Re: [PATCH] io_uring: initialize task context before running the BPF loop

[Pavel Begunkov]

On 9/22/26 07:54, Yao Kai wrote:
...> ---
>   io_uring/loop.c | 5 +++++
>   1 file changed, 5 insertions(+)
>
> diff --git a/io_uring/loop.c b/io_uring/loop.c
> index bbbb6ef14e6a..c923cae6dfc7 100644
> --- a/io_uring/loop.c
> +++ b/io_uring/loop.c
> @@ -2,6 +2,7 @@
>   #include "io_uring.h"
>   #include "wait.h"
>   #include "loop.h"
> +#include "tctx.h"
>   
>   static inline int io_loop_nr_cqes(const struct io_ring_ctx *ctx,
>   				  const struct iou_loop_params *lp)
> @@ -84,6 +85,10 @@ int io_run_loop(struct io_ring_ctx *ctx)
>   	if (!io_allowed_run_tw(ctx))
>   		return -EEXIST;
>   
> +	ret = io_uring_add_tctx_node(ctx);
> +	if (unlikely(ret))
> +		return ret;
> +

Looks good in general, but can you move it a couple of lines
up before io_allowed_run_tw()? Thanks

>   	mutex_lock(&ctx->uring_lock);
>   	ret = __io_run_loop(ctx);
>   	mutex_unlock(&ctx->uring_lock);

--
Pavel Begunkov