uringmachine 0.1

Sign up to get free protection for your applications and to get access to all the features.
Files changed (307) hide show
  1. checksums.yaml +7 -0
  2. data/.github/dependabot.yml +12 -0
  3. data/.github/workflows/test.yml +35 -0
  4. data/.gitignore +59 -0
  5. data/.gitmodules +3 -0
  6. data/CHANGELOG.md +7 -0
  7. data/Gemfile +3 -0
  8. data/LICENSE +21 -0
  9. data/README.md +11 -0
  10. data/Rakefile +39 -0
  11. data/TODO.md +0 -0
  12. data/examples/echo_server.rb +52 -0
  13. data/examples/event_loop.rb +69 -0
  14. data/examples/fibers.rb +105 -0
  15. data/examples/http_server.rb +56 -0
  16. data/examples/http_server_multishot.rb +57 -0
  17. data/examples/http_server_simpler.rb +34 -0
  18. data/ext/um/extconf.rb +71 -0
  19. data/ext/um/iou.h +101 -0
  20. data/ext/um/op_ctx.c +138 -0
  21. data/ext/um/ring.c +755 -0
  22. data/ext/um/um.c +267 -0
  23. data/ext/um/um.h +97 -0
  24. data/ext/um/um_class.c +175 -0
  25. data/ext/um/um_ext.c +11 -0
  26. data/ext/um/um_op.c +87 -0
  27. data/ext/um/um_utils.c +23 -0
  28. data/lib/uringmachine/version.rb +3 -0
  29. data/lib/uringmachine.rb +8 -0
  30. data/test/helper.rb +70 -0
  31. data/test/test_iou.rb +876 -0
  32. data/test/test_um.rb +168 -0
  33. data/uringmachine.gemspec +27 -0
  34. data/vendor/liburing/.github/actions/codespell/stopwords +7 -0
  35. data/vendor/liburing/.github/pull_request_template.md +86 -0
  36. data/vendor/liburing/.github/workflows/build.yml +137 -0
  37. data/vendor/liburing/.github/workflows/codespell.yml +25 -0
  38. data/vendor/liburing/.github/workflows/shellcheck.yml +20 -0
  39. data/vendor/liburing/.gitignore +41 -0
  40. data/vendor/liburing/CHANGELOG +111 -0
  41. data/vendor/liburing/CITATION.cff +11 -0
  42. data/vendor/liburing/COPYING +502 -0
  43. data/vendor/liburing/COPYING.GPL +339 -0
  44. data/vendor/liburing/LICENSE +20 -0
  45. data/vendor/liburing/Makefile +96 -0
  46. data/vendor/liburing/Makefile.common +7 -0
  47. data/vendor/liburing/Makefile.quiet +11 -0
  48. data/vendor/liburing/README +106 -0
  49. data/vendor/liburing/SECURITY.md +6 -0
  50. data/vendor/liburing/configure +624 -0
  51. data/vendor/liburing/debian/README.Debian +7 -0
  52. data/vendor/liburing/debian/changelog +38 -0
  53. data/vendor/liburing/debian/control +39 -0
  54. data/vendor/liburing/debian/copyright +49 -0
  55. data/vendor/liburing/debian/liburing-dev.install +4 -0
  56. data/vendor/liburing/debian/liburing-dev.manpages +5 -0
  57. data/vendor/liburing/debian/liburing2.install +1 -0
  58. data/vendor/liburing/debian/liburing2.symbols +56 -0
  59. data/vendor/liburing/debian/patches/series +1 -0
  60. data/vendor/liburing/debian/rules +29 -0
  61. data/vendor/liburing/debian/source/format +1 -0
  62. data/vendor/liburing/debian/source/local-options +2 -0
  63. data/vendor/liburing/debian/source/options +1 -0
  64. data/vendor/liburing/debian/watch +3 -0
  65. data/vendor/liburing/examples/Makefile +53 -0
  66. data/vendor/liburing/examples/helpers.c +62 -0
  67. data/vendor/liburing/examples/helpers.h +7 -0
  68. data/vendor/liburing/examples/io_uring-close-test.c +123 -0
  69. data/vendor/liburing/examples/io_uring-cp.c +282 -0
  70. data/vendor/liburing/examples/io_uring-test.c +112 -0
  71. data/vendor/liburing/examples/io_uring-udp.c +403 -0
  72. data/vendor/liburing/examples/link-cp.c +193 -0
  73. data/vendor/liburing/examples/napi-busy-poll-client.c +509 -0
  74. data/vendor/liburing/examples/napi-busy-poll-server.c +450 -0
  75. data/vendor/liburing/examples/poll-bench.c +101 -0
  76. data/vendor/liburing/examples/proxy.c +2461 -0
  77. data/vendor/liburing/examples/proxy.h +102 -0
  78. data/vendor/liburing/examples/rsrc-update-bench.c +100 -0
  79. data/vendor/liburing/examples/send-zerocopy.c +658 -0
  80. data/vendor/liburing/examples/ucontext-cp.c +258 -0
  81. data/vendor/liburing/liburing-ffi.pc.in +12 -0
  82. data/vendor/liburing/liburing.pc.in +12 -0
  83. data/vendor/liburing/liburing.spec +66 -0
  84. data/vendor/liburing/make-debs.sh +55 -0
  85. data/vendor/liburing/src/Makefile +129 -0
  86. data/vendor/liburing/src/arch/aarch64/lib.h +47 -0
  87. data/vendor/liburing/src/arch/aarch64/syscall.h +91 -0
  88. data/vendor/liburing/src/arch/generic/lib.h +17 -0
  89. data/vendor/liburing/src/arch/generic/syscall.h +100 -0
  90. data/vendor/liburing/src/arch/riscv64/lib.h +48 -0
  91. data/vendor/liburing/src/arch/riscv64/syscall.h +100 -0
  92. data/vendor/liburing/src/arch/syscall-defs.h +94 -0
  93. data/vendor/liburing/src/arch/x86/lib.h +11 -0
  94. data/vendor/liburing/src/arch/x86/syscall.h +296 -0
  95. data/vendor/liburing/src/ffi.c +15 -0
  96. data/vendor/liburing/src/include/liburing/barrier.h +81 -0
  97. data/vendor/liburing/src/include/liburing/io_uring.h +818 -0
  98. data/vendor/liburing/src/include/liburing.h +1602 -0
  99. data/vendor/liburing/src/int_flags.h +11 -0
  100. data/vendor/liburing/src/lib.h +52 -0
  101. data/vendor/liburing/src/liburing-ffi.map +211 -0
  102. data/vendor/liburing/src/liburing.map +104 -0
  103. data/vendor/liburing/src/nolibc.c +55 -0
  104. data/vendor/liburing/src/queue.c +468 -0
  105. data/vendor/liburing/src/register.c +374 -0
  106. data/vendor/liburing/src/setup.c +689 -0
  107. data/vendor/liburing/src/setup.h +9 -0
  108. data/vendor/liburing/src/syscall.c +29 -0
  109. data/vendor/liburing/src/syscall.h +53 -0
  110. data/vendor/liburing/src/version.c +21 -0
  111. data/vendor/liburing/test/232c93d07b74.c +305 -0
  112. data/vendor/liburing/test/35fa71a030ca.c +329 -0
  113. data/vendor/liburing/test/500f9fbadef8.c +91 -0
  114. data/vendor/liburing/test/7ad0e4b2f83c.c +94 -0
  115. data/vendor/liburing/test/8a9973408177.c +107 -0
  116. data/vendor/liburing/test/917257daa0fe.c +54 -0
  117. data/vendor/liburing/test/Makefile +297 -0
  118. data/vendor/liburing/test/a0908ae19763.c +59 -0
  119. data/vendor/liburing/test/a4c0b3decb33.c +181 -0
  120. data/vendor/liburing/test/accept-link.c +255 -0
  121. data/vendor/liburing/test/accept-non-empty.c +256 -0
  122. data/vendor/liburing/test/accept-reuse.c +163 -0
  123. data/vendor/liburing/test/accept-test.c +83 -0
  124. data/vendor/liburing/test/accept.c +919 -0
  125. data/vendor/liburing/test/across-fork.c +284 -0
  126. data/vendor/liburing/test/b19062a56726.c +54 -0
  127. data/vendor/liburing/test/b5837bd5311d.c +78 -0
  128. data/vendor/liburing/test/bind-listen.c +408 -0
  129. data/vendor/liburing/test/buf-ring-nommap.c +123 -0
  130. data/vendor/liburing/test/buf-ring-put.c +83 -0
  131. data/vendor/liburing/test/buf-ring.c +473 -0
  132. data/vendor/liburing/test/ce593a6c480a.c +139 -0
  133. data/vendor/liburing/test/close-opath.c +123 -0
  134. data/vendor/liburing/test/config +14 -0
  135. data/vendor/liburing/test/connect-rep.c +204 -0
  136. data/vendor/liburing/test/connect.c +442 -0
  137. data/vendor/liburing/test/coredump.c +60 -0
  138. data/vendor/liburing/test/cq-full.c +97 -0
  139. data/vendor/liburing/test/cq-overflow.c +530 -0
  140. data/vendor/liburing/test/cq-peek-batch.c +103 -0
  141. data/vendor/liburing/test/cq-ready.c +95 -0
  142. data/vendor/liburing/test/cq-size.c +65 -0
  143. data/vendor/liburing/test/d4ae271dfaae.c +96 -0
  144. data/vendor/liburing/test/d77a67ed5f27.c +65 -0
  145. data/vendor/liburing/test/defer-taskrun.c +391 -0
  146. data/vendor/liburing/test/defer-tw-timeout.c +173 -0
  147. data/vendor/liburing/test/defer.c +319 -0
  148. data/vendor/liburing/test/double-poll-crash.c +195 -0
  149. data/vendor/liburing/test/drop-submit.c +94 -0
  150. data/vendor/liburing/test/eeed8b54e0df.c +120 -0
  151. data/vendor/liburing/test/empty-eownerdead.c +45 -0
  152. data/vendor/liburing/test/eploop.c +74 -0
  153. data/vendor/liburing/test/eventfd-disable.c +179 -0
  154. data/vendor/liburing/test/eventfd-reg.c +77 -0
  155. data/vendor/liburing/test/eventfd-ring.c +98 -0
  156. data/vendor/liburing/test/eventfd.c +113 -0
  157. data/vendor/liburing/test/evloop.c +73 -0
  158. data/vendor/liburing/test/exec-target.c +6 -0
  159. data/vendor/liburing/test/exit-no-cleanup.c +117 -0
  160. data/vendor/liburing/test/fadvise.c +202 -0
  161. data/vendor/liburing/test/fallocate.c +265 -0
  162. data/vendor/liburing/test/fc2a85cb02ef.c +132 -0
  163. data/vendor/liburing/test/fd-install.c +500 -0
  164. data/vendor/liburing/test/fd-pass.c +237 -0
  165. data/vendor/liburing/test/fdinfo.c +419 -0
  166. data/vendor/liburing/test/file-register.c +1189 -0
  167. data/vendor/liburing/test/file-update.c +231 -0
  168. data/vendor/liburing/test/file-verify.c +654 -0
  169. data/vendor/liburing/test/files-exit-hang-poll.c +114 -0
  170. data/vendor/liburing/test/files-exit-hang-timeout.c +137 -0
  171. data/vendor/liburing/test/fixed-buf-iter.c +115 -0
  172. data/vendor/liburing/test/fixed-buf-merge.c +101 -0
  173. data/vendor/liburing/test/fixed-hugepage.c +411 -0
  174. data/vendor/liburing/test/fixed-link.c +90 -0
  175. data/vendor/liburing/test/fixed-reuse.c +160 -0
  176. data/vendor/liburing/test/fpos.c +255 -0
  177. data/vendor/liburing/test/fsnotify.c +118 -0
  178. data/vendor/liburing/test/fsync.c +224 -0
  179. data/vendor/liburing/test/futex.c +571 -0
  180. data/vendor/liburing/test/hardlink.c +170 -0
  181. data/vendor/liburing/test/helpers.c +318 -0
  182. data/vendor/liburing/test/helpers.h +108 -0
  183. data/vendor/liburing/test/ignore-single-mmap.c +48 -0
  184. data/vendor/liburing/test/init-mem.c +164 -0
  185. data/vendor/liburing/test/io-cancel.c +561 -0
  186. data/vendor/liburing/test/io_uring_enter.c +264 -0
  187. data/vendor/liburing/test/io_uring_passthrough.c +482 -0
  188. data/vendor/liburing/test/io_uring_register.c +503 -0
  189. data/vendor/liburing/test/io_uring_setup.c +110 -0
  190. data/vendor/liburing/test/iopoll-leak.c +85 -0
  191. data/vendor/liburing/test/iopoll-overflow.c +118 -0
  192. data/vendor/liburing/test/iopoll.c +465 -0
  193. data/vendor/liburing/test/lfs-openat-write.c +119 -0
  194. data/vendor/liburing/test/lfs-openat.c +273 -0
  195. data/vendor/liburing/test/link-timeout.c +1108 -0
  196. data/vendor/liburing/test/link.c +497 -0
  197. data/vendor/liburing/test/link_drain.c +255 -0
  198. data/vendor/liburing/test/madvise.c +195 -0
  199. data/vendor/liburing/test/min-timeout-wait.c +354 -0
  200. data/vendor/liburing/test/min-timeout.c +233 -0
  201. data/vendor/liburing/test/mkdir.c +112 -0
  202. data/vendor/liburing/test/msg-ring-fd.c +331 -0
  203. data/vendor/liburing/test/msg-ring-flags.c +212 -0
  204. data/vendor/liburing/test/msg-ring-overflow.c +159 -0
  205. data/vendor/liburing/test/msg-ring.c +467 -0
  206. data/vendor/liburing/test/multicqes_drain.c +429 -0
  207. data/vendor/liburing/test/napi-test.c +215 -0
  208. data/vendor/liburing/test/napi-test.sh +48 -0
  209. data/vendor/liburing/test/no-mmap-inval.c +42 -0
  210. data/vendor/liburing/test/nolibc.c +62 -0
  211. data/vendor/liburing/test/nop-all-sizes.c +99 -0
  212. data/vendor/liburing/test/nop.c +177 -0
  213. data/vendor/liburing/test/nvme.h +169 -0
  214. data/vendor/liburing/test/ooo-file-unreg.c +82 -0
  215. data/vendor/liburing/test/open-close.c +261 -0
  216. data/vendor/liburing/test/open-direct-link.c +188 -0
  217. data/vendor/liburing/test/open-direct-pick.c +180 -0
  218. data/vendor/liburing/test/openat2.c +312 -0
  219. data/vendor/liburing/test/personality.c +204 -0
  220. data/vendor/liburing/test/pipe-bug.c +95 -0
  221. data/vendor/liburing/test/pipe-eof.c +83 -0
  222. data/vendor/liburing/test/pipe-reuse.c +105 -0
  223. data/vendor/liburing/test/poll-cancel-all.c +496 -0
  224. data/vendor/liburing/test/poll-cancel-ton.c +135 -0
  225. data/vendor/liburing/test/poll-cancel.c +228 -0
  226. data/vendor/liburing/test/poll-link.c +221 -0
  227. data/vendor/liburing/test/poll-many.c +230 -0
  228. data/vendor/liburing/test/poll-mshot-overflow.c +265 -0
  229. data/vendor/liburing/test/poll-mshot-update.c +323 -0
  230. data/vendor/liburing/test/poll-race-mshot.c +276 -0
  231. data/vendor/liburing/test/poll-race.c +105 -0
  232. data/vendor/liburing/test/poll-ring.c +48 -0
  233. data/vendor/liburing/test/poll-v-poll.c +353 -0
  234. data/vendor/liburing/test/poll.c +327 -0
  235. data/vendor/liburing/test/probe.c +135 -0
  236. data/vendor/liburing/test/read-before-exit.c +129 -0
  237. data/vendor/liburing/test/read-mshot-empty.c +153 -0
  238. data/vendor/liburing/test/read-mshot.c +404 -0
  239. data/vendor/liburing/test/read-write.c +1013 -0
  240. data/vendor/liburing/test/recv-msgall-stream.c +398 -0
  241. data/vendor/liburing/test/recv-msgall.c +263 -0
  242. data/vendor/liburing/test/recv-multishot.c +602 -0
  243. data/vendor/liburing/test/recvsend_bundle.c +691 -0
  244. data/vendor/liburing/test/reg-fd-only.c +131 -0
  245. data/vendor/liburing/test/reg-hint.c +56 -0
  246. data/vendor/liburing/test/reg-reg-ring.c +90 -0
  247. data/vendor/liburing/test/regbuf-merge.c +91 -0
  248. data/vendor/liburing/test/register-restrictions.c +633 -0
  249. data/vendor/liburing/test/rename.c +132 -0
  250. data/vendor/liburing/test/ring-leak.c +283 -0
  251. data/vendor/liburing/test/ring-leak2.c +249 -0
  252. data/vendor/liburing/test/ringbuf-read.c +196 -0
  253. data/vendor/liburing/test/ringbuf-status.c +242 -0
  254. data/vendor/liburing/test/rsrc_tags.c +461 -0
  255. data/vendor/liburing/test/runtests-loop.sh +16 -0
  256. data/vendor/liburing/test/runtests-quiet.sh +11 -0
  257. data/vendor/liburing/test/runtests.sh +168 -0
  258. data/vendor/liburing/test/rw_merge_test.c +98 -0
  259. data/vendor/liburing/test/self.c +91 -0
  260. data/vendor/liburing/test/send-zerocopy.c +971 -0
  261. data/vendor/liburing/test/send_recv.c +412 -0
  262. data/vendor/liburing/test/send_recvmsg.c +444 -0
  263. data/vendor/liburing/test/shared-wq.c +84 -0
  264. data/vendor/liburing/test/short-read.c +75 -0
  265. data/vendor/liburing/test/shutdown.c +165 -0
  266. data/vendor/liburing/test/sigfd-deadlock.c +88 -0
  267. data/vendor/liburing/test/single-issuer.c +169 -0
  268. data/vendor/liburing/test/skip-cqe.c +428 -0
  269. data/vendor/liburing/test/socket-getsetsock-cmd.c +346 -0
  270. data/vendor/liburing/test/socket-io-cmd.c +237 -0
  271. data/vendor/liburing/test/socket-rw-eagain.c +149 -0
  272. data/vendor/liburing/test/socket-rw-offset.c +149 -0
  273. data/vendor/liburing/test/socket-rw.c +137 -0
  274. data/vendor/liburing/test/socket.c +408 -0
  275. data/vendor/liburing/test/splice.c +512 -0
  276. data/vendor/liburing/test/sq-full-cpp.cc +45 -0
  277. data/vendor/liburing/test/sq-full.c +45 -0
  278. data/vendor/liburing/test/sq-poll-dup.c +211 -0
  279. data/vendor/liburing/test/sq-poll-kthread.c +169 -0
  280. data/vendor/liburing/test/sq-poll-share.c +138 -0
  281. data/vendor/liburing/test/sq-space_left.c +159 -0
  282. data/vendor/liburing/test/sqpoll-disable-exit.c +196 -0
  283. data/vendor/liburing/test/sqpoll-exec.c +132 -0
  284. data/vendor/liburing/test/sqpoll-exit-hang.c +78 -0
  285. data/vendor/liburing/test/sqpoll-sleep.c +69 -0
  286. data/vendor/liburing/test/statx.c +172 -0
  287. data/vendor/liburing/test/stdout.c +232 -0
  288. data/vendor/liburing/test/submit-and-wait.c +108 -0
  289. data/vendor/liburing/test/submit-link-fail.c +156 -0
  290. data/vendor/liburing/test/submit-reuse.c +237 -0
  291. data/vendor/liburing/test/symlink.c +117 -0
  292. data/vendor/liburing/test/sync-cancel.c +235 -0
  293. data/vendor/liburing/test/teardowns.c +58 -0
  294. data/vendor/liburing/test/test.h +36 -0
  295. data/vendor/liburing/test/thread-exit.c +143 -0
  296. data/vendor/liburing/test/timeout-new.c +256 -0
  297. data/vendor/liburing/test/timeout.c +1798 -0
  298. data/vendor/liburing/test/truncate.c +186 -0
  299. data/vendor/liburing/test/tty-write-dpoll.c +60 -0
  300. data/vendor/liburing/test/unlink.c +112 -0
  301. data/vendor/liburing/test/version.c +25 -0
  302. data/vendor/liburing/test/wait-timeout.c +287 -0
  303. data/vendor/liburing/test/waitid.c +373 -0
  304. data/vendor/liburing/test/wakeup-hang.c +162 -0
  305. data/vendor/liburing/test/wq-aff.c +146 -0
  306. data/vendor/liburing/test/xattr.c +442 -0
  307. metadata +412 -0
@@ -0,0 +1,1602 @@
1
+ /* SPDX-License-Identifier: MIT */
2
+ #ifndef LIB_URING_H
3
+ #define LIB_URING_H
4
+
5
+ #include <sys/socket.h>
6
+ #include <sys/stat.h>
7
+ #include <sys/uio.h>
8
+ #include <errno.h>
9
+ #include <signal.h>
10
+ #include <stdbool.h>
11
+ #include <inttypes.h>
12
+ #include <time.h>
13
+ #include <fcntl.h>
14
+ #include <sched.h>
15
+ #include <linux/swab.h>
16
+ #include <sys/wait.h>
17
+ #include "liburing/compat.h"
18
+ #include "liburing/io_uring.h"
19
+ #include "liburing/io_uring_version.h"
20
+ #include "liburing/barrier.h"
21
+
22
+ #ifndef uring_unlikely
23
+ #define uring_unlikely(cond) __builtin_expect(!!(cond), 0)
24
+ #endif
25
+
26
+ #ifndef uring_likely
27
+ #define uring_likely(cond) __builtin_expect(!!(cond), 1)
28
+ #endif
29
+
30
+ #ifndef IOURINGINLINE
31
+ #define IOURINGINLINE static inline
32
+ #endif
33
+
34
+ #ifdef __alpha__
35
+ /*
36
+ * alpha and mips are the exceptions, all other architectures have
37
+ * common numbers for new system calls.
38
+ */
39
+ #ifndef __NR_io_uring_setup
40
+ #define __NR_io_uring_setup 535
41
+ #endif
42
+ #ifndef __NR_io_uring_enter
43
+ #define __NR_io_uring_enter 536
44
+ #endif
45
+ #ifndef __NR_io_uring_register
46
+ #define __NR_io_uring_register 537
47
+ #endif
48
+ #elif defined __mips__
49
+ #ifndef __NR_io_uring_setup
50
+ #define __NR_io_uring_setup (__NR_Linux + 425)
51
+ #endif
52
+ #ifndef __NR_io_uring_enter
53
+ #define __NR_io_uring_enter (__NR_Linux + 426)
54
+ #endif
55
+ #ifndef __NR_io_uring_register
56
+ #define __NR_io_uring_register (__NR_Linux + 427)
57
+ #endif
58
+ #else /* !__alpha__ and !__mips__ */
59
+ #ifndef __NR_io_uring_setup
60
+ #define __NR_io_uring_setup 425
61
+ #endif
62
+ #ifndef __NR_io_uring_enter
63
+ #define __NR_io_uring_enter 426
64
+ #endif
65
+ #ifndef __NR_io_uring_register
66
+ #define __NR_io_uring_register 427
67
+ #endif
68
+ #endif
69
+
70
+ #ifdef __cplusplus
71
+ extern "C" {
72
+ #endif
73
+
74
+ /*
75
+ * Library interface to io_uring
76
+ */
77
+ struct io_uring_sq {
78
+ unsigned *khead;
79
+ unsigned *ktail;
80
+ // Deprecated: use `ring_mask` instead of `*kring_mask`
81
+ unsigned *kring_mask;
82
+ // Deprecated: use `ring_entries` instead of `*kring_entries`
83
+ unsigned *kring_entries;
84
+ unsigned *kflags;
85
+ unsigned *kdropped;
86
+ unsigned *array;
87
+ struct io_uring_sqe *sqes;
88
+
89
+ unsigned sqe_head;
90
+ unsigned sqe_tail;
91
+
92
+ size_t ring_sz;
93
+ void *ring_ptr;
94
+
95
+ unsigned ring_mask;
96
+ unsigned ring_entries;
97
+
98
+ unsigned pad[2];
99
+ };
100
+
101
+ struct io_uring_cq {
102
+ unsigned *khead;
103
+ unsigned *ktail;
104
+ // Deprecated: use `ring_mask` instead of `*kring_mask`
105
+ unsigned *kring_mask;
106
+ // Deprecated: use `ring_entries` instead of `*kring_entries`
107
+ unsigned *kring_entries;
108
+ unsigned *kflags;
109
+ unsigned *koverflow;
110
+ struct io_uring_cqe *cqes;
111
+
112
+ size_t ring_sz;
113
+ void *ring_ptr;
114
+
115
+ unsigned ring_mask;
116
+ unsigned ring_entries;
117
+
118
+ unsigned pad[2];
119
+ };
120
+
121
+ struct io_uring {
122
+ struct io_uring_sq sq;
123
+ struct io_uring_cq cq;
124
+ unsigned flags;
125
+ int ring_fd;
126
+
127
+ unsigned features;
128
+ int enter_ring_fd;
129
+ __u8 int_flags;
130
+ __u8 pad[3];
131
+ unsigned pad2;
132
+ };
133
+
134
+ /*
135
+ * Library interface
136
+ */
137
+
138
+ /*
139
+ * return an allocated io_uring_probe structure, or NULL if probe fails (for
140
+ * example, if it is not available). The caller is responsible for freeing it
141
+ */
142
+ struct io_uring_probe *io_uring_get_probe_ring(struct io_uring *ring);
143
+ /* same as io_uring_get_probe_ring, but takes care of ring init and teardown */
144
+ struct io_uring_probe *io_uring_get_probe(void);
145
+
146
+ /*
147
+ * frees a probe allocated through io_uring_get_probe() or
148
+ * io_uring_get_probe_ring()
149
+ */
150
+ void io_uring_free_probe(struct io_uring_probe *probe);
151
+
152
+ IOURINGINLINE int io_uring_opcode_supported(const struct io_uring_probe *p,
153
+ int op)
154
+ {
155
+ if (op > p->last_op)
156
+ return 0;
157
+ return (p->ops[op].flags & IO_URING_OP_SUPPORTED) != 0;
158
+ }
159
+
160
+ int io_uring_queue_init_mem(unsigned entries, struct io_uring *ring,
161
+ struct io_uring_params *p,
162
+ void *buf, size_t buf_size);
163
+ int io_uring_queue_init_params(unsigned entries, struct io_uring *ring,
164
+ struct io_uring_params *p);
165
+ int io_uring_queue_init(unsigned entries, struct io_uring *ring,
166
+ unsigned flags);
167
+ int io_uring_queue_mmap(int fd, struct io_uring_params *p,
168
+ struct io_uring *ring);
169
+ int io_uring_ring_dontfork(struct io_uring *ring);
170
+ void io_uring_queue_exit(struct io_uring *ring);
171
+ unsigned io_uring_peek_batch_cqe(struct io_uring *ring,
172
+ struct io_uring_cqe **cqes, unsigned count);
173
+ int io_uring_wait_cqes(struct io_uring *ring, struct io_uring_cqe **cqe_ptr,
174
+ unsigned wait_nr, struct __kernel_timespec *ts,
175
+ sigset_t *sigmask);
176
+ int io_uring_wait_cqes_min_timeout(struct io_uring *ring,
177
+ struct io_uring_cqe **cqe_ptr,
178
+ unsigned wait_nr,
179
+ struct __kernel_timespec *ts,
180
+ unsigned int min_ts_usec,
181
+ sigset_t *sigmask);
182
+ int io_uring_wait_cqe_timeout(struct io_uring *ring,
183
+ struct io_uring_cqe **cqe_ptr,
184
+ struct __kernel_timespec *ts);
185
+ int io_uring_submit(struct io_uring *ring);
186
+ int io_uring_submit_and_wait(struct io_uring *ring, unsigned wait_nr);
187
+ int io_uring_submit_and_wait_timeout(struct io_uring *ring,
188
+ struct io_uring_cqe **cqe_ptr,
189
+ unsigned wait_nr,
190
+ struct __kernel_timespec *ts,
191
+ sigset_t *sigmask);
192
+ int io_uring_submit_and_wait_min_timeout(struct io_uring *ring,
193
+ struct io_uring_cqe **cqe_ptr,
194
+ unsigned wait_nr,
195
+ struct __kernel_timespec *ts,
196
+ unsigned min_wait,
197
+ sigset_t *sigmask);
198
+
199
+ int io_uring_register_buffers(struct io_uring *ring, const struct iovec *iovecs,
200
+ unsigned nr_iovecs);
201
+ int io_uring_register_buffers_tags(struct io_uring *ring,
202
+ const struct iovec *iovecs,
203
+ const __u64 *tags, unsigned nr);
204
+ int io_uring_register_buffers_sparse(struct io_uring *ring, unsigned nr);
205
+ int io_uring_register_buffers_update_tag(struct io_uring *ring,
206
+ unsigned off,
207
+ const struct iovec *iovecs,
208
+ const __u64 *tags, unsigned nr);
209
+ int io_uring_unregister_buffers(struct io_uring *ring);
210
+
211
+ int io_uring_register_files(struct io_uring *ring, const int *files,
212
+ unsigned nr_files);
213
+ int io_uring_register_files_tags(struct io_uring *ring, const int *files,
214
+ const __u64 *tags, unsigned nr);
215
+ int io_uring_register_files_sparse(struct io_uring *ring, unsigned nr);
216
+ int io_uring_register_files_update_tag(struct io_uring *ring, unsigned off,
217
+ const int *files, const __u64 *tags,
218
+ unsigned nr_files);
219
+
220
+ int io_uring_unregister_files(struct io_uring *ring);
221
+ int io_uring_register_files_update(struct io_uring *ring, unsigned off,
222
+ const int *files, unsigned nr_files);
223
+ int io_uring_register_eventfd(struct io_uring *ring, int fd);
224
+ int io_uring_register_eventfd_async(struct io_uring *ring, int fd);
225
+ int io_uring_unregister_eventfd(struct io_uring *ring);
226
+ int io_uring_register_probe(struct io_uring *ring, struct io_uring_probe *p,
227
+ unsigned nr);
228
+ int io_uring_register_personality(struct io_uring *ring);
229
+ int io_uring_unregister_personality(struct io_uring *ring, int id);
230
+ int io_uring_register_restrictions(struct io_uring *ring,
231
+ struct io_uring_restriction *res,
232
+ unsigned int nr_res);
233
+ int io_uring_enable_rings(struct io_uring *ring);
234
+ int __io_uring_sqring_wait(struct io_uring *ring);
235
+ int io_uring_register_iowq_aff(struct io_uring *ring, size_t cpusz,
236
+ const cpu_set_t *mask);
237
+ int io_uring_unregister_iowq_aff(struct io_uring *ring);
238
+ int io_uring_register_iowq_max_workers(struct io_uring *ring,
239
+ unsigned int *values);
240
+ int io_uring_register_ring_fd(struct io_uring *ring);
241
+ int io_uring_unregister_ring_fd(struct io_uring *ring);
242
+ int io_uring_close_ring_fd(struct io_uring *ring);
243
+ int io_uring_register_buf_ring(struct io_uring *ring,
244
+ struct io_uring_buf_reg *reg, unsigned int flags);
245
+ int io_uring_unregister_buf_ring(struct io_uring *ring, int bgid);
246
+ int io_uring_buf_ring_head(struct io_uring *ring, int buf_group, uint16_t *head);
247
+ int io_uring_register_sync_cancel(struct io_uring *ring,
248
+ struct io_uring_sync_cancel_reg *reg);
249
+
250
+ int io_uring_register_file_alloc_range(struct io_uring *ring,
251
+ unsigned off, unsigned len);
252
+
253
+ int io_uring_register_napi(struct io_uring *ring, struct io_uring_napi *napi);
254
+ int io_uring_unregister_napi(struct io_uring *ring, struct io_uring_napi *napi);
255
+
256
+ int io_uring_register_clock(struct io_uring *ring,
257
+ struct io_uring_clock_register *arg);
258
+
259
+ int io_uring_get_events(struct io_uring *ring);
260
+ int io_uring_submit_and_get_events(struct io_uring *ring);
261
+
262
+ /*
263
+ * io_uring syscalls.
264
+ */
265
+ int io_uring_enter(unsigned int fd, unsigned int to_submit,
266
+ unsigned int min_complete, unsigned int flags, sigset_t *sig);
267
+ int io_uring_enter2(unsigned int fd, unsigned int to_submit,
268
+ unsigned int min_complete, unsigned int flags,
269
+ sigset_t *sig, size_t sz);
270
+ int io_uring_setup(unsigned int entries, struct io_uring_params *p);
271
+ int io_uring_register(unsigned int fd, unsigned int opcode, const void *arg,
272
+ unsigned int nr_args);
273
+
274
+ /*
275
+ * Mapped buffer ring alloc/register + unregister/free helpers
276
+ */
277
+ struct io_uring_buf_ring *io_uring_setup_buf_ring(struct io_uring *ring,
278
+ unsigned int nentries,
279
+ int bgid, unsigned int flags,
280
+ int *ret);
281
+ int io_uring_free_buf_ring(struct io_uring *ring, struct io_uring_buf_ring *br,
282
+ unsigned int nentries, int bgid);
283
+
284
+ /*
285
+ * Helper for the peek/wait single cqe functions. Exported because of that,
286
+ * but probably shouldn't be used directly in an application.
287
+ */
288
+ int __io_uring_get_cqe(struct io_uring *ring,
289
+ struct io_uring_cqe **cqe_ptr, unsigned submit,
290
+ unsigned wait_nr, sigset_t *sigmask);
291
+
292
+ #define LIBURING_UDATA_TIMEOUT ((__u64) -1)
293
+
294
+ /*
295
+ * Calculates the step size for CQE iteration.
296
+ * For standard CQE's its 1, for big CQE's its two.
297
+ */
298
+ #define io_uring_cqe_shift(ring) \
299
+ (!!((ring)->flags & IORING_SETUP_CQE32))
300
+
301
+ #define io_uring_cqe_index(ring,ptr,mask) \
302
+ (((ptr) & (mask)) << io_uring_cqe_shift(ring))
303
+
304
+ #define io_uring_for_each_cqe(ring, head, cqe) \
305
+ /* \
306
+ * io_uring_smp_load_acquire() enforces the order of tail \
307
+ * and CQE reads. \
308
+ */ \
309
+ for (head = *(ring)->cq.khead; \
310
+ (cqe = (head != io_uring_smp_load_acquire((ring)->cq.ktail) ? \
311
+ &(ring)->cq.cqes[io_uring_cqe_index(ring, head, (ring)->cq.ring_mask)] : NULL)); \
312
+ head++) \
313
+
314
+ /*
315
+ * Must be called after io_uring_for_each_cqe()
316
+ */
317
+ IOURINGINLINE void io_uring_cq_advance(struct io_uring *ring, unsigned nr)
318
+ {
319
+ if (nr) {
320
+ struct io_uring_cq *cq = &ring->cq;
321
+
322
+ /*
323
+ * Ensure that the kernel only sees the new value of the head
324
+ * index after the CQEs have been read.
325
+ */
326
+ io_uring_smp_store_release(cq->khead, *cq->khead + nr);
327
+ }
328
+ }
329
+
330
+ /*
331
+ * Must be called after io_uring_{peek,wait}_cqe() after the cqe has
332
+ * been processed by the application.
