Re: SQPOLL question

[Jens Axboe <axboe@kernel.dk>]

On 9/8/20 8:36 AM, Jens Axboe wrote:
> On 9/8/20 12:47 AM, Josef wrote:
>>> If you're up for it, you could just clone my for-5.10/io_uring and base
>>> your SQPOLL testing on that. Should be finished, modulo bugs...
>>
>> yeah did some benchmark tests and I'm quite impressed, however accept
>> op seems to fail with -EBADF when the flag IOSQE_ASYNC is set, is that
>> known?
> 
> Nope, ran a quick test case here on the current tree, works for me.
> 
> Are you using for-5.10 and SQEPOLL + ASYNC accept? I'll give that a
> test spin.

This should do it for your testing, need to confirm this is absolutely
safe. But it'll make it work for the 5.10/io_uring setup of allowing
file open/closes.

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 80913973337a..e21a7a9c6a59 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -6757,7 +6757,7 @@ static enum sq_ret __io_sq_thread(struct io_ring_ctx *ctx,
 
 	mutex_lock(&ctx->uring_lock);
 	if (likely(!percpu_ref_is_dying(&ctx->refs)))
-		ret = io_submit_sqes(ctx, to_submit, NULL, -1);
+		ret = io_submit_sqes(ctx, to_submit, ctx->ring_file, ctx->ring_fd);
 	mutex_unlock(&ctx->uring_lock);
 
 	if (!io_sqring_full(ctx) && wq_has_sleeper(&ctx->sqo_sq_wait))
@@ -8966,6 +8966,11 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p,
 		goto err;
 	}
 
+	if (p->flags & IORING_SETUP_SQPOLL) {
+		ctx->ring_fd = fd;
+		ctx->ring_file = file;
+	}
+
 	ret = io_sq_offload_create(ctx, p);
 	if (ret)
 		goto err;

-- 
Jens Axboe