polyphony 0.45.1 → 0.46.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 5f54b88851ca73f121975839812e0a1ad5686d017199b83a6365deda2b656d3a
-  data.tar.gz: 92671a26215f19e36c0b0f49edb990a6818e2c64d7a8ed8bc9c4f2621a5de170
+  metadata.gz: 831c85a16a22fe3877044430a4d1cc4a270d18dc73e9ceca8d4826623ab0ddc8
+  data.tar.gz: f0432473abb769be2805ad354dddf44465789ebe2d5220f194660e4c6521ec87
 SHA512:
-  metadata.gz: fd3caf340b523eccb1f86b10e06996ec26704ed1906bd82188363fb5142901b585992df8d88d0ca20adc188e51a010144fba8f10843ba79503d2984f63a08a05
-  data.tar.gz: 2b44d80a6ac0bf3f0f42ece1b7d2008c3d0b933c1df11be7f698bbc9dae80341bd1d302a1e0ccaa56b223468b607b7f024bbaa949a88ff1c343028d6cd256ba5
+  metadata.gz: 9bfecbed04a8052c3a885dfde73975694313801d2a8341260041be84e4910ac459af71766f7df6fa96e5afad66a737b97e36531811bd47cab407ad43b239e073
+  data.tar.gz: 2caa40f7193cf1954b69adc5810844ad4e578a4472382602a6ee4204d548e8327344f913bc0aef076030e7deacb911cab1733b281f7a2e8849ffcbd009a00ebc

data/.github/workflows/test.yml

@@ -23,6 +23,8 @@ jobs:
       run: |
         gem install bundler
         bundle install
+    - name: Show Linux kernel version
+      run: uname -r
     - name: Compile C-extension
       run: bundle exec rake compile
     - name: Run tests

data/CHANGELOG.md

@@ -1,3 +1,38 @@
+## 0.46.1
+
+* Add `TCPServer#accept_loop`, `OpenSSL::SSL::SSLSocket#accept_loop` method
+* Fix compilation error on MacOS (#43)
+* Fix backtrace for `Timeout.timeout`
+* Add `Backend#timer_loop`
+
+## 0.46.0
+
+* Implement [io_uring backend](https://github.com/digital-fabric/polyphony/pull/44)
+
+## 0.45.5
+
+* Fix compilation error (#43)
+* Add support for resetting move_on_after, cancel_after timeouts
+* Optimize anti-event starvation polling
+* Implement optimized runqueue for better performance
+* Schedule parent with priority on uncaught exception
+* Fix race condition in `Mutex#synchronize` (#41)
+
+## 0.45.4
+
+* Improve signal trapping mechanism
+
+## 0.45.3
+
+* Don't swallow error in `Process#kill_and_await`
+* Add `Fiber#mailbox` attribute reader
+* Fix bug in `Fiber.await`
+* Implement `IO#getc`, `IO#getbyte`
+
+## 0.45.2
+
+* Rewrite `Fiber#<<`, `Fiber#await`, `Fiber#receive` in C
+
 ## 0.45.1
 
 * Fix Net::HTTP compatibility

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    polyphony (0.45.1)
+    polyphony (0.46.1)
 
 GEM
   remote: https://rubygems.org/
@@ -87,7 +87,7 @@ GEM
     rack (2.2.3)
     rainbow (3.0.0)
     rake (12.3.3)
-    rake-compiler (1.0.5)
+    rake-compiler (1.1.1)
       rake
     rb-fsevent (0.10.3)
     rb-inotify (0.10.1)
@@ -141,7 +141,7 @@ DEPENDENCIES
   polyphony!
   pry (= 0.13.1)
   rack (>= 2.0.8, < 2.3.0)
-  rake-compiler (= 1.0.5)
+  rake-compiler (= 1.1.1)
   redis (= 4.1.0)
   rubocop (= 0.85.1)
   sequel (= 5.34.0)

data/README.md

@@ -35,9 +35,9 @@
 Polyphony is a library for building concurrent applications in Ruby. Polyphony
 harnesses the power of [Ruby fibers](https://ruby-doc.org/core-2.5.1/Fiber.html)
 to provide a cooperative, sequential coroutine-based concurrency model. Under
-the hood, Polyphony uses [libev](https://github.com/enki/libev) as a
-high-performance event reactor that provides timers, I/O watchers and other
-asynchronous event primitives.
+the hood, Polyphony uses
+[io_uring](https://unixism.net/loti/what_is_io_uring.html) or
+[libev](https://github.com/enki/libev) to maximize I/O performance.
 
 ## Features
 

data/Rakefile

@@ -23,4 +23,4 @@ task :docs do
   exec 'RUBYOPT=-W0 jekyll serve -s docs -H ec2-18-156-117-172.eu-central-1.compute.amazonaws.com'
 end
 
-CLEAN.include "**/*.o", "**/*.so", "**/*.bundle", "**/*.jar", "pkg", "tmp"
+CLEAN.include "**/*.o", "**/*.so", "**/*.so.*", "**/*.a", "**/*.bundle", "**/*.jar", "pkg", "tmp"

data/TODO.md

@@ -1,15 +1,24 @@
-0.45.1
+## Roadmap for Polyphony 1.0
 
-- Adapter for Pry and IRB (Which fixes #5 and #6)
-- Redesign signal handling - the current mechanism is problematic in that it
-  does not address signals that do not kill, for instance HUP or USR1.
+- Check why worker-thread example doesn't work.
+- Add test that mimics the original design for Monocrono:
+  - 256 fibers each waiting for a message
+  - When message received do some blocking work using a `ThreadPool`
+  - Send messages, collect responses, check for correctness
 - Improve `#supervise`. It does not work as advertised, and seems to exhibit an
   inconsistent behaviour (see supervisor example).
-- Fix backtrace for `Timeout.timeout` API (see timeout example).
-- Check why worker-thread example doesn't work.
 
-0.46.0
+- io_uring
+  - Use playground.c to find out why we when submitting and waiting for
+    completion in single syscall signals seem to be blocked until the syscall
+    returns. Is this a bug in io_uring/liburing?
 
+-----------------------------------------------------
+
+- Add `Backend#splice(in, out, nbytes)` API
+- Adapter for io/console (what does `IO#raw` do?)
+- Adapter for Pry and IRB (Which fixes #5 and #6)
+- allow backend selection at runtime
 - Debugging
   - Eat your own dogfood: need a good tool to check what's going on when some
     test fails
@@ -122,8 +131,6 @@
   - discuss using `snooze` for ensuring responsiveness when executing CPU-bound work
 
 
-## 0.47
-
 ### Some more API work, more docs
 
 - sintra app with database access (postgresql)
@@ -135,14 +142,10 @@
   - proceed from there
 
 
-## 0.48
-
 ### Sinatra / Sidekiq
 
 - Pull out redis/postgres code, put into new `polyphony-xxx` gems
 
-## 0.49
-
 ### Testing && Docs
 
 - More tests
@@ -153,7 +156,10 @@
   - `IO.foreach`
   - `Process.waitpid`
 
-## 0.50 DNS
+### Quic / HTTP/3
+
+- Python impl: https://github.com/aiortc/aioquic/
+- Go impl: https://github.com/lucas-clemente/quic-go
 
 ### DNS client
 

data/bin/test

@@ -0,0 +1,4 @@
+#!/usr/bin/env bash
+set -e
+clear && POLYPHONY_USE_LIBEV=1 rake recompile && ruby test/run.rb
+clear && rake recompile && ruby test/run.rb

data/examples/io/raw.rb

@@ -0,0 +1,14 @@
+# frozen_string_literal: true
+
+require 'bundler/setup'
+require 'polyphony'
+require 'io/console'
+
+c = STDIN.raw(min: 1, tim: 0, &:getbyte)
+p result: c
+exit
+
+puts '?' * 40
+c = STDIN.getbyte
+puts '*' * 40
+p c

data/examples/io/reline.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+require 'bundler/setup'
+require 'polyphony'
+require 'polyphony/adapters/readline'
+require 'pry'
+
+$counter = 0
+timer = spin do
+  throttled_loop(5) do
+    $counter += 1
+  end
+end
+
+at_exit { timer.stop }
+
+puts 'try typing $counter to see the counter incremented in the background'
+binding.pry

data/examples/performance/fiber_transfer.rb

@@ -12,6 +12,7 @@ end
 def run(num_fibers)
   count = 0
 
+  GC.start
   GC.disable
 
   first = nil
@@ -36,13 +37,21 @@ def run(num_fibers)
   last.next = first
   
   t0 = Time.now
+  puts "start transfer..."
   first.transfer
   elapsed = Time.now - t0
 
-  puts "fibers: #{num_fibers} count: #{count} rate: #{count / elapsed}"
-  GC.start
+  rss = `ps -o rss= -p #{Process.pid}`.to_i
+
+  puts "fibers: #{num_fibers} rss: #{rss} count: #{count} rate: #{count / elapsed}"
+rescue Exception => e
+  puts "Stopped at #{count} fibers"
+  p e
 end
 
+puts "pid: #{Process.pid}"
 run(100)
-run(1000)
-run(10000)
+# run(1000)
+# run(10000)
+# run(100000)
+# run(400000)

data/examples/performance/multi_snooze.rb

@@ -18,7 +18,6 @@ def bm(fibers, iterations)
   Fiber.current.await_all_children
   dt = Time.now - t0
   puts "#{[fibers, iterations].inspect} setup: #{t0 - t_pre}s count: #{count} #{count / dt.to_f}/s"
-  Thread.current.run_queue_trace
 end
 
 GC.disable

data/examples/performance/thread-vs-fiber/polyphony_server.rb

@@ -4,42 +4,30 @@ require 'bundler/setup'
 require 'polyphony'
 require 'http/parser'
 
-$connection_count = 0
-
 def handle_client(socket)
-  $connection_count += 1
+  pending_requests = []
   parser = Http::Parser.new
-  reqs = []
-  parser.on_message_complete = proc do |env|
-    reqs << Object.new # parser
-  end
-  socket.read_loop |data|
+  parser.on_message_complete = proc { pending_requests << parser }
+
+  socket.recv_loop do |data|
     parser << data
-    while (req = reqs.shift)
-      handle_request(socket, req)
-      req = nil
-      snooze
-    end
+    write_response(socket) while pending_requests.shift
   end
 rescue IOError, SystemCallError => e
   # do nothing
 ensure
-  $connection_count -= 1
   socket&.close
 end
 
-def handle_request(client, parser)
+def write_response(socket)
   status_code = "200 OK"
   data = "Hello world!\n"
   headers = "Content-Type: text/plain\r\nContent-Length: #{data.bytesize}\r\n"
-  client.write "HTTP/1.1 #{status_code}\r\n#{headers}\r\n#{data}"
+  socket.write "HTTP/1.1 #{status_code}\r\n#{headers}\r\n#{data}"
 end
 
 server = TCPServer.open('0.0.0.0', 1234)
 puts "pid #{Process.pid}"
 puts "listening on port 1234"
 
-loop do
-  client = server.accept
-  spin { handle_client(client) }
-end
+server.accept_loop { |c| handle_client(c) }

data/ext/liburing/liburing.h

@@ -0,0 +1,585 @@
+/* SPDX-License-Identifier: MIT */
+#ifndef LIB_URING_H
+#define LIB_URING_H
+
+#include <sys/socket.h>
+#include <sys/uio.h>
+#include <sys/stat.h>
+#include <errno.h>
+#include <signal.h>
+#include <stdbool.h>
+#include <inttypes.h>
+#include <time.h>
+#include <linux/swab.h>
+#include "liburing/compat.h"
+#include "liburing/io_uring.h"
+#include "liburing/barrier.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Library interface to io_uring
+ */
+struct io_uring_sq {
+	unsigned *khead;
+	unsigned *ktail;
+	unsigned *kring_mask;
+	unsigned *kring_entries;
+	unsigned *kflags;
+	unsigned *kdropped;
+	unsigned *array;
+	struct io_uring_sqe *sqes;
+
+	unsigned sqe_head;
+	unsigned sqe_tail;
+
+	size_t ring_sz;
+	void *ring_ptr;
+
+	unsigned pad[4];
+};
+
+struct io_uring_cq {
+	unsigned *khead;
+	unsigned *ktail;
+	unsigned *kring_mask;
+	unsigned *kring_entries;
+	unsigned *kflags;
+	unsigned *koverflow;
+	struct io_uring_cqe *cqes;
+
+	size_t ring_sz;
+	void *ring_ptr;
+
+	unsigned pad[4];
+};
+
+struct io_uring {
+	struct io_uring_sq sq;
+	struct io_uring_cq cq;
+	unsigned flags;
+	int ring_fd;
+
+	unsigned pad[4];
+};
+
+/*
+ * Library interface
+ */
+
+/*
+ * return an allocated io_uring_probe structure, or NULL if probe fails (for
+ * example, if it is not available). The caller is responsible for freeing it
+ */
+extern struct io_uring_probe *io_uring_get_probe_ring(struct io_uring *ring);
+/* same as io_uring_get_probe_ring, but takes care of ring init and teardown */
+extern struct io_uring_probe *io_uring_get_probe(void);
+
+/*
+ * frees a probe allocated through io_uring_get_probe() or
+ * io_uring_get_probe_ring()
+ */
+extern void io_uring_free_probe(struct io_uring_probe *probe);
+
+static inline int io_uring_opcode_supported(struct io_uring_probe *p, int op)
+{
+	if (op > p->last_op)
+		return 0;
+	return (p->ops[op].flags & IO_URING_OP_SUPPORTED) != 0;
+}
+
+extern int io_uring_queue_init_params(unsigned entries, struct io_uring *ring,
+	struct io_uring_params *p);
+extern int io_uring_queue_init(unsigned entries, struct io_uring *ring,
+	unsigned flags);
+extern int io_uring_queue_mmap(int fd, struct io_uring_params *p,
+	struct io_uring *ring);
+extern int io_uring_ring_dontfork(struct io_uring *ring);
+extern void io_uring_queue_exit(struct io_uring *ring);
+unsigned io_uring_peek_batch_cqe(struct io_uring *ring,
+	struct io_uring_cqe **cqes, unsigned count);
+extern int io_uring_wait_cqes(struct io_uring *ring,
+	struct io_uring_cqe **cqe_ptr, unsigned wait_nr,
+	struct __kernel_timespec *ts, sigset_t *sigmask);
+extern int io_uring_wait_cqe_timeout(struct io_uring *ring,
+	struct io_uring_cqe **cqe_ptr, struct __kernel_timespec *ts);
+extern int io_uring_submit(struct io_uring *ring);
+extern int io_uring_submit_and_wait(struct io_uring *ring, unsigned wait_nr);
+extern struct io_uring_sqe *io_uring_get_sqe(struct io_uring *ring);
+
+extern int io_uring_register_buffers(struct io_uring *ring,
+					const struct iovec *iovecs,
+					unsigned nr_iovecs);
+extern int io_uring_unregister_buffers(struct io_uring *ring);
+extern int io_uring_register_files(struct io_uring *ring, const int *files,
+					unsigned nr_files);
+extern int io_uring_unregister_files(struct io_uring *ring);
+extern int io_uring_register_files_update(struct io_uring *ring, unsigned off,
+					int *files, unsigned nr_files);
+extern int io_uring_register_eventfd(struct io_uring *ring, int fd);
+extern int io_uring_register_eventfd_async(struct io_uring *ring, int fd);
+extern int io_uring_unregister_eventfd(struct io_uring *ring);
+extern int io_uring_register_probe(struct io_uring *ring,
+					struct io_uring_probe *p, unsigned nr);
+extern int io_uring_register_personality(struct io_uring *ring);
+extern int io_uring_unregister_personality(struct io_uring *ring, int id);
+extern int io_uring_register_restrictions(struct io_uring *ring,
+					  struct io_uring_restriction *res,
+					  unsigned int nr_res);
+extern int io_uring_enable_rings(struct io_uring *ring);
+extern int __io_uring_sqring_wait(struct io_uring *ring);
+
+/*
+ * Helper for the peek/wait single cqe functions. Exported because of that,
+ * but probably shouldn't be used directly in an application.
+ */
+extern int __io_uring_get_cqe(struct io_uring *ring,
+			      struct io_uring_cqe **cqe_ptr, unsigned submit,
+			      unsigned wait_nr, sigset_t *sigmask);
+
+#define LIBURING_UDATA_TIMEOUT	((__u64) -1)
+
+#define io_uring_for_each_cqe(ring, head, cqe)				\
+	/*								\
+	 * io_uring_smp_load_acquire() enforces the order of tail	\
+	 * and CQE reads.						\
+	 */								\
+	for (head = *(ring)->cq.khead;					\
+	     (cqe = (head != io_uring_smp_load_acquire((ring)->cq.ktail) ? \
+		&(ring)->cq.cqes[head & (*(ring)->cq.kring_mask)] : NULL)); \
+	     head++)							\
+
+/*
+ * Must be called after io_uring_for_each_cqe()
+ */
+static inline void io_uring_cq_advance(struct io_uring *ring,
+				       unsigned nr)
+{
+	if (nr) {
+		struct io_uring_cq *cq = &ring->cq;
+
+		/*
+		 * Ensure that the kernel only sees the new value of the head
+		 * index after the CQEs have been read.
+		 */
+		io_uring_smp_store_release(cq->khead, *cq->khead + nr);
+	}
+}
+
+/*
+ * Must be called after io_uring_{peek,wait}_cqe() after the cqe has
+ * been processed by the application.
+ */
+static inline void io_uring_cqe_seen(struct io_uring *ring,
+				     struct io_uring_cqe *cqe)
+{
+	if (cqe)
+		io_uring_cq_advance(ring, 1);
+}
+
+/*
+ * Command prep helpers
+ */
+static inline void io_uring_sqe_set_data(struct io_uring_sqe *sqe, void *data)
+{
+	sqe->user_data = (unsigned long) data;
+}
+
+static inline void *io_uring_cqe_get_data(const struct io_uring_cqe *cqe)
+{
+	return (void *) (uintptr_t) cqe->user_data;
+}
+
+static inline void io_uring_sqe_set_flags(struct io_uring_sqe *sqe,
+					  unsigned flags)
+{
+	sqe->flags = flags;
+}
+
+static inline void io_uring_prep_rw(int op, struct io_uring_sqe *sqe, int fd,
+				    const void *addr, unsigned len,
+				    __u64 offset)
+{
+	sqe->opcode = op;
+	sqe->flags = 0;
+	sqe->ioprio = 0;
+	sqe->fd = fd;
+	sqe->off = offset;
+	sqe->addr = (unsigned long) addr;
+	sqe->len = len;
+	sqe->rw_flags = 0;
+	sqe->user_data = 0;
+	sqe->__pad2[0] = sqe->__pad2[1] = sqe->__pad2[2] = 0;
+}
+
+static inline void io_uring_prep_splice(struct io_uring_sqe *sqe,
+					int fd_in, int64_t off_in,
+					int fd_out, int64_t off_out,
+					unsigned int nbytes,
+					unsigned int splice_flags)
+{
+	io_uring_prep_rw(IORING_OP_SPLICE, sqe, fd_out, NULL, nbytes, off_out);
+	sqe->splice_off_in = off_in;
+	sqe->splice_fd_in = fd_in;
+	sqe->splice_flags = splice_flags;
+}
+
+static inline void io_uring_prep_readv(struct io_uring_sqe *sqe, int fd,
+				       const struct iovec *iovecs,
+				       unsigned nr_vecs, off_t offset)
+{