333
+ */
334
+ IOURINGINLINE void io_uring_cqe_seen(struct io_uring *ring,
335
+ struct io_uring_cqe *cqe)
336
+ {
337
+ if (cqe)
338
+ io_uring_cq_advance(ring, 1);
339
+ }
340
+
341
+ /*
342
+ * Command prep helpers
343
+ */
344
+
345
+ /*
346
+ * Associate pointer @data with the sqe, for later retrieval from the cqe
347
+ * at command completion time with io_uring_cqe_get_data().
348
+ */
349
+ IOURINGINLINE void io_uring_sqe_set_data(struct io_uring_sqe *sqe, void *data)
350
+ {
351
+ sqe->user_data = (unsigned long) data;
352
+ }
353
+
354
+ IOURINGINLINE void *io_uring_cqe_get_data(const struct io_uring_cqe *cqe)
355
+ {
356
+ return (void *) (uintptr_t) cqe->user_data;
357
+ }
358
+
359
+ /*
360
+ * Assign a 64-bit value to this sqe, which can get retrieved at completion
361
+ * time with io_uring_cqe_get_data64. Just like the non-64 variants, except
362
+ * these store a 64-bit type rather than a data pointer.
363
+ */
364
+ IOURINGINLINE void io_uring_sqe_set_data64(struct io_uring_sqe *sqe,
365
+ __u64 data)
366
+ {
367
+ sqe->user_data = data;
368
+ }
369
+
370
+ IOURINGINLINE __u64 io_uring_cqe_get_data64(const struct io_uring_cqe *cqe)
371
+ {
372
+ return cqe->user_data;
373
+ }
374
+
375
+ /*
376
+ * Tell the app the have the 64-bit variants of the get/set userdata
377
+ */
378
+ #define LIBURING_HAVE_DATA64
379
+
380
+ IOURINGINLINE void io_uring_sqe_set_flags(struct io_uring_sqe *sqe,
381
+ unsigned flags)
382
+ {
383
+ sqe->flags = (__u8) flags;
384
+ }
385
+
386
+ IOURINGINLINE void __io_uring_set_target_fixed_file(struct io_uring_sqe *sqe,
387
+ unsigned int file_index)
388
+ {
389
+ /* 0 means no fixed files, indexes should be encoded as "index + 1" */
390
+ sqe->file_index = file_index + 1;
391
+ }
392
+
393
+ IOURINGINLINE void io_uring_initialize_sqe(struct io_uring_sqe *sqe)
394
+ {
395
+ sqe->flags = 0;
396
+ sqe->ioprio = 0;
397
+ sqe->rw_flags = 0;
398
+ sqe->buf_index = 0;
399
+ sqe->personality = 0;
400
+ sqe->file_index = 0;
401
+ sqe->addr3 = 0;
402
+ sqe->__pad2[0] = 0;
403
+ }
404
+
405
+ IOURINGINLINE void io_uring_prep_rw(int op, struct io_uring_sqe *sqe, int fd,
406
+ const void *addr, unsigned len,
407
+ __u64 offset)
408
+ {
409
+ sqe->opcode = (__u8) op;
410
+ sqe->fd = fd;
411
+ sqe->off = offset;
412
+ sqe->addr = (unsigned long) addr;
413
+ sqe->len = len;
414
+ }
415
+
416
+ /*
417
+ * io_uring_prep_splice() - Either @fd_in or @fd_out must be a pipe.
418
+ *
419
+ * - If @fd_in refers to a pipe, @off_in is ignored and must be set to -1.
420
+ *
421
+ * - If @fd_in does not refer to a pipe and @off_in is -1, then @nbytes are read
422
+ * from @fd_in starting from the file offset, which is incremented by the
423
+ * number of bytes read.
424
+ *
425
+ * - If @fd_in does not refer to a pipe and @off_in is not -1, then the starting
426
+ * offset of @fd_in will be @off_in.
427
+ *
428
+ * This splice operation can be used to implement sendfile by splicing to an
429
+ * intermediate pipe first, then splice to the final destination.
430
+ * In fact, the implementation of sendfile in kernel uses splice internally.
431
+ *
432
+ * NOTE that even if fd_in or fd_out refers to a pipe, the splice operation
433
+ * can still fail with EINVAL if one of the fd doesn't explicitly support splice
434
+ * operation, e.g. reading from terminal is unsupported from kernel 5.7 to 5.11.
435
+ * Check issue #291 for more information.
436
+ */
437
+ IOURINGINLINE void io_uring_prep_splice(struct io_uring_sqe *sqe,
438
+ int fd_in, int64_t off_in,
439
+ int fd_out, int64_t off_out,
440
+ unsigned int nbytes,
441
+ unsigned int splice_flags)
442
+ {
443
+ io_uring_prep_rw(IORING_OP_SPLICE, sqe, fd_out, NULL, nbytes,
444
+ (__u64) off_out);
445
+ sqe->splice_off_in = (__u64) off_in;
446
+ sqe->splice_fd_in = fd_in;
447
+ sqe->splice_flags = splice_flags;
448
+ }
449
+
450
+ IOURINGINLINE void io_uring_prep_tee(struct io_uring_sqe *sqe,
451
+ int fd_in, int fd_out,
452
+ unsigned int nbytes,
453
+ unsigned int splice_flags)
454
+ {
455
+ io_uring_prep_rw(IORING_OP_TEE, sqe, fd_out, NULL, nbytes, 0);
456
+ sqe->splice_off_in = 0;
457
+ sqe->splice_fd_in = fd_in;
458
+ sqe->splice_flags = splice_flags;
459
+ }
460
+
461
+ IOURINGINLINE void io_uring_prep_readv(struct io_uring_sqe *sqe, int fd,
462
+ const struct iovec *iovecs,
463
+ unsigned nr_vecs, __u64 offset)
464
+ {
465
+ io_uring_prep_rw(IORING_OP_READV, sqe, fd, iovecs, nr_vecs, offset);
466
+ }
467
+
468
+ IOURINGINLINE void io_uring_prep_readv2(struct io_uring_sqe *sqe, int fd,
469
+ const struct iovec *iovecs,
470
+ unsigned nr_vecs, __u64 offset,
471
+ int flags)
472
+ {
473
+ io_uring_prep_readv(sqe, fd, iovecs, nr_vecs, offset);
474
+ sqe->rw_flags = flags;
475
+ }
476
+
477
+ IOURINGINLINE void io_uring_prep_read_fixed(struct io_uring_sqe *sqe, int fd,
478
+ void *buf, unsigned nbytes,
479
+ __u64 offset, int buf_index)
480
+ {
481
+ io_uring_prep_rw(IORING_OP_READ_FIXED, sqe, fd, buf, nbytes, offset);
482
+ sqe->buf_index = (__u16) buf_index;
483
+ }
484
+
485
+ IOURINGINLINE void io_uring_prep_writev(struct io_uring_sqe *sqe, int fd,
486
+ const struct iovec *iovecs,
487
+ unsigned nr_vecs, __u64 offset)
488
+ {
489
+ io_uring_prep_rw(IORING_OP_WRITEV, sqe, fd, iovecs, nr_vecs, offset);
490
+ }
491
+
492
+ IOURINGINLINE void io_uring_prep_writev2(struct io_uring_sqe *sqe, int fd,
493
+ const struct iovec *iovecs,
494
+ unsigned nr_vecs, __u64 offset,
495
+ int flags)
496
+ {
497
+ io_uring_prep_writev(sqe, fd, iovecs, nr_vecs, offset);
498
+ sqe->rw_flags = flags;
499
+ }
500
+
501
+ IOURINGINLINE void io_uring_prep_write_fixed(struct io_uring_sqe *sqe, int fd,
502
+ const void *buf, unsigned nbytes,
503
+ __u64 offset, int buf_index)
504
+ {
505
+ io_uring_prep_rw(IORING_OP_WRITE_FIXED, sqe, fd, buf, nbytes, offset);
506
+ sqe->buf_index = (__u16) buf_index;
507
+ }
508
+
509
+ IOURINGINLINE void io_uring_prep_recvmsg(struct io_uring_sqe *sqe, int fd,
510
+ struct msghdr *msg, unsigned flags)
511
+ {
512
+ io_uring_prep_rw(IORING_OP_RECVMSG, sqe, fd, msg, 1, 0);
513
+ sqe->msg_flags = flags;
514
+ }
515
+
516
+ IOURINGINLINE void io_uring_prep_recvmsg_multishot(struct io_uring_sqe *sqe,
517
+ int fd, struct msghdr *msg,
518
+ unsigned flags)
519
+ {
520
+ io_uring_prep_recvmsg(sqe, fd, msg, flags);
521
+ sqe->ioprio |= IORING_RECV_MULTISHOT;
522
+ }
523
+
524
+ IOURINGINLINE void io_uring_prep_sendmsg(struct io_uring_sqe *sqe, int fd,
525
+ const struct msghdr *msg,
526
+ unsigned flags)
527
+ {
528
+ io_uring_prep_rw(IORING_OP_SENDMSG, sqe, fd, msg, 1, 0);
529
+ sqe->msg_flags = flags;
530
+ }
531
+
532
+ IOURINGINLINE unsigned __io_uring_prep_poll_mask(unsigned poll_mask)
533
+ {
534
+ #if __BYTE_ORDER == __BIG_ENDIAN
535
+ poll_mask = __swahw32(poll_mask);
536
+ #endif
537
+ return poll_mask;
538
+ }
539
+
540
+ IOURINGINLINE void io_uring_prep_poll_add(struct io_uring_sqe *sqe, int fd,
541
+ unsigned poll_mask)
542
+ {
543
+ io_uring_prep_rw(IORING_OP_POLL_ADD, sqe, fd, NULL, 0, 0);
544
+ sqe->poll32_events = __io_uring_prep_poll_mask(poll_mask);
545
+ }
546
+
547
+ IOURINGINLINE void io_uring_prep_poll_multishot(struct io_uring_sqe *sqe,
548
+ int fd, unsigned poll_mask)
549
+ {
550
+ io_uring_prep_poll_add(sqe, fd, poll_mask);
551
+ sqe->len = IORING_POLL_ADD_MULTI;
552
+ }
553
+
554
+ IOURINGINLINE void io_uring_prep_poll_remove(struct io_uring_sqe *sqe,
555
+ __u64 user_data)
556
+ {
557
+ io_uring_prep_rw(IORING_OP_POLL_REMOVE, sqe, -1, NULL, 0, 0);
558
+ sqe->addr = user_data;
559
+ }
560
+
561
+ IOURINGINLINE void io_uring_prep_poll_update(struct io_uring_sqe *sqe,
562
+ __u64 old_user_data,
563
+ __u64 new_user_data,
564
+ unsigned poll_mask, unsigned flags)
565
+ {
566
+ io_uring_prep_rw(IORING_OP_POLL_REMOVE, sqe, -1, NULL, flags,
567
+ new_user_data);
568
+ sqe->addr = old_user_data;
569
+ sqe->poll32_events = __io_uring_prep_poll_mask(poll_mask);
570
+ }
571
+
572
+ IOURINGINLINE void io_uring_prep_fsync(struct io_uring_sqe *sqe, int fd,
573
+ unsigned fsync_flags)
574
+ {
575
+ io_uring_prep_rw(IORING_OP_FSYNC, sqe, fd, NULL, 0, 0);
576
+ sqe->fsync_flags = fsync_flags;
577
+ }
578
+
579
+ IOURINGINLINE void io_uring_prep_nop(struct io_uring_sqe *sqe)
580
+ {
581
+ io_uring_prep_rw(IORING_OP_NOP, sqe, -1, NULL, 0, 0);
