buffers_comp);
-	return ret;
+	if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING))
+		__io_put_kbuf_list(req, &req->ctx->io_buffers_comp);
 }

If you post a v2 with the above suggestion, let's please just keep the flags checking how it is. It's consistent with what we do elsewhere.

OK, I can rename it as io_kbuf_drop() in v2 and keep the original check style.

Also the following patch removes one io_put_kbuf_comp(), I will post v2 when that patch gets feedback or merged.

https://lore.kernel.org/io-uring/[email protected]/

thanks,
Ming