+	io_uring_prep_rw(IORING_OP_READV, sqe, fd, iovecs, nr_vecs, offset);
+}
+
+static inline void io_uring_prep_read_fixed(struct io_uring_sqe *sqe, int fd,
+					    void *buf, unsigned nbytes,
+					    off_t offset, int buf_index)
+{
+	io_uring_prep_rw(IORING_OP_READ_FIXED, sqe, fd, buf, nbytes, offset);
+	sqe->buf_index = buf_index;
+}
+
+static inline void io_uring_prep_writev(struct io_uring_sqe *sqe, int fd,
+					const struct iovec *iovecs,
+					unsigned nr_vecs, off_t offset)
+{
+	io_uring_prep_rw(IORING_OP_WRITEV, sqe, fd, iovecs, nr_vecs, offset);
+}
+
+static inline void io_uring_prep_write_fixed(struct io_uring_sqe *sqe, int fd,
+					     const void *buf, unsigned nbytes,
+					     off_t offset, int buf_index)
+{
+	io_uring_prep_rw(IORING_OP_WRITE_FIXED, sqe, fd, buf, nbytes, offset);
+	sqe->buf_index = buf_index;
+}
+
+static inline void io_uring_prep_recvmsg(struct io_uring_sqe *sqe, int fd,
+					 struct msghdr *msg, unsigned flags)
+{
+	io_uring_prep_rw(IORING_OP_RECVMSG, sqe, fd, msg, 1, 0);
+	sqe->msg_flags = flags;
+}
+
+static inline void io_uring_prep_sendmsg(struct io_uring_sqe *sqe, int fd,
+					 const struct msghdr *msg, unsigned flags)
+{
+	io_uring_prep_rw(IORING_OP_SENDMSG, sqe, fd, msg, 1, 0);
+	sqe->msg_flags = flags;
+}
+
+static inline void io_uring_prep_poll_add(struct io_uring_sqe *sqe, int fd,
+					  unsigned poll_mask)
+{
+	io_uring_prep_rw(IORING_OP_POLL_ADD, sqe, fd, NULL, 0, 0);
+#if __BYTE_ORDER == __BIG_ENDIAN
+	poll_mask = __swahw32(poll_mask);
+#endif
+	sqe->poll32_events = poll_mask;
+}
+
+static inline void io_uring_prep_poll_remove(struct io_uring_sqe *sqe,
+					     void *user_data)
+{
+	io_uring_prep_rw(IORING_OP_POLL_REMOVE, sqe, -1, user_data, 0, 0);
+}
+
+static inline void io_uring_prep_fsync(struct io_uring_sqe *sqe, int fd,
+				       unsigned fsync_flags)
+{
+	io_uring_prep_rw(IORING_OP_FSYNC, sqe, fd, NULL, 0, 0);
+	sqe->fsync_flags = fsync_flags;
+}
+
+static inline void io_uring_prep_nop(struct io_uring_sqe *sqe)
+{
+	io_uring_prep_rw(IORING_OP_NOP, sqe, -1, NULL, 0, 0);
+}
+
+static inline void io_uring_prep_timeout(struct io_uring_sqe *sqe,
+					 struct __kernel_timespec *ts,
+					 unsigned count, unsigned flags)
+{
+	io_uring_prep_rw(IORING_OP_TIMEOUT, sqe, -1, ts, 1, count);
+	sqe->timeout_flags = flags;
+}
+
+static inline void io_uring_prep_timeout_remove(struct io_uring_sqe *sqe,
+						__u64 user_data, unsigned flags)
+{
+	io_uring_prep_rw(IORING_OP_TIMEOUT_REMOVE, sqe, -1,
+				(void *)(unsigned long)user_data, 0, 0);
+	sqe->timeout_flags = flags;
+}
+
+static inline void io_uring_prep_accept(struct io_uring_sqe *sqe, int fd,
+					struct sockaddr *addr,
+					socklen_t *addrlen, int flags)
+{
+	io_uring_prep_rw(IORING_OP_ACCEPT, sqe, fd, addr, 0,
+				(__u64) (unsigned long) addrlen);
+	sqe->accept_flags = flags;
+}
+
+static inline void io_uring_prep_cancel(struct io_uring_sqe *sqe, void *user_data,
+					int flags)
+{
+	io_uring_prep_rw(IORING_OP_ASYNC_CANCEL, sqe, -1, user_data, 0, 0);
+	sqe->cancel_flags = flags;
+}
+
+static inline void io_uring_prep_link_timeout(struct io_uring_sqe *sqe,
+					      struct __kernel_timespec *ts,
+					      unsigned flags)
+{
+	io_uring_prep_rw(IORING_OP_LINK_TIMEOUT, sqe, -1, ts, 1, 0);
+	sqe->timeout_flags = flags;
+}
+
+static inline void io_uring_prep_connect(struct io_uring_sqe *sqe, int fd,
+					 const struct sockaddr *addr,
+					 socklen_t addrlen)
+{
+	io_uring_prep_rw(IORING_OP_CONNECT, sqe, fd, addr, 0, addrlen);
+}
+
+static inline void io_uring_prep_files_update(struct io_uring_sqe *sqe,
+					      int *fds, unsigned nr_fds,
+					      int offset)
+{
+	io_uring_prep_rw(IORING_OP_FILES_UPDATE, sqe, -1, fds, nr_fds, offset);
+}
+
+static inline void io_uring_prep_fallocate(struct io_uring_sqe *sqe, int fd,
+					   int mode, off_t offset, off_t len)
+{
+
+	io_uring_prep_rw(IORING_OP_FALLOCATE, sqe, fd,
+			(const uintptr_t *) (unsigned long) len, mode, offset);
+}
+
+static inline void io_uring_prep_openat(struct io_uring_sqe *sqe, int dfd,
+					const char *path, int flags, mode_t mode)
+{
+	io_uring_prep_rw(IORING_OP_OPENAT, sqe, dfd, path, mode, 0);
+	sqe->open_flags = flags;
+}
+
+static inline void io_uring_prep_close(struct io_uring_sqe *sqe, int fd)
+{
+	io_uring_prep_rw(IORING_OP_CLOSE, sqe, fd, NULL, 0, 0);
+}
+
+static inline void io_uring_prep_read(struct io_uring_sqe *sqe, int fd,
+				      void *buf, unsigned nbytes, off_t offset)
+{
+	io_uring_prep_rw(IORING_OP_READ, sqe, fd, buf, nbytes, offset);
+}
+
+static inline void io_uring_prep_write(struct io_uring_sqe *sqe, int fd,
+				       const void *buf, unsigned nbytes, off_t offset)
+{
+	io_uring_prep_rw(IORING_OP_WRITE, sqe, fd, buf, nbytes, offset);
+}
+
+struct statx;
+static inline void io_uring_prep_statx(struct io_uring_sqe *sqe, int dfd,
+				const char *path, int flags, unsigned mask,
+				struct statx *statxbuf)
+{
+	io_uring_prep_rw(IORING_OP_STATX, sqe, dfd, path, mask,
+				(__u64) (unsigned long) statxbuf);
+	sqe->statx_flags = flags;
+}
+
+static inline void io_uring_prep_fadvise(struct io_uring_sqe *sqe, int fd,
+					 off_t offset, off_t len, int advice)
+{
+	io_uring_prep_rw(IORING_OP_FADVISE, sqe, fd, NULL, len, offset);
+	sqe->fadvise_advice = advice;
+}
+
+static inline void io_uring_prep_madvise(struct io_uring_sqe *sqe, void *addr,
+					 off_t length, int advice)
+{
+	io_uring_prep_rw(IORING_OP_MADVISE, sqe, -1, addr, length, 0);
+	sqe->fadvise_advice = advice;
+}
+
+static inline void io_uring_prep_send(struct io_uring_sqe *sqe, int sockfd,
+				      const void *buf, size_t len, int flags)
+{
+	io_uring_prep_rw(IORING_OP_SEND, sqe, sockfd, buf, len, 0);
+	sqe->msg_flags = flags;
+}
+
+static inline void io_uring_prep_recv(struct io_uring_sqe *sqe, int sockfd,
+				      void *buf, size_t len, int flags)
+{
+	io_uring_prep_rw(IORING_OP_RECV, sqe, sockfd, buf, len, 0);
+	sqe->msg_flags = flags;
+}
+
+static inline void io_uring_prep_openat2(struct io_uring_sqe *sqe, int dfd,
+					const char *path, struct open_how *how)
+{
+	io_uring_prep_rw(IORING_OP_OPENAT2, sqe, dfd, path, sizeof(*how),
+				(uint64_t) (uintptr_t) how);
+}
+
+struct epoll_event;
+static inline void io_uring_prep_epoll_ctl(struct io_uring_sqe *sqe, int epfd,
+					   int fd, int op,
+					   struct epoll_event *ev)
+{
+	io_uring_prep_rw(IORING_OP_EPOLL_CTL, sqe, epfd, ev, op, fd);
+}
+
+static inline void io_uring_prep_provide_buffers(struct io_uring_sqe *sqe,
+						 void *addr, int len, int nr,
+						 int bgid, int bid)
+{
+	io_uring_prep_rw(IORING_OP_PROVIDE_BUFFERS, sqe, nr, addr, len, bid);
+	sqe->buf_group = bgid;
+}
+
+static inline void io_uring_prep_remove_buffers(struct io_uring_sqe *sqe,
+						int nr, int bgid)
+{
+	io_uring_prep_rw(IORING_OP_REMOVE_BUFFERS, sqe, nr, NULL, 0, 0);
+	sqe->buf_group = bgid;
+}
+
+static inline void io_uring_prep_shutdown(struct io_uring_sqe *sqe, int fd,
