public inbox for [email protected]
 help / color / mirror / Atom feed
* [PATCH liburing] rem_buf/test: inital testing for OP_REMOVE_BUFFERS
@ 2020-12-08 21:57 Pavel Begunkov
  2020-12-08 22:42 ` Jens Axboe
  0 siblings, 1 reply; 4+ messages in thread
From: Pavel Begunkov @ 2020-12-08 21:57 UTC (permalink / raw)
  To: Jens Axboe, io-uring

Basic testing for IORING_OP_REMOVE_BUFFERS. test_rem_buf(IOSQE_ASYNC)
should check that it's doing locking right when punted async.

Signed-off-by: Pavel Begunkov <[email protected]>
---

NOTICE:
test_rem_buf(IOSQE_ASYNC) hangs with 5.10
because of double io_ring_submit_lock(). One of the iopoll patches
for 5.11 fixes that.

 test/read-write.c | 112 ++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 103 insertions(+), 9 deletions(-)

diff --git a/test/read-write.c b/test/read-write.c
index 84ea3a2..7f33ad4 100644
--- a/test/read-write.c
+++ b/test/read-write.c
@@ -472,10 +472,42 @@ static int test_buf_select_short(const char *filename, int nonvec)
 	return ret;
 }
 
-static int test_buf_select(const char *filename, int nonvec)
+static int provide_buffers_iovec(struct io_uring *ring, int bgid)
 {
 	struct io_uring_sqe *sqe;
 	struct io_uring_cqe *cqe;
+	int i, ret;
+
+	for (i = 0; i < BUFFERS; i++) {
+		sqe = io_uring_get_sqe(ring);
+		io_uring_prep_provide_buffers(sqe, vecs[i].iov_base,
+						vecs[i].iov_len, 1, bgid, i);
+	}
+
+	ret = io_uring_submit(ring);
+	if (ret != BUFFERS) {
+		fprintf(stderr, "submit: %d\n", ret);
+		return -1;
+	}
+
+	for (i = 0; i < BUFFERS; i++) {
+		ret = io_uring_wait_cqe(ring, &cqe);
+		if (ret) {
+			fprintf(stderr, "wait_cqe=%d\n", ret);
+			return 1;
+		}
+		if (cqe->res < 0) {
+			fprintf(stderr, "cqe->res=%d\n", cqe->res);
+			return 1;
+		}
+		io_uring_cqe_seen(ring, cqe);
+	}
+
+	return 0;
+}
+
+static int test_buf_select(const char *filename, int nonvec)
+{
 	struct io_uring_probe *p;
 	struct io_uring ring;
 	int ret, i;
@@ -509,29 +541,67 @@ static int test_buf_select(const char *filename, int nonvec)
 	for (i = 0; i < BUFFERS; i++)
 		memset(vecs[i].iov_base, 0x55, vecs[i].iov_len);
 
-	for (i = 0; i < BUFFERS; i++) {
+	ret = provide_buffers_iovec(&ring, 1);
+	if (ret)
+		return ret;
+
+	ret = __test_io(filename, &ring, 0, 0, 0, 0, nonvec, 1, 1, BS);
+	io_uring_queue_exit(&ring);
+	return ret;
+}
+
+static int test_rem_buf(int batch, int sqe_flags)
+{
+	struct io_uring_sqe *sqe;
+	struct io_uring_cqe *cqe;
+	struct io_uring ring;
+	int left, ret, nr = 0;
+	int bgid = 1;
+
+	if (no_buf_select)
+		return 0;
+
+	ret = io_uring_queue_init(64, &ring, 0);
+	if (ret) {
+		fprintf(stderr, "ring create failed: %d\n", ret);
+		return 1;
+	}
+
+	ret = provide_buffers_iovec(&ring, bgid);
+	if (ret)
+		return ret;
+
+	left = BUFFERS;
+	while (left) {
+		int to_rem = (left < batch) ? left : batch;
+
+		left -= to_rem;
 		sqe = io_uring_get_sqe(&ring);
-		io_uring_prep_provide_buffers(sqe, vecs[i].iov_base,
-						vecs[i].iov_len, 1, 1, i);
+		io_uring_prep_remove_buffers(sqe, to_rem, bgid);
+		sqe->user_data = to_rem;
+		sqe->flags |= sqe_flags;
+		++nr;
 	}
 
 	ret = io_uring_submit(&ring);
-	if (ret != BUFFERS) {
+	if (ret != nr) {
 		fprintf(stderr, "submit: %d\n", ret);
 		return -1;
 	}
 
-	for (i = 0; i < BUFFERS; i++) {
+	for (; nr > 0; nr--) {
 		ret = io_uring_wait_cqe(&ring, &cqe);
-		if (cqe->res < 0) {
+		if (ret) {
+			fprintf(stderr, "wait_cqe=%d\n", ret);
+			return 1;
+		}
+		if (cqe->res != cqe->user_data) {
 			fprintf(stderr, "cqe->res=%d\n", cqe->res);
 			return 1;
 		}
 		io_uring_cqe_seen(&ring, cqe);
 	}
 
-	ret = __test_io(filename, &ring, 0, 0, 0, 0, nonvec, 1, 1, BS);
-
 	io_uring_queue_exit(&ring);
 	return ret;
 }
@@ -786,6 +856,30 @@ int main(int argc, char *argv[])
 		goto err;
 	}
 
+	ret = test_rem_buf(1, 0);
+	if (ret) {
+		fprintf(stderr, "test_rem_buf by 1 failed\n");
+		goto err;
+	}
+
+	ret = test_rem_buf(10, 0);
+	if (ret) {
+		fprintf(stderr, "test_rem_buf by 10 failed\n");
+		goto err;
+	}
+
+	ret = test_rem_buf(2, IOSQE_IO_LINK);
+	if (ret) {
+		fprintf(stderr, "test_rem_buf link failed\n");
+		goto err;
+	}
+
+	ret = test_rem_buf(2, IOSQE_ASYNC);
+	if (ret) {
+		fprintf(stderr, "test_rem_buf async failed\n");
+		goto err;
+	}
+
 	srand((unsigned)time(NULL));
 	if (create_nonaligned_buffers()) {
 		fprintf(stderr, "file creation failed\n");
-- 
2.24.0


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH liburing] rem_buf/test: inital testing for OP_REMOVE_BUFFERS
  2020-12-08 21:57 [PATCH liburing] rem_buf/test: inital testing for OP_REMOVE_BUFFERS Pavel Begunkov
@ 2020-12-08 22:42 ` Jens Axboe
  2020-12-08 22:59   ` Pavel Begunkov
  0 siblings, 1 reply; 4+ messages in thread
From: Jens Axboe @ 2020-12-08 22:42 UTC (permalink / raw)
  To: Pavel Begunkov, io-uring

On 12/8/20 2:57 PM, Pavel Begunkov wrote:
> Basic testing for IORING_OP_REMOVE_BUFFERS. test_rem_buf(IOSQE_ASYNC)
> should check that it's doing locking right when punted async.
> 
> Signed-off-by: Pavel Begunkov <[email protected]>
> ---
> 
> NOTICE:
> test_rem_buf(IOSQE_ASYNC) hangs with 5.10
> because of double io_ring_submit_lock(). One of the iopoll patches
> for 5.11 fixes that.

Let's get just that into 5.10, and then we can fix up 5.11 around that.

-- 
Jens Axboe


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH liburing] rem_buf/test: inital testing for OP_REMOVE_BUFFERS
  2020-12-08 22:42 ` Jens Axboe
@ 2020-12-08 22:59   ` Pavel Begunkov
  2020-12-09  0:24     ` Jens Axboe
  0 siblings, 1 reply; 4+ messages in thread
From: Pavel Begunkov @ 2020-12-08 22:59 UTC (permalink / raw)
  To: Jens Axboe, io-uring

On 08/12/2020 22:42, Jens Axboe wrote:
> On 12/8/20 2:57 PM, Pavel Begunkov wrote:
>> Basic testing for IORING_OP_REMOVE_BUFFERS. test_rem_buf(IOSQE_ASYNC)
>> should check that it's doing locking right when punted async.
>>
>> Signed-off-by: Pavel Begunkov <[email protected]>
>> ---
>>
>> NOTICE:
>> test_rem_buf(IOSQE_ASYNC) hangs with 5.10
>> because of double io_ring_submit_lock(). One of the iopoll patches
>> for 5.11 fixes that.
> 
> Let's get just that into 5.10, and then we can fix up 5.11 around that.

If you mean get that kernel patch, then it's the one I commented on
yesterday.
[2/5] io_uring: fix racy IOPOLL completion

Either can split this test

-- 
Pavel Begunkov

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH liburing] rem_buf/test: inital testing for OP_REMOVE_BUFFERS
  2020-12-08 22:59   ` Pavel Begunkov
@ 2020-12-09  0:24     ` Jens Axboe
  0 siblings, 0 replies; 4+ messages in thread
From: Jens Axboe @ 2020-12-09  0:24 UTC (permalink / raw)
  To: Pavel Begunkov, io-uring

On 12/8/20 3:59 PM, Pavel Begunkov wrote:
> On 08/12/2020 22:42, Jens Axboe wrote:
>> On 12/8/20 2:57 PM, Pavel Begunkov wrote:
>>> Basic testing for IORING_OP_REMOVE_BUFFERS. test_rem_buf(IOSQE_ASYNC)
>>> should check that it's doing locking right when punted async.
>>>
>>> Signed-off-by: Pavel Begunkov <[email protected]>
>>> ---
>>>
>>> NOTICE:
>>> test_rem_buf(IOSQE_ASYNC) hangs with 5.10
>>> because of double io_ring_submit_lock(). One of the iopoll patches
>>> for 5.11 fixes that.
>>
>> Let's get just that into 5.10, and then we can fix up 5.11 around that.
> 
> If you mean get that kernel patch, then it's the one I commented on
> yesterday.
> [2/5] io_uring: fix racy IOPOLL completion

Right, of course...

> Either can split this test

Nah, that's fine. I'll get it applied, thanks.

-- 
Jens Axboe


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2020-12-09  0:25 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-12-08 21:57 [PATCH liburing] rem_buf/test: inital testing for OP_REMOVE_BUFFERS Pavel Begunkov
2020-12-08 22:42 ` Jens Axboe
2020-12-08 22:59   ` Pavel Begunkov
2020-12-09  0:24     ` Jens Axboe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox