Re: [PATCH 6/6] io_uring: add flag for allocating a fully sparse direct descriptor space

[Hao Xu <haoxu.linux@xxxxxxxxx>]

在 2022/5/9 下午11:50, Jens Axboe 写道:
> Currently to setup a fully sparse descriptor space upfront, the app needs
> to alloate an array of the full size and memset it to -1 and then pass
> that in. Make this a bit easier by allowing a flag that simply does
> this internally rather than needing to copy each slot separately.
>
> This works with IORING_REGISTER_FILES2 as the flag is set in struct
> io_uring_rsrc_register, and is only allow when the type is
> IORING_RSRC_FILE as this doesn't make sense for registered buffers.
>
> Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
> ---
>   fs/io_uring.c                 | 15 ++++++++++++---
>   include/uapi/linux/io_uring.h |  8 +++++++-
>   2 files changed, 19 insertions(+), 4 deletions(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index 644f57a46c5f..fe67fe939fac 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -9107,12 +9107,12 @@ static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
>   	for (i = 0; i < nr_args; i++, ctx->nr_user_files++) {
>   		struct io_fixed_file *file_slot;
>   
> -		if (copy_from_user(&fd, &fds[i], sizeof(fd))) {
> +		if (fds && copy_from_user(&fd, &fds[i], sizeof(fd))) {
>   			ret = -EFAULT;
>   			goto fail;
>   		}
>   		/* allow sparse sets */
> -		if (fd == -1) {
> +		if (!fds || fd == -1) {
>   			ret = -EINVAL;
>   			if (unlikely(*io_get_tag_slot(ctx->file_data, i)))
>   				goto fail;
> @@ -11755,14 +11755,20 @@ static __cold int io_register_rsrc(struct io_ring_ctx *ctx, void __user *arg,
>   	memset(&rr, 0, sizeof(rr));
>   	if (copy_from_user(&rr, arg, size))
>   		return -EFAULT;
> -	if (!rr.nr || rr.resv || rr.resv2)
> +	if (!rr.nr || rr.resv2)
> +		return -EINVAL;
> +	if (rr.flags & ~IORING_RSRC_REGISTER_SPARSE)
>   		return -EINVAL;
>   
>   	switch (type) {
>   	case IORING_RSRC_FILE:
> +		if (rr.flags & IORING_RSRC_REGISTER_SPARSE && rr.data)
> +			break;
>   		return io_sqe_files_register(ctx, u64_to_user_ptr(rr.data),
>   					     rr.nr, u64_to_user_ptr(rr.tags));
>   	case IORING_RSRC_BUFFER:
> +		if (rr.flags & IORING_RSRC_REGISTER_SPARSE)
> +			break;
>   		return io_sqe_buffers_register(ctx, u64_to_user_ptr(rr.data),
>   					       rr.nr, u64_to_user_ptr(rr.tags));
>   	}
> @@ -11931,6 +11937,9 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
>   		ret = io_sqe_buffers_unregister(ctx);
>   		break;
>   	case IORING_REGISTER_FILES:
> +		ret = -EFAULT;
> +		if (!arg)
> +			break;
>   		ret = io_sqe_files_register(ctx, arg, nr_args, NULL);
>   		break;
>   	case IORING_UNREGISTER_FILES:
> diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
> index b7f02a55032a..d09cf7c0d1fe 100644
> --- a/include/uapi/linux/io_uring.h
> +++ b/include/uapi/linux/io_uring.h
> @@ -396,9 +396,15 @@ struct io_uring_files_update {
>   	__aligned_u64 /* __s32 * */ fds;
>   };
>   
> +/*
> + * Register a fully sparse file sparse, rather than pass in an array of all

                                    ^space

> + * -1 file descriptors.
> + */
> +#define IORING_RSRC_REGISTER_SPARSE	(1U << 0)
> +
>   struct io_uring_rsrc_register {
>   	__u32 nr;
> -	__u32 resv;
> +	__u32 flags;
>   	__u64 resv2;
>   	__aligned_u64 data;
>   	__aligned_u64 tags;

This looks promising, we may eliminate cqes for open/accept_direct reqs.

feel free to add,
Reviewed-by: Hao Xu <howeyxu@xxxxxxxxxxx>