582
+ }
583
+
584
+ IOURINGINLINE void io_uring_prep_timeout(struct io_uring_sqe *sqe,
585
+ struct __kernel_timespec *ts,
586
+ unsigned count, unsigned flags)
587
+ {
588
+ io_uring_prep_rw(IORING_OP_TIMEOUT, sqe, -1, ts, 1, count);
589
+ sqe->timeout_flags = flags;
590
+ }
591
+
592
+ IOURINGINLINE void io_uring_prep_timeout_remove(struct io_uring_sqe *sqe,
593
+ __u64 user_data, unsigned flags)
594
+ {
595
+ io_uring_prep_rw(IORING_OP_TIMEOUT_REMOVE, sqe, -1, NULL, 0, 0);
596
+ sqe->addr = user_data;
597
+ sqe->timeout_flags = flags;
598
+ }
599
+
600
+ IOURINGINLINE void io_uring_prep_timeout_update(struct io_uring_sqe *sqe,
601
+ struct __kernel_timespec *ts,
602
+ __u64 user_data, unsigned flags)
603
+ {
604
+ io_uring_prep_rw(IORING_OP_TIMEOUT_REMOVE, sqe, -1, NULL, 0,
605
+ (uintptr_t) ts);
606
+ sqe->addr = user_data;
607
+ sqe->timeout_flags = flags | IORING_TIMEOUT_UPDATE;
608
+ }
609
+
610
+ IOURINGINLINE void io_uring_prep_accept(struct io_uring_sqe *sqe, int fd,
611
+ struct sockaddr *addr,
612
+ socklen_t *addrlen, int flags)
613
+ {
614
+ io_uring_prep_rw(IORING_OP_ACCEPT, sqe, fd, addr, 0,
615
+ (__u64) (unsigned long) addrlen);
616
+ sqe->accept_flags = (__u32) flags;
617
+ }
618
+
619
+ /* accept directly into the fixed file table */
620
+ IOURINGINLINE void io_uring_prep_accept_direct(struct io_uring_sqe *sqe, int fd,
621
+ struct sockaddr *addr,
622
+ socklen_t *addrlen, int flags,
623
+ unsigned int file_index)
624
+ {
625
+ io_uring_prep_accept(sqe, fd, addr, addrlen, flags);
626
+ /* offset by 1 for allocation */
627
+ if (file_index == IORING_FILE_INDEX_ALLOC)
628
+ file_index--;
629
+ __io_uring_set_target_fixed_file(sqe, file_index);
630
+ }
631
+
632
+ IOURINGINLINE void io_uring_prep_multishot_accept(struct io_uring_sqe *sqe,
633
+ int fd, struct sockaddr *addr,
634
+ socklen_t *addrlen, int flags)
635
+ {
636
+ io_uring_prep_accept(sqe, fd, addr, addrlen, flags);
637
+ sqe->ioprio |= IORING_ACCEPT_MULTISHOT;
638
+ }
639
+
640
+ /* multishot accept directly into the fixed file table */
641
+ IOURINGINLINE void io_uring_prep_multishot_accept_direct(struct io_uring_sqe *sqe,
642
+ int fd,
643
+ struct sockaddr *addr,
644
+ socklen_t *addrlen,
645
+ int flags)
646
+ {
647
+ io_uring_prep_multishot_accept(sqe, fd, addr, addrlen, flags);
648
+ __io_uring_set_target_fixed_file(sqe, IORING_FILE_INDEX_ALLOC - 1);
649
+ }
650
+
651
+ IOURINGINLINE void io_uring_prep_cancel64(struct io_uring_sqe *sqe,
652
+ __u64 user_data, int flags)
653
+ {
654
+ io_uring_prep_rw(IORING_OP_ASYNC_CANCEL, sqe, -1, NULL, 0, 0);
655
+ sqe->addr = user_data;
656
+ sqe->cancel_flags = (__u32) flags;
657
+ }
658
+
659
+ IOURINGINLINE void io_uring_prep_cancel(struct io_uring_sqe *sqe,
660
+ void *user_data, int flags)
661
+ {
662
+ io_uring_prep_cancel64(sqe, (__u64) (uintptr_t) user_data, flags);
663
+ }
664
+
665
+ IOURINGINLINE void io_uring_prep_cancel_fd(struct io_uring_sqe *sqe, int fd,
666
+ unsigned int flags)
667
+ {
668
+ io_uring_prep_rw(IORING_OP_ASYNC_CANCEL, sqe, fd, NULL, 0, 0);
669
+ sqe->cancel_flags = (__u32) flags | IORING_ASYNC_CANCEL_FD;
670
+ }
671
+
672
+ IOURINGINLINE void io_uring_prep_link_timeout(struct io_uring_sqe *sqe,
673
+ struct __kernel_timespec *ts,
674
+ unsigned flags)
675
+ {
676
+ io_uring_prep_rw(IORING_OP_LINK_TIMEOUT, sqe, -1, ts, 1, 0);
677
+ sqe->timeout_flags = flags;
678
+ }
679
+
680
+ IOURINGINLINE void io_uring_prep_connect(struct io_uring_sqe *sqe, int fd,
681
+ const struct sockaddr *addr,
682
+ socklen_t addrlen)
683
+ {
684
+ io_uring_prep_rw(IORING_OP_CONNECT, sqe, fd, addr, 0, addrlen);
685
+ }
686
+
687
+ IOURINGINLINE void io_uring_prep_bind(struct io_uring_sqe *sqe, int fd,
688
+ struct sockaddr *addr,
689
+ socklen_t addrlen)
690
+ {
691
+ io_uring_prep_rw(IORING_OP_BIND, sqe, fd, addr, 0, addrlen);
692
+ }
693
+
694
+ IOURINGINLINE void io_uring_prep_listen(struct io_uring_sqe *sqe, int fd,
695
+ int backlog)
696
+ {
697
+ io_uring_prep_rw(IORING_OP_LISTEN, sqe, fd, 0, backlog, 0);
698
+ }
699
+
700
+ IOURINGINLINE void io_uring_prep_files_update(struct io_uring_sqe *sqe,
701
+ int *fds, unsigned nr_fds,
702
+ int offset)
703
+ {
704
+ io_uring_prep_rw(IORING_OP_FILES_UPDATE, sqe, -1, fds, nr_fds,
705
+ (__u64) offset);
706
+ }
707
+
708
+ IOURINGINLINE void io_uring_prep_fallocate(struct io_uring_sqe *sqe, int fd,
709
+ int mode, __u64 offset, __u64 len)
710
+ {
711
+ io_uring_prep_rw(IORING_OP_FALLOCATE, sqe, fd,
712
+ 0, (unsigned int) mode, (__u64) offset);
713
+ sqe->addr = (__u64) len;
714
+ }
715
+
716
+ IOURINGINLINE void io_uring_prep_openat(struct io_uring_sqe *sqe, int dfd,
717
+ const char *path, int flags,
718
+ mode_t mode)
719
+ {
720
+ io_uring_prep_rw(IORING_OP_OPENAT, sqe, dfd, path, mode, 0);
721
+ sqe->open_flags = (__u32) flags;
722
+ }
723
+
724
+ /* open directly into the fixed file table */
725
+ IOURINGINLINE void io_uring_prep_openat_direct(struct io_uring_sqe *sqe,
726
+ int dfd, const char *path,
727
+ int flags, mode_t mode,
728
+ unsigned file_index)
729
+ {
730
+ io_uring_prep_openat(sqe, dfd, path, flags, mode);
731
+ /* offset by 1 for allocation */
732
+ if (file_index == IORING_FILE_INDEX_ALLOC)
733
+ file_index--;
734
+ __io_uring_set_target_fixed_file(sqe, file_index);
735
+ }
736
+
737
+ IOURINGINLINE void io_uring_prep_close(struct io_uring_sqe *sqe, int fd)
738
+ {
739
+ io_uring_prep_rw(IORING_OP_CLOSE, sqe, fd, NULL, 0, 0);
740
+ }
741
+
742
+ IOURINGINLINE void io_uring_prep_close_direct(struct io_uring_sqe *sqe,
743
+ unsigned file_index)
744
+ {
745
+ io_uring_prep_close(sqe, 0);
746
+ __io_uring_set_target_fixed_file(sqe, file_index);
747
+ }
748
+
749
+ IOURINGINLINE void io_uring_prep_read(struct io_uring_sqe *sqe, int fd,
750
+ void *buf, unsigned nbytes, __u64 offset)
751
+ {
752
+ io_uring_prep_rw(IORING_OP_READ, sqe, fd, buf, nbytes, offset);
753
+ }
754
+
755
+ IOURINGINLINE void io_uring_prep_read_multishot(struct io_uring_sqe *sqe,
756
+ int fd, unsigned nbytes,
757
+ __u64 offset, int buf_group)
758
+ {
759
+ io_uring_prep_rw(IORING_OP_READ_MULTISHOT, sqe, fd, NULL, nbytes,
760
+ offset);
761
+ sqe->buf_group = buf_group;
762
+ sqe->flags = IOSQE_BUFFER_SELECT;
763
+ }
764
+
765
+ IOURINGINLINE void io_uring_prep_write(struct io_uring_sqe *sqe, int fd,
766
+ const void *buf, unsigned nbytes,
767
+ __u64 offset)
768
+ {
769
+ io_uring_prep_rw(IORING_OP_WRITE, sqe, fd, buf, nbytes, offset);
770
+ }
771
+
772
+ struct statx;
773
+ IOURINGINLINE void io_uring_prep_statx(struct io_uring_sqe *sqe, int dfd,
774
+ const char *path, int flags,
775
+ unsigned mask, struct statx *statxbuf)
776
+ {
777
+ io_uring_prep_rw(IORING_OP_STATX, sqe, dfd, path, mask,
778
+ (__u64) (unsigned long) statxbuf);
779
+ sqe->statx_flags = (__u32) flags;
780
+ }
781
+
782
+ IOURINGINLINE void io_uring_prep_fadvise(struct io_uring_sqe *sqe, int fd,
783
+ __u64 offset, __u32 len, int advice)
784
+ {
785
+ io_uring_prep_rw(IORING_OP_FADVISE, sqe, fd, NULL, (__u32) len, offset);
786
+ sqe->fadvise_advice = (__u32) advice;
787
+ }
788
+
789
+ IOURINGINLINE void io_uring_prep_madvise(struct io_uring_sqe *sqe, void *addr,
790
+ __u32 length, int advice)
791
+ {
792
+ io_uring_prep_rw(IORING_OP_MADVISE, sqe, -1, addr, (__u32) length, 0);
793
+ sqe->fadvise_advice = (__u32) advice;
794
+ }
795
+
796
+ IOURINGINLINE void io_uring_prep_fadvise64(struct io_uring_sqe *sqe, int fd,
797
+ __u64 offset, off_t len, int advice)
798
+ {
799
+ io_uring_prep_rw(IORING_OP_FADVISE, sqe, fd, NULL, 0, offset);
800
+ sqe->addr = len;
801
+ sqe->fadvise_advice = (__u32) advice;
802
+ }
803
+
804
+ IOURINGINLINE void io_uring_prep_madvise64(struct io_uring_sqe *sqe, void *addr,
805
+ off_t length, int advice)
806
+ {
807
+ io_uring_prep_rw(IORING_OP_MADVISE, sqe, -1, addr, 0, length);
808
+ sqe->fadvise_advice = (__u32) advice;
809
+ }
810
+
811
+ IOURINGINLINE void io_uring_prep_send(struct io_uring_sqe *sqe, int sockfd,
812
+ const void *buf, size_t len, int flags)
813
+ {
814
+ io_uring_prep_rw(IORING_OP_SEND, sqe, sockfd, buf, (__u32) len, 0);
815
+ sqe->msg_flags = (__u32) flags;
816
+ }
817
+
818
+ IOURINGINLINE void io_uring_prep_send_bundle(struct io_uring_sqe *sqe,
819
+ int sockfd, size_t len, int flags)
820
+ {
821
+ io_uring_prep_send(sqe, sockfd, NULL, len, flags);
822
+ sqe->ioprio |= IORING_RECVSEND_BUNDLE;
823
+ }
824
+
825
+ IOURINGINLINE void io_uring_prep_send_set_addr(struct io_uring_sqe *sqe,
826
+ const struct sockaddr *dest_addr,
827
+ __u16 addr_len)
828
+ {
829
+ sqe->addr2 = (unsigned long)(const void *)dest_addr;
830
+ sqe->addr_len = addr_len;
831
+ }
832
+
833
+ IOURINGINLINE void io_uring_prep_sendto(struct io_uring_sqe *sqe, int sockfd,
834
+ const void *buf, size_t len, int flags,
835
+ const struct sockaddr *addr,
836
+ socklen_t addrlen)
837
+ {
838
+ io_uring_prep_send(sqe, sockfd, buf, len, flags);
839
+ io_uring_prep_send_set_addr(sqe, addr, addrlen);
840
+ }
841
+
842
+ IOURINGINLINE void io_uring_prep_send_zc(struct io_uring_sqe *sqe, int sockfd,
843
+ const void *buf, size_t len, int flags,
844
+ unsigned zc_flags)
845
+ {
846
+ io_uring_prep_rw(IORING_OP_SEND_ZC, sqe, sockfd, buf, (__u32) len, 0);
847
+ sqe->msg_flags = (__u32) flags;
848
+ sqe->ioprio = zc_flags;
849
+ }
850
+
851
+ IOURINGINLINE void io_uring_prep_send_zc_fixed(struct io_uring_sqe *sqe,
852
+ int sockfd, const void *buf,
853
+ size_t len, int flags,
854
+ unsigned zc_flags,
855
+ unsigned buf_index)
856
+ {
857
+ io_uring_prep_send_zc(sqe, sockfd, buf, len, flags, zc_flags);
858
+ sqe->ioprio |= IORING_RECVSEND_FIXED_BUF;
859
+ sqe->buf_index = buf_index;
860
+ }
861
+
862
+ IOURINGINLINE void io_uring_prep_sendmsg_zc(struct io_uring_sqe *sqe, int fd,
863
+ const struct msghdr *msg,
864
+ unsigned flags)
865
+ {
866
+ io_uring_prep_sendmsg(sqe, fd, msg, flags);
867
+ sqe->opcode = IORING_OP_SENDMSG_ZC;
868
+ }
869
+
870
+ IOURINGINLINE void io_uring_prep_recv(struct io_uring_sqe *sqe, int sockfd,
871
+ void *buf, size_t len, int flags)
872
+ {
873
+ io_uring_prep_rw(IORING_OP_RECV, sqe, sockfd, buf, (__u32) len, 0);
874
+ sqe->msg_flags = (__u32) flags;
875
+ }
876
+
877
+ IOURINGINLINE void io_uring_prep_recv_multishot(struct io_uring_sqe *sqe,
878
+ int sockfd, void *buf,
879
+ size_t len, int flags)
880
+ {
881
+ io_uring_prep_recv(sqe, sockfd, buf, len, flags);
882
+ sqe->ioprio |= IORING_RECV_MULTISHOT;
883
+ }
884
+
885
+ IOURINGINLINE struct io_uring_recvmsg_out *
886
+ io_uring_recvmsg_validate(void *buf, int buf_len, struct msghdr *msgh)
887
+ {
888
+ unsigned long header = msgh->msg_controllen + msgh->msg_namelen +
889
+ sizeof(struct io_uring_recvmsg_out);
890
+ if (buf_len < 0 || (unsigned long)buf_len < header)
891
+ return NULL;
892
+ return (struct io_uring_recvmsg_out *)buf;
893
+ }
894
+
895
+ IOURINGINLINE void *io_uring_recvmsg_name(struct io_uring_recvmsg_out *o)
896
+ {
897
+ return (void *) &o[1];
898
+ }
899
+
900
+ IOURINGINLINE struct cmsghdr *
901
+ io_uring_recvmsg_cmsg_firsthdr(struct io_uring_recvmsg_out *o,
902
+ struct msghdr *msgh)
903
+ {
904
+ if (o->controllen < sizeof(struct cmsghdr))
905
+ return NULL;
906
+
907
+ return (struct cmsghdr *)((unsigned char *) io_uring_recvmsg_name(o) +
908
+ msgh->msg_namelen);
909
+ }
910
+
911
+ IOURINGINLINE struct cmsghdr *
912
+ io_uring_recvmsg_cmsg_nexthdr(struct io_uring_recvmsg_out *o, struct msghdr *msgh,
913
+ struct cmsghdr *cmsg)
914
+ {
915
+ unsigned char *end;
916
+
917
+ if (cmsg->cmsg_len < sizeof(struct cmsghdr))
918
+ return NULL;
919
+ end = (unsigned char *) io_uring_recvmsg_cmsg_firsthdr(o, msgh) +
920
+ o->controllen;
921
+ cmsg = (struct cmsghdr *)((unsigned char *) cmsg +
922
+ CMSG_ALIGN(cmsg->cmsg_len));
923
+
924
+ if ((unsigned char *) (cmsg + 1) > end)
925
+ return NULL;
926
+ if (((unsigned char *) cmsg) + CMSG_ALIGN(cmsg->cmsg_len) > end)
927
+ return NULL;
928
+
929
+ return cmsg;
930
+ }
931
+
932
+ IOURINGINLINE void *io_uring_recvmsg_payload(struct io_uring_recvmsg_out *o,
933
+ struct msghdr *msgh)
934
+ {
935
+ return (void *)((unsigned char *)io_uring_recvmsg_name(o) +
936
+ msgh->msg_namelen + msgh->msg_controllen);
937
+ }
938
+
939
+ IOURINGINLINE unsigned int
940
+ io_uring_recvmsg_payload_length(struct io_uring_recvmsg_out *o,
941
+ int buf_len, struct msghdr *msgh)
942
+ {
943
+ unsigned long payload_start, payload_end;
944
+
945
+ payload_start = (unsigned long) io_uring_recvmsg_payload(o, msgh);
946
+ payload_end = (unsigned long) o + buf_len;
947
+ return (unsigned int) (payload_end - payload_start);
948
+ }
949
+
950
+ IOURINGINLINE void io_uring_prep_openat2(struct io_uring_sqe *sqe, int dfd,
951
+ const char *path, struct open_how *how)
952
+ {
953
+ io_uring_prep_rw(IORING_OP_OPENAT2, sqe, dfd, path, sizeof(*how),
954
+ (uint64_t) (uintptr_t) how);
955
+ }
956
+
957
+ /* open directly into the fixed file table */
958
+ IOURINGINLINE void io_uring_prep_openat2_direct(struct io_uring_sqe *sqe,
959
+ int dfd, const char *path,
960
+ struct open_how *how,
961
+ unsigned file_index)
962
+ {
963
+ io_uring_prep_openat2(sqe, dfd, path, how);
964
+ /* offset by 1 for allocation */
965
+ if (file_index == IORING_FILE_INDEX_ALLOC)
966
+ file_index--;
967
+ __io_uring_set_target_fixed_file(sqe, file_index);
968
+ }
969
+
970
+ struct epoll_event;
971
+ IOURINGINLINE void io_uring_prep_epoll_ctl(struct io_uring_sqe *sqe, int epfd,
972
+ int fd, int op,
973
+ struct epoll_event *ev)
974
+ {
975
+ io_uring_prep_rw(IORING_OP_EPOLL_CTL, sqe, epfd, ev,
976
+ (__u32) op, (__u32) fd);
977
+ }
978
+
979
+ IOURINGINLINE void io_uring_prep_provide_buffers(struct io_uring_sqe *sqe,
980
+ void *addr, int len, int nr,
981
+ int bgid, int bid)
982
+ {
983
+ io_uring_prep_rw(IORING_OP_PROVIDE_BUFFERS, sqe, nr, addr, (__u32) len,
984
+ (__u64) bid);
985
+ sqe->buf_group = (__u16) bgid;
986
+ }
987
+
988
+ IOURINGINLINE void io_uring_prep_remove_buffers(struct io_uring_sqe *sqe,
989
+ int nr, int bgid)
990
+ {
991
+ io_uring_prep_rw(IORING_OP_REMOVE_BUFFERS, sqe, nr, NULL, 0, 0);
992
+ sqe->buf_group = (__u16) bgid;
993
+ }
994
+
995
+ IOURINGINLINE void io_uring_prep_shutdown(struct io_uring_sqe *sqe, int fd,
996
+ int how)
997
+ {
998
+ io_uring_prep_rw(IORING_OP_SHUTDOWN, sqe, fd, NULL, (__u32) how, 0);
999
+ }
1000
+
1001
+ IOURINGINLINE void io_uring_prep_unlinkat(struct io_uring_sqe *sqe, int dfd,
1002
+ const char *path, int flags)
1003
+ {
1004
+ io_uring_prep_rw(IORING_OP_UNLINKAT, sqe, dfd, path, 0, 0);
1005
+ sqe->unlink_flags = (__u32) flags;
1006
+ }
1007
+
1008
+ IOURINGINLINE void io_uring_prep_unlink(struct io_uring_sqe *sqe,
1009
+ const char *path, int flags)
1010
+ {
1011
+ io_uring_prep_unlinkat(sqe, AT_FDCWD, path, flags);
1012
+ }
1013
+
1014
+ IOURINGINLINE void io_uring_prep_renameat(struct io_uring_sqe *sqe, int olddfd,
1015
+ const char *oldpath, int newdfd,
1016
+ const char *newpath, unsigned int flags)
1017
+ {
1018
+ io_uring_prep_rw(IORING_OP_RENAMEAT, sqe, olddfd, oldpath,
1019
+ (__u32) newdfd,
1020
+ (uint64_t) (uintptr_t) newpath);
1021
+ sqe->rename_flags = (__u32) flags;
1022
+ }
1023
+
1024
+ IOURINGINLINE void io_uring_prep_rename(struct io_uring_sqe *sqe,
1025
+ const char *oldpath,
1026
+ const char *newpath)
1027
+ {
1028
+ io_uring_prep_renameat(sqe, AT_FDCWD, oldpath, AT_FDCWD, newpath, 0);
1029
+ }
1030
+
1031
+ IOURINGINLINE void io_uring_prep_sync_file_range(struct io_uring_sqe *sqe,
1032
+ int fd, unsigned len,
1033
+ __u64 offset, int flags)
1034
+ {
1035
+ io_uring_prep_rw(IORING_OP_SYNC_FILE_RANGE, sqe, fd, NULL, len, offset);
1036
+ sqe->sync_range_flags = (__u32) flags;
1037
+ }
1038
+
1039
+ IOURINGINLINE void io_uring_prep_mkdirat(struct io_uring_sqe *sqe, int dfd,
1040
+ const char *path, mode_t mode)
1041
+ {
1042
+ io_uring_prep_rw(IORING_OP_MKDIRAT, sqe, dfd, path, mode, 0);
1043
+ }
1044
+
1045
+ IOURINGINLINE void io_uring_prep_mkdir(struct io_uring_sqe *sqe,
1046
+ const char *path, mode_t mode)
1047
+ {
1048
+ io_uring_prep_mkdirat(sqe, AT_FDCWD, path, mode);
1049
+ }
1050
+
1051
+ IOURINGINLINE void io_uring_prep_symlinkat(struct io_uring_sqe *sqe,
1052
+ const char *target, int newdirfd,
1053
+ const char *linkpath)
1054
+ {
1055
+ io_uring_prep_rw(IORING_OP_SYMLINKAT, sqe, newdirfd, target, 0,
1056
+ (uint64_t) (uintptr_t) linkpath);
1057
+ }
1058
+
1059
+ IOURINGINLINE void io_uring_prep_symlink(struct io_uring_sqe *sqe,
1060
+ const char *target,
1061
+ const char *linkpath)
1062
+ {
1063
+ io_uring_prep_symlinkat(sqe, target, AT_FDCWD, linkpath);
1064
+ }
1065
+
1066
+ IOURINGINLINE void io_uring_prep_linkat(struct io_uring_sqe *sqe, int olddfd,
1067
+ const char *oldpath, int newdfd,
1068
+ const char *newpath, int flags)
1069
+ {
1070
+ io_uring_prep_rw(IORING_OP_LINKAT, sqe, olddfd, oldpath, (__u32) newdfd,
1071
+ (uint64_t) (uintptr_t) newpath);
1072
+ sqe->hardlink_flags = (__u32) flags;
1073
+ }
1074
+
1075
+ IOURINGINLINE void io_uring_prep_link(struct io_uring_sqe *sqe,
1076
+ const char *oldpath, const char *newpath,
1077
+ int flags)
1078
+ {
1079
+ io_uring_prep_linkat(sqe, AT_FDCWD, oldpath, AT_FDCWD, newpath, flags);
1080
+ }
1081
+
1082
+ IOURINGINLINE void io_uring_prep_msg_ring_cqe_flags(struct io_uring_sqe *sqe,
1083
+ int fd, unsigned int len, __u64 data,
1084
+ unsigned int flags, unsigned int cqe_flags)
1085
+ {
1086
+ io_uring_prep_rw(IORING_OP_MSG_RING, sqe, fd, NULL, len, data);
1087
+ sqe->msg_ring_flags = IORING_MSG_RING_FLAGS_PASS | flags;
1088
+ sqe->file_index = cqe_flags;
1089
+ }
1090
+
1091
+ IOURINGINLINE void io_uring_prep_msg_ring(struct io_uring_sqe *sqe, int fd,
1092
+ unsigned int len, __u64 data,
1093
+ unsigned int flags)
1094
+ {
1095
+ io_uring_prep_rw(IORING_OP_MSG_RING, sqe, fd, NULL, len, data);
1096
+ sqe->msg_ring_flags = flags;
1097
+ }
1098
+
1099
+ IOURINGINLINE void io_uring_prep_msg_ring_fd(struct io_uring_sqe *sqe, int fd,
1100
+ int source_fd, int target_fd,
1101
+ __u64 data, unsigned int flags)
1102
+ {
1103
+ io_uring_prep_rw(IORING_OP_MSG_RING, sqe, fd,
1104
+ (void *) (uintptr_t) IORING_MSG_SEND_FD, 0, data);
1105
+ sqe->addr3 = source_fd;
1106
+ /* offset by 1 for allocation */
1107
+ if ((unsigned int) target_fd == IORING_FILE_INDEX_ALLOC)
1108
+ target_fd--;
1109
+ __io_uring_set_target_fixed_file(sqe, target_fd);
1110
+ sqe->msg_ring_flags = flags;
1111
+ }
1112
+
1113
+ IOURINGINLINE void io_uring_prep_msg_ring_fd_alloc(struct io_uring_sqe *sqe,
1114
+ int fd, int source_fd,
1115
+ __u64 data, unsigned int flags)
1116
+ {
1117
+ io_uring_prep_msg_ring_fd(sqe, fd, source_fd, IORING_FILE_INDEX_ALLOC,
1118
+ data, flags);
1119
+ }
1120
+
1121
+ IOURINGINLINE void io_uring_prep_getxattr(struct io_uring_sqe *sqe,
1122
+ const char *name, char *value,
1123
+ const char *path, unsigned int len)
1124
+ {
1125
+ io_uring_prep_rw(IORING_OP_GETXATTR, sqe, 0, name, len,
1126
+ (__u64) (uintptr_t) value);
1127
+ sqe->addr3 = (__u64) (uintptr_t) path;
1128
+ sqe->xattr_flags = 0;
1129
+ }
1130
+
1131
+ IOURINGINLINE void io_uring_prep_setxattr(struct io_uring_sqe *sqe,
1132
+ const char *name, const char *value,
1133
+ const char *path, int flags,
1134
+ unsigned int len)
1135
+ {
1136
+ io_uring_prep_rw(IORING_OP_SETXATTR, sqe, 0, name, len,
1137
+ (__u64) (uintptr_t) value);
1138
+ sqe->addr3 = (__u64) (uintptr_t) path;
1139
+ sqe->xattr_flags = flags;
1140
+ }
1141
+
1142
+ IOURINGINLINE void io_uring_prep_fgetxattr(struct io_uring_sqe *sqe,
1143
+ int fd, const char *name,
1144
+ char *value, unsigned int len)
1145
+ {
1146
+ io_uring_prep_rw(IORING_OP_FGETXATTR, sqe, fd, name, len,
1147
+ (__u64) (uintptr_t) value);
1148
+ sqe->xattr_flags = 0;
1149
+ }
1150
+
1151
+ IOURINGINLINE void io_uring_prep_fsetxattr(struct io_uring_sqe *sqe, int fd,
1152
+ const char *name, const char *value,
1153
+ int flags, unsigned int len)
1154
+ {
1155
+ io_uring_prep_rw(IORING_OP_FSETXATTR, sqe, fd, name, len,
1156
+ (__u64) (uintptr_t) value);
1157
+ sqe->xattr_flags = flags;
1158
+ }
1159
+
1160
+ IOURINGINLINE void io_uring_prep_socket(struct io_uring_sqe *sqe, int domain,
1161
+ int type, int protocol,
1162
+ unsigned int flags)
1163
+ {
1164
+ io_uring_prep_rw(IORING_OP_SOCKET, sqe, domain, NULL, protocol, type);
1165
+ sqe->rw_flags = flags;
1166
+ }
1167
+
1168
+ IOURINGINLINE void io_uring_prep_socket_direct(struct io_uring_sqe *sqe,
1169
+ int domain, int type,
1170
+ int protocol,
1171
+ unsigned file_index,
1172
+ unsigned int flags)
1173
+ {
1174
+ io_uring_prep_rw(IORING_OP_SOCKET, sqe, domain, NULL, protocol, type);
1175
+ sqe->rw_flags = flags;
1176
+ /* offset by 1 for allocation */
1177
+ if (file_index == IORING_FILE_INDEX_ALLOC)
1178
+ file_index--;
1179
+ __io_uring_set_target_fixed_file(sqe, file_index);
1180
+ }
1181
+
1182
+ IOURINGINLINE void io_uring_prep_socket_direct_alloc(struct io_uring_sqe *sqe,
1183
+ int domain, int type,
1184
+ int protocol,
1185
+ unsigned int flags)
1186
+ {
1187
+ io_uring_prep_rw(IORING_OP_SOCKET, sqe, domain, NULL, protocol, type);
1188
+ sqe->rw_flags = flags;
1189
+ __io_uring_set_target_fixed_file(sqe, IORING_FILE_INDEX_ALLOC - 1);
1190
+ }
1191
+
1192
+ /*
1193
+ * Prepare commands for sockets
1194
+ */
1195
+ IOURINGINLINE void io_uring_prep_cmd_sock(struct io_uring_sqe *sqe,
1196
+ int cmd_op,
1197
+ int fd,
1198
+ int level,
1199
+ int optname,
1200
+ void *optval,
1201
+ int optlen)
1202
+ {
1203
+ io_uring_prep_rw(IORING_OP_URING_CMD, sqe, fd, NULL, 0, 0);
1204
+ sqe->optval = (unsigned long) (uintptr_t) optval;
1205
+ sqe->optname = optname;
1206
+ sqe->optlen = optlen;
1207
+ sqe->cmd_op = cmd_op;
1208
+ sqe->level = level;
1209
+ }
1210
+
1211
+ IOURINGINLINE void io_uring_prep_waitid(struct io_uring_sqe *sqe,
1212
+ idtype_t idtype,
1213
+ id_t id,
1214
+ siginfo_t *infop,
1215
+ int options, unsigned int flags)
1216
+ {
1217
+ io_uring_prep_rw(IORING_OP_WAITID, sqe, id, NULL, (unsigned) idtype, 0);
1218
+ sqe->waitid_flags = flags;
1219
+ sqe->file_index = options;
1220
+ sqe->addr2 = (unsigned long) infop;
1221
+ }
1222
+
1223
+ IOURINGINLINE void io_uring_prep_futex_wake(struct io_uring_sqe *sqe,
1224
+ uint32_t *futex, uint64_t val,
1225
+ uint64_t mask, uint32_t futex_flags,
1226
+ unsigned int flags)
1227
+ {
1228
+ io_uring_prep_rw(IORING_OP_FUTEX_WAKE, sqe, futex_flags, futex, 0, val);
1229
+ sqe->futex_flags = flags;
1230
+ sqe->addr3 = mask;
1231
+ }
1232
+
1233
+ IOURINGINLINE void io_uring_prep_futex_wait(struct io_uring_sqe *sqe,
1234
+ uint32_t *futex, uint64_t val,
1235
+ uint64_t mask, uint32_t futex_flags,
1236
+ unsigned int flags)
1237
+ {
1238
+ io_uring_prep_rw(IORING_OP_FUTEX_WAIT, sqe, futex_flags, futex, 0, val);
1239
+ sqe->futex_flags = flags;
1240
+ sqe->addr3 = mask;
1241
+ }
1242
+
1243
+ struct futex_waitv;
1244
+ IOURINGINLINE void io_uring_prep_futex_waitv(struct io_uring_sqe *sqe,
1245
+ struct futex_waitv *futex,
1246
+ uint32_t nr_futex,
1247
+ unsigned int flags)
1248
+ {
1249
+ io_uring_prep_rw(IORING_OP_FUTEX_WAITV, sqe, 0, futex, nr_futex, 0);
1250
+ sqe->futex_flags = flags;
1251
+ }
1252
+
1253
+ IOURINGINLINE void io_uring_prep_fixed_fd_install(struct io_uring_sqe *sqe,
1254
+ int fd,
1255
+ unsigned int flags)
1256
+ {
1257
+ io_uring_prep_rw(IORING_OP_FIXED_FD_INSTALL, sqe, fd, NULL, 0, 0);
1258
+ sqe->flags = IOSQE_FIXED_FILE;
1259
+ sqe->install_fd_flags = flags;
1260
+ }
1261
+
1262
+ IOURINGINLINE void io_uring_prep_ftruncate(struct io_uring_sqe *sqe,
1263
+ int fd, loff_t len)
1264
+ {
1265
+ io_uring_prep_rw(IORING_OP_FTRUNCATE, sqe, fd, 0, 0, len);
1266
+ }
1267
+
1268
+ /*
1269
+ * Returns number of unconsumed (if SQPOLL) or unsubmitted entries exist in
1270
+ * the SQ ring
1271
+ */
1272
+ IOURINGINLINE unsigned io_uring_sq_ready(const struct io_uring *ring)
1273
+ {
1274
+ unsigned khead;
1275
+
1276
+ /*
1277
+ * Without a barrier, we could miss an update and think the SQ wasn't
1278
+ * ready. We don't need the load acquire for non-SQPOLL since then we
1279
+ * drive updates.
1280
+ */
1281
+ if (ring->flags & IORING_SETUP_SQPOLL)
1282
+ khead = io_uring_smp_load_acquire(ring->sq.khead);
1283
+ else
1284
+ khead = *ring->sq.khead;
1285
+
1286
+ /* always use real head, to avoid losing sync for short submit */
1287
+ return ring->sq.sqe_tail - khead;
1288
+ }
1289
+
1290
+ /*
1291
+ * Returns how much space is left in the SQ ring.
1292
+ */
1293
+ IOURINGINLINE unsigned io_uring_sq_space_left(const struct io_uring *ring)
1294
+ {
1295
+ return ring->sq.ring_entries - io_uring_sq_ready(ring);
1296
+ }
1297
+
1298
+ /*
1299
+ * Only applicable when using SQPOLL - allows the caller to wait for space
1300
+ * to free up in the SQ ring, which happens when the kernel side thread has
1301
+ * consumed one or more entries. If the SQ ring is currently non-full, no
1302
+ * action is taken. Note: may return -EINVAL if the kernel doesn't support
1303
+ * this feature.
1304
+ */
1305
+ IOURINGINLINE int io_uring_sqring_wait(struct io_uring *ring)
1306
+ {
1307
+ if (!(ring->flags & IORING_SETUP_SQPOLL))
1308
+ return 0;
1309
+ if (io_uring_sq_space_left(ring))
1310
+ return 0;
1311
+
1312
+ return __io_uring_sqring_wait(ring);
1313
+ }
1314
+
1315
+ /*
1316
+ * Returns how many unconsumed entries are ready in the CQ ring
1317
+ */
1318
+ IOURINGINLINE unsigned io_uring_cq_ready(const struct io_uring *ring)
1319
+ {
1320
+ return io_uring_smp_load_acquire(ring->cq.ktail) - *ring->cq.khead;
1321
+ }
1322
+
1323
+ /*
1324
+ * Returns true if there are overflow entries waiting to be flushed onto
1325
+ * the CQ ring
1326
+ */
1327
+ IOURINGINLINE bool io_uring_cq_has_overflow(const struct io_uring *ring)
1328
+ {
1329
+ return IO_URING_READ_ONCE(*ring->sq.kflags) & IORING_SQ_CQ_OVERFLOW;
1330
+ }
1331
+
1332
+ /*
1333
+ * Returns true if the eventfd notification is currently enabled
1334
+ */
1335
+ IOURINGINLINE bool io_uring_cq_eventfd_enabled(const struct io_uring *ring)
1336
+ {
1337
+ if (!ring->cq.kflags)
1338
+ return true;
1339
+
1340
+ return !(*ring->cq.kflags & IORING_CQ_EVENTFD_DISABLED);
1341
+ }
1342
+
1343
+ /*
1344
+ * Toggle eventfd notification on or off, if an eventfd is registered with
1345
+ * the ring.
1346
+ */
1347
+ IOURINGINLINE int io_uring_cq_eventfd_toggle(struct io_uring *ring,
1348
+ bool enabled)
1349
+ {
1350
+ uint32_t flags;
1351
+
1352
+ if (!!enabled == io_uring_cq_eventfd_enabled(ring))
1353
+ return 0;
1354
+
1355
+ if (!ring->cq.kflags)
1356
+ return -EOPNOTSUPP;
1357
+
1358
+ flags = *ring->cq.kflags;
1359
+
1360
+ if (enabled)
1361
+ flags &= ~IORING_CQ_EVENTFD_DISABLED;
1362
+ else
1363
+ flags |= IORING_CQ_EVENTFD_DISABLED;
1364
+
1365
+ IO_URING_WRITE_ONCE(*ring->cq.kflags, flags);
1366
+
1367
+ return 0;
1368
+ }
1369
+
1370
+ /*
1371
+ * Return an IO completion, waiting for 'wait_nr' completions if one isn't
1372
+ * readily available. Returns 0 with cqe_ptr filled in on success, -errno on
1373
+ * failure.
1374
+ */
1375
+ IOURINGINLINE int io_uring_wait_cqe_nr(struct io_uring *ring,
1376
+ struct io_uring_cqe **cqe_ptr,
1377
+ unsigned wait_nr)
1378
+ {
1379
+ return __io_uring_get_cqe(ring, cqe_ptr, 0, wait_nr, NULL);
1380
+ }
1381
+
1382
+ /*
1383
+ * Internal helper, don't use directly in applications. Use one of the
1384
+ * "official" versions of this, io_uring_peek_cqe(), io_uring_wait_cqe(),
1385
+ * or io_uring_wait_cqes*().
1386
+ */
1387
+ IOURINGINLINE int __io_uring_peek_cqe(struct io_uring *ring,
1388
+ struct io_uring_cqe **cqe_ptr,
1389
+ unsigned *nr_available)
1390
+ {
1391
+ struct io_uring_cqe *cqe;
1392
+ int err = 0;
1393
+ unsigned available;
1394
+ unsigned mask = ring->cq.ring_mask;
1395
+ int shift = 0;
1396
+
1397
+ if (ring->flags & IORING_SETUP_CQE32)
1398
+ shift = 1;
1399
+
1400
+ do {
1401
+ unsigned tail = io_uring_smp_load_acquire(ring->cq.ktail);
1402
+ unsigned head = *ring->cq.khead;
1403
+
1404
+ cqe = NULL;
1405
+ available = tail - head;
1406
+ if (!available)
1407
+ break;
1408
+
1409
+ cqe = &ring->cq.cqes[(head & mask) << shift];
1410
+ if (!(ring->features & IORING_FEAT_EXT_ARG) &&
1411
+ cqe->user_data == LIBURING_UDATA_TIMEOUT) {
1412
+ if (cqe->res < 0)
1413
+ err = cqe->res;
1414
+ io_uring_cq_advance(ring, 1);
1415
+ if (!err)
1416
+ continue;
1417
+ cqe = NULL;
1418
+ }
1419
+
1420
+ break;
1421
+ } while (1);
1422
+
1423
+ *cqe_ptr = cqe;
1424
+ if (nr_available)
1425
+ *nr_available = available;
1426
+ return err;
1427
+ }
1428
+
1429
+ /*
1430
+ * Return an IO completion, if one is readily available. Returns 0 with
1431
+ * cqe_ptr filled in on success, -errno on failure.
1432
+ */
1433
+ IOURINGINLINE int io_uring_peek_cqe(struct io_uring *ring,
1434
+ struct io_uring_cqe **cqe_ptr)
1435
+ {
1436
+ if (!__io_uring_peek_cqe(ring, cqe_ptr, NULL) && *cqe_ptr)
1437
+ return 0;
1438
+
1439
+ return io_uring_wait_cqe_nr(ring, cqe_ptr, 0);
1440
+ }
1441
+
1442
+ /*
1443
+ * Return an IO completion, waiting for it if necessary. Returns 0 with
1444
+ * cqe_ptr filled in on success, -errno on failure.
1445
+ */
1446
+ IOURINGINLINE int io_uring_wait_cqe(struct io_uring *ring,
1447
+ struct io_uring_cqe **cqe_ptr)
1448
+ {
1449
+ if (!__io_uring_peek_cqe(ring, cqe_ptr, NULL) && *cqe_ptr)
1450
+ return 0;
1451
+
1452
+ return io_uring_wait_cqe_nr(ring, cqe_ptr, 1);
1453
+ }
1454
+
1455
+ /*
1456
+ * Return an sqe to fill. Application must later call io_uring_submit()
1457
+ * when it's ready to tell the kernel about it. The caller may call this
1458
+ * function multiple times before calling io_uring_submit().
1459
+ *
1460
+ * Returns a vacant sqe, or NULL if we're full.
1461
+ */
1462
+ IOURINGINLINE struct io_uring_sqe *_io_uring_get_sqe(struct io_uring *ring)
1463
+ {
1464
+ struct io_uring_sq *sq = &ring->sq;
1465
+ unsigned int head, next = sq->sqe_tail + 1;
1466
+ int shift = 0;
1467
+
1468
+ if (ring->flags & IORING_SETUP_SQE128)
1469
+ shift = 1;
1470
+ if (!(ring->flags & IORING_SETUP_SQPOLL))
1471
+ head = *sq->khead;
1472
+ else
1473
+ head = io_uring_smp_load_acquire(sq->khead);
1474
+
1475
+ if (next - head <= sq->ring_entries) {
1476
+ struct io_uring_sqe *sqe;
1477
+
1478
+ sqe = &sq->sqes[(sq->sqe_tail & sq->ring_mask) << shift];
1479
+ sq->sqe_tail = next;
1480
+ io_uring_initialize_sqe(sqe);
1481
+ return sqe;
1482
+ }
1483
+
1484
+ return NULL;
1485
+ }
1486
+
1487
+ /*
1488
+ * Return the appropriate mask for a buffer ring of size 'ring_entries'
1489
+ */
1490
+ IOURINGINLINE int io_uring_buf_ring_mask(__u32 ring_entries)
1491
+ {
1492
+ return ring_entries - 1;
1493
+ }
1494
+
1495
+ IOURINGINLINE void io_uring_buf_ring_init(struct io_uring_buf_ring *br)
1496
+ {
1497
+ br->tail = 0;
1498
+ }
1499
+
1500
+ /*
1501
+ * Assign 'buf' with the addr/len/buffer ID supplied
1502
+ */
1503
+ IOURINGINLINE void io_uring_buf_ring_add(struct io_uring_buf_ring *br,
1504
+ void *addr, unsigned int len,
1505
+ unsigned short bid, int mask,
1506
+ int buf_offset)
1507
+ {
1508
+ struct io_uring_buf *buf = &br->bufs[(br->tail + buf_offset) & mask];
1509
+
1510
+ buf->addr = (unsigned long) (uintptr_t) addr;
1511
+ buf->len = len;
1512
+ buf->bid = bid;
1513
+ }
1514
+
1515
+ /*
1516
+ * Make 'count' new buffers visible to the kernel. Called after
1517
+ * io_uring_buf_ring_add() has been called 'count' times to fill in new
1518
+ * buffers.
1519
+ */
1520
+ IOURINGINLINE void io_uring_buf_ring_advance(struct io_uring_buf_ring *br,
1521
+ int count)
1522
+ {
1523
+ unsigned short new_tail = br->tail + count;
1524
+
1525
+ io_uring_smp_store_release(&br->tail, new_tail);
1526
+ }
1527
+
1528
+ IOURINGINLINE void __io_uring_buf_ring_cq_advance(struct io_uring *ring,
1529
+ struct io_uring_buf_ring *br,
1530
+ int cq_count, int buf_count)
1531
+ {
1532
+ io_uring_buf_ring_advance(br, buf_count);
1533
+ io_uring_cq_advance(ring, cq_count);
1534
+ }
1535
+
1536
+ /*
1537
+ * Make 'count' new buffers visible to the kernel while at the same time
1538
+ * advancing the CQ ring seen entries. This can be used when the application
1539
+ * is using ring provided buffers and returns buffers while processing CQEs,
1540
+ * avoiding an extra atomic when needing to increment both the CQ ring and
1541
+ * the ring buffer index at the same time.
1542
+ */
1543
+ IOURINGINLINE void io_uring_buf_ring_cq_advance(struct io_uring *ring,
1544
+ struct io_uring_buf_ring *br,
1545
+ int count)
1546
+ {
1547
+ __io_uring_buf_ring_cq_advance(ring, br, count, count);
1548
+ }
1549
+
1550
+ IOURINGINLINE int io_uring_buf_ring_available(struct io_uring *ring,
1551
+ struct io_uring_buf_ring *br,
1552
+ unsigned short bgid)
1553
+ {
1554
+ uint16_t head;
1555
+ int ret;
1556
+
1557
+ ret = io_uring_buf_ring_head(ring, bgid, &head);
1558
+ if (ret)
1559
+ return ret;
1560
+
1561
+ return (uint16_t) (br->tail - head);
1562
+ }
1563
+
1564
+ #ifndef LIBURING_INTERNAL
1565
+ IOURINGINLINE struct io_uring_sqe *io_uring_get_sqe(struct io_uring *ring)
1566
+ {
1567
+ return _io_uring_get_sqe(ring);
1568
+ }
1569
+ #else
1570
+ struct io_uring_sqe *io_uring_get_sqe(struct io_uring *ring);
1571
+ #endif
1572
+
1573
+ ssize_t io_uring_mlock_size(unsigned entries, unsigned flags);
1574
+ ssize_t io_uring_mlock_size_params(unsigned entries, struct io_uring_params *p);
1575
+
1576
+ /*
1577
+ * Versioning information for liburing.
1578
+ *
1579
+ * Use IO_URING_CHECK_VERSION() for compile time checks including from
1580
+ * preprocessor directives.
1581
+ *
1582
+ * Use io_uring_check_version() for runtime checks of the version of
1583
+ * liburing that was loaded by the dynamic linker.
1584
+ */
1585
+ int io_uring_major_version(void);
1586
+ int io_uring_minor_version(void);
1587
+ bool io_uring_check_version(int major, int minor);
1588
+
1589
+ #define IO_URING_CHECK_VERSION(major,minor) \
1590
+ (major > IO_URING_VERSION_MAJOR || \
1591
+ (major == IO_URING_VERSION_MAJOR && \
1592
+ minor > IO_URING_VERSION_MINOR))
1593
+
1594
+ #ifdef __cplusplus
1595
+ }
1596
+ #endif
1597
+
1598
+ #ifdef IOURINGINLINE
1599
+ #undef IOURINGINLINE
1600
+ #endif
1601
+
1602
+ #endif