+					  int how)
+{
+	io_uring_prep_rw(IORING_OP_SHUTDOWN, sqe, fd, NULL, how, 0);
+}
+
+/*
+ * Returns number of unconsumed (if SQPOLL) or unsubmitted entries exist in
+ * the SQ ring
+ */
+static inline unsigned io_uring_sq_ready(struct io_uring *ring)
+{
+	/*
+	 * Without a barrier, we could miss an update and think the SQ wasn't ready.
+	 * We don't need the load acquire for non-SQPOLL since then we drive updates.
+	 */
+	if (ring->flags & IORING_SETUP_SQPOLL)
+		return ring->sq.sqe_tail - io_uring_smp_load_acquire(ring->sq.khead);
+
+	/* always use real head, to avoid losing sync for short submit */
+	return ring->sq.sqe_tail - *ring->sq.khead;
+}
+
+/*
+ * Returns how much space is left in the SQ ring.
+ */
+static inline unsigned io_uring_sq_space_left(struct io_uring *ring)
+{
+	return *ring->sq.kring_entries - io_uring_sq_ready(ring);
+}
+
+/*
+ * Only applicable when using SQPOLL - allows the caller to wait for space
+ * to free up in the SQ ring, which happens when the kernel side thread has
+ * consumed one or more entries. If the SQ ring is currently non-full, no
+ * action is taken. Note: may return -EINVAL if the kernel doesn't support
+ * this feature.
+ */
+static inline int io_uring_sqring_wait(struct io_uring *ring)
+{
+	if (!(ring->flags & IORING_SETUP_SQPOLL))
+		return 0;
+	if (io_uring_sq_space_left(ring))
+		return 0;
+
+	return __io_uring_sqring_wait(ring);
+}
+
+/*
+ * Returns how many unconsumed entries are ready in the CQ ring
+ */
+static inline unsigned io_uring_cq_ready(struct io_uring *ring)
+{
+	return io_uring_smp_load_acquire(ring->cq.ktail) - *ring->cq.khead;
+}
+
+/*
+ * Returns true if the eventfd notification is currently enabled
+ */
+static inline bool io_uring_cq_eventfd_enabled(struct io_uring *ring)
+{
+	if (!ring->cq.kflags)
+		return true;
+
+	return !(*ring->cq.kflags & IORING_CQ_EVENTFD_DISABLED);
+}
+
+/*
+ * Toggle eventfd notification on or off, if an eventfd is registered with
+ * the ring.
+ */
+static inline int io_uring_cq_eventfd_toggle(struct io_uring *ring,
+					     bool enabled)
+{
+	uint32_t flags;
+
+	if (!!enabled == io_uring_cq_eventfd_enabled(ring))
+		return 0;
+
+	if (!ring->cq.kflags)
+		return -EOPNOTSUPP;
+
+	flags = *ring->cq.kflags;
+
+	if (enabled)
+		flags &= ~IORING_CQ_EVENTFD_DISABLED;
+	else
+		flags |= IORING_CQ_EVENTFD_DISABLED;
+
+	IO_URING_WRITE_ONCE(*ring->cq.kflags, flags);
+
+	return 0;
+}
+
+/*
+ * Return an IO completion, waiting for 'wait_nr' completions if one isn't
+ * readily available. Returns 0 with cqe_ptr filled in on success, -errno on
+ * failure.
+ */
+static inline int io_uring_wait_cqe_nr(struct io_uring *ring,
+				      struct io_uring_cqe **cqe_ptr,
+				      unsigned wait_nr)
+{
+	return __io_uring_get_cqe(ring, cqe_ptr, 0, wait_nr, NULL);
+}
+
+/*
+ * Return an IO completion, if one is readily available. Returns 0 with
+ * cqe_ptr filled in on success, -errno on failure.
+ */
+static inline int io_uring_peek_cqe(struct io_uring *ring,
+				    struct io_uring_cqe **cqe_ptr)
+{
+	return io_uring_wait_cqe_nr(ring, cqe_ptr, 0);
+}
+
+/*
+ * Return an IO completion, waiting for it if necessary. Returns 0 with
+ * cqe_ptr filled in on success, -errno on failure.
+ */
+static inline int io_uring_wait_cqe(struct io_uring *ring,
+				    struct io_uring_cqe **cqe_ptr)
+{
+	return io_uring_wait_cqe_nr(ring, cqe_ptr, 1);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif