public inbox for [email protected]
 help / color / mirror / Atom feed
From: Pavel Begunkov <[email protected]>
To: Jens Axboe <[email protected]>, [email protected]
Subject: [PATCH 3/4] io_uring: add IORING_OP_BPF
Date: Wed, 17 Feb 2021 12:38:05 +0000	[thread overview]
Message-ID: <b9ff5972067ba23026e2340d492bf0d06c538468.1613563964.git.asml.silence@gmail.com> (raw)
In-Reply-To: <[email protected]>

Wire up a new io_uring operation type IORING_OP_BPF, which executes a
specified BPF program from the registered prog table. It doesn't allow
to do anything useful for now, no BPF functions are allowed apart from
basic ones.

Signed-off-by: Pavel Begunkov <[email protected]>
---
 fs/io_uring.c                 | 83 +++++++++++++++++++++++++++++++++++
 include/uapi/linux/io_uring.h |  1 +
 2 files changed, 84 insertions(+)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 524cf1eb1cec..716881ca0b48 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -637,6 +637,11 @@ struct io_unlink {
 	struct filename			*filename;
 };
 
+struct io_bpf {
+	struct file			*file;
+	struct bpf_prog			*prog;
+};
+
 struct io_completion {
 	struct file			*file;
 	struct list_head		list;
@@ -773,6 +778,7 @@ struct io_kiocb {
 		struct io_shutdown	shutdown;
 		struct io_rename	rename;
 		struct io_unlink	unlink;
+		struct io_bpf		bpf;
 		/* use only after cleaning per-op data, see io_clean_op() */
 		struct io_completion	compl;
 	};
@@ -839,6 +845,10 @@ struct io_op_def {
 	unsigned		work_flags;
 };
 
+
+struct io_bpf_ctx {
+};
+
 static const struct io_op_def io_op_defs[] = {
 	[IORING_OP_NOP] = {},
 	[IORING_OP_READV] = {
@@ -1029,6 +1039,9 @@ static const struct io_op_def io_op_defs[] = {
 		.work_flags		= IO_WQ_WORK_MM | IO_WQ_WORK_FILES |
 						IO_WQ_WORK_FS | IO_WQ_WORK_BLKCG,
 	},
+	[IORING_OP_BPF]	= {
+		.work_flags		= IO_WQ_WORK_MM,
+	},
 };
 
 static void io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
@@ -1068,6 +1081,7 @@ static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
 static void io_req_task_queue(struct io_kiocb *req);
 static void io_submit_flush_completions(struct io_comp_state *cs,
 					struct io_ring_ctx *ctx);
+static void io_bpf_run(struct io_kiocb *req);
 
 static struct kmem_cache *req_cachep;
 
@@ -4208,6 +4222,53 @@ static int io_openat(struct io_kiocb *req, unsigned int issue_flags)
 	return io_openat2(req, issue_flags & IO_URING_F_NONBLOCK);
 }
 
+static int io_bpf_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	struct io_ring_ctx *ctx = req->ctx;
+	struct bpf_prog *prog;
+	unsigned int idx;
+
+	if (unlikely(ctx->flags & (IORING_SETUP_IOPOLL|IORING_SETUP_SQPOLL)))
+		return -EINVAL;
+	if (unlikely(req->flags & (REQ_F_FIXED_FILE | REQ_F_BUFFER_SELECT)))
+		return -EINVAL;
+	if (sqe->ioprio || sqe->len || sqe->cancel_flags)
+		return -EINVAL;
+	if (sqe->addr)
+		return -EINVAL;
+
+	idx = READ_ONCE(sqe->off);
+	if (unlikely(idx >= ctx->nr_bpf_progs))
+		return -EFAULT;
+	idx = array_index_nospec(idx, ctx->nr_bpf_progs);
+	prog = ctx->bpf_progs[idx].prog;
+	if (!prog)
+		return -EFAULT;
+
+	req->bpf.prog = prog;
+	return 0;
+}
+
+static void io_bpf_run_task_work(struct callback_head *cb)
+{
+	struct io_kiocb *req = container_of(cb, struct io_kiocb, task_work);
+	struct io_ring_ctx *ctx = req->ctx;
+
+	mutex_lock(&ctx->uring_lock);
+	io_bpf_run(req);
+	mutex_unlock(&ctx->uring_lock);
+}
+
+static int io_bpf(struct io_kiocb *req, unsigned int issue_flags)
+{
+	init_task_work(&req->task_work, io_bpf_run_task_work);
+	if (unlikely(io_req_task_work_add(req))) {
+		percpu_ref_get(&req->ctx->refs);
+		io_req_task_work_add_fallback(req, io_req_task_cancel);
+	}
+	return 0;
+}
+
 static int io_remove_buffers_prep(struct io_kiocb *req,
 				  const struct io_uring_sqe *sqe)
 {
@@ -6142,6 +6203,8 @@ static int io_req_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		return io_renameat_prep(req, sqe);
 	case IORING_OP_UNLINKAT:
 		return io_unlinkat_prep(req, sqe);
+	case IORING_OP_BPF:
+		return io_bpf_prep(req, sqe);
 	}
 
 	printk_once(KERN_WARNING "io_uring: unhandled opcode %d\n",
@@ -6380,6 +6443,9 @@ static int io_issue_sqe(struct io_kiocb *req, unsigned int issue_flags)
 	case IORING_OP_UNLINKAT:
 		ret = io_unlinkat(req, issue_flags);
 		break;
+	case IORING_OP_BPF:
+		ret = io_bpf(req, issue_flags);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
@@ -10267,6 +10333,23 @@ const struct bpf_verifier_ops bpf_io_uring_verifier_ops = {
 	.is_valid_access	= io_bpf_is_valid_access,
 };
 
+static void io_bpf_run(struct io_kiocb *req)
+{
+	struct io_ring_ctx *ctx = req->ctx;
+	struct io_bpf_ctx bpf_ctx;
+
+	lockdep_assert_held(&req->ctx->uring_lock);
+
+	if (unlikely(percpu_ref_is_dying(&ctx->refs))) {
+		io_req_complete(req, -EAGAIN);
+		return;
+	}
+
+	memset(&bpf_ctx, 0, sizeof(bpf_ctx));
+	BPF_PROG_RUN(req->bpf.prog, &bpf_ctx);
+	io_req_complete(req, 0);
+}
+
 SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
 		void __user *, arg, unsigned int, nr_args)
 {
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index d95e04d6d316..b75dfbf4f2cb 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -137,6 +137,7 @@ enum {
 	IORING_OP_SHUTDOWN,
 	IORING_OP_RENAMEAT,
 	IORING_OP_UNLINKAT,
+	IORING_OP_BPF,
 
 	/* this goes last, obviously */
 	IORING_OP_LAST,
-- 
2.24.0


  parent reply	other threads:[~2021-02-17 12:51 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-02-17 12:38 [RFC 0/4] add BPF-driven requests Pavel Begunkov
2021-02-17 12:38 ` [PATCH 1/4] bpf: add IOURING program type Pavel Begunkov
2021-02-17 12:38 ` [PATCH 2/4] io_uring: implement bpf prog registration Pavel Begunkov
2021-02-17 12:38 ` Pavel Begunkov [this message]
2021-02-17 12:38 ` [PATCH 4/4] io_uring: enable BPF to submit SQEs Pavel Begunkov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=b9ff5972067ba23026e2340d492bf0d06c538468.1613563964.git.asml.silence@gmail.com \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox