xref: /qemu/block/io_uring.c (revision 27a4a30e)
1 /*
2  * Linux io_uring support.
3  *
4  * Copyright (C) 2009 IBM, Corp.
5  * Copyright (C) 2009 Red Hat, Inc.
6  * Copyright (C) 2019 Aarushi Mehta
7  *
8  * This work is licensed under the terms of the GNU GPL, version 2 or later.
9  * See the COPYING file in the top-level directory.
10  */
11 #include "qemu/osdep.h"
12 #include <liburing.h>
13 #include "qemu-common.h"
14 #include "block/aio.h"
15 #include "qemu/queue.h"
16 #include "block/block.h"
17 #include "block/raw-aio.h"
18 #include "qemu/coroutine.h"
19 #include "qapi/error.h"
20 #include "trace.h"
21 
22 /* io_uring ring size */
23 #define MAX_ENTRIES 128
24 
25 typedef struct LuringAIOCB {
26     Coroutine *co;
27     struct io_uring_sqe sqeq;
28     ssize_t ret;
29     QEMUIOVector *qiov;
30     bool is_read;
31     QSIMPLEQ_ENTRY(LuringAIOCB) next;
32 
33     /*
34      * Buffered reads may require resubmission, see
35      * luring_resubmit_short_read().
36      */
37     int total_read;
38     QEMUIOVector resubmit_qiov;
39 } LuringAIOCB;
40 
41 typedef struct LuringQueue {
42     int plugged;
43     unsigned int in_queue;
44     unsigned int in_flight;
45     bool blocked;
46     QSIMPLEQ_HEAD(, LuringAIOCB) submit_queue;
47 } LuringQueue;
48 
49 typedef struct LuringState {
50     AioContext *aio_context;
51 
52     struct io_uring ring;
53 
54     /* io queue for submit at batch.  Protected by AioContext lock. */
55     LuringQueue io_q;
56 
57     /* I/O completion processing.  Only runs in I/O thread.  */
58     QEMUBH *completion_bh;
59 } LuringState;
60 
61 /**
62  * luring_resubmit:
63  *
64  * Resubmit a request by appending it to submit_queue.  The caller must ensure
65  * that ioq_submit() is called later so that submit_queue requests are started.
66  */
67 static void luring_resubmit(LuringState *s, LuringAIOCB *luringcb)
68 {
69     QSIMPLEQ_INSERT_TAIL(&s->io_q.submit_queue, luringcb, next);
70     s->io_q.in_queue++;
71 }
72 
73 /**
74  * luring_resubmit_short_read:
75  *
76  * Before Linux commit 9d93a3f5a0c ("io_uring: punt short reads to async
77  * context") a buffered I/O request with the start of the file range in the
78  * page cache could result in a short read.  Applications need to resubmit the
79  * remaining read request.
80  *
81  * This is a slow path but recent kernels never take it.
82  */
83 static void luring_resubmit_short_read(LuringState *s, LuringAIOCB *luringcb,
84                                        int nread)
85 {
86     QEMUIOVector *resubmit_qiov;
87     size_t remaining;
88 
89     trace_luring_resubmit_short_read(s, luringcb, nread);
90 
91     /* Update read position */
92     luringcb->total_read = nread;
93     remaining = luringcb->qiov->size - luringcb->total_read;
94 
95     /* Shorten qiov */
96     resubmit_qiov = &luringcb->resubmit_qiov;
97     if (resubmit_qiov->iov == NULL) {
98         qemu_iovec_init(resubmit_qiov, luringcb->qiov->niov);
99     } else {
100         qemu_iovec_reset(resubmit_qiov);
101     }
102     qemu_iovec_concat(resubmit_qiov, luringcb->qiov, luringcb->total_read,
103                       remaining);
104 
105     /* Update sqe */
106     luringcb->sqeq.off = nread;
107     luringcb->sqeq.addr = (__u64)(uintptr_t)luringcb->resubmit_qiov.iov;
108     luringcb->sqeq.len = luringcb->resubmit_qiov.niov;
109 
110     luring_resubmit(s, luringcb);
111 }
112 
113 /**
114  * luring_process_completions:
115  * @s: AIO state
116  *
117  * Fetches completed I/O requests, consumes cqes and invokes their callbacks
118  * The function is somewhat tricky because it supports nested event loops, for
119  * example when a request callback invokes aio_poll().
120  *
121  * Function schedules BH completion so it  can be called again in a nested
122  * event loop.  When there are no events left  to complete the BH is being
123  * canceled.
124  *
125  */
126 static void luring_process_completions(LuringState *s)
127 {
128     struct io_uring_cqe *cqes;
129     int total_bytes;
130     /*
131      * Request completion callbacks can run the nested event loop.
132      * Schedule ourselves so the nested event loop will "see" remaining
133      * completed requests and process them.  Without this, completion
134      * callbacks that wait for other requests using a nested event loop
135      * would hang forever.
136      *
137      * This workaround is needed because io_uring uses poll_wait, which
138      * is woken up when new events are added to the uring, thus polling on
139      * the same uring fd will block unless more events are received.
140      *
141      * Other leaf block drivers (drivers that access the data themselves)
142      * are networking based, so they poll sockets for data and run the
143      * correct coroutine.
144      */
145     qemu_bh_schedule(s->completion_bh);
146 
147     while (io_uring_peek_cqe(&s->ring, &cqes) == 0) {
148         LuringAIOCB *luringcb;
149         int ret;
150 
151         if (!cqes) {
152             break;
153         }
154 
155         luringcb = io_uring_cqe_get_data(cqes);
156         ret = cqes->res;
157         io_uring_cqe_seen(&s->ring, cqes);
158         cqes = NULL;
159 
160         /* Change counters one-by-one because we can be nested. */
161         s->io_q.in_flight--;
162         trace_luring_process_completion(s, luringcb, ret);
163 
164         /* total_read is non-zero only for resubmitted read requests */
165         total_bytes = ret + luringcb->total_read;
166 
167         if (ret < 0) {
168             if (ret == -EINTR) {
169                 luring_resubmit(s, luringcb);
170                 continue;
171             }
172         } else if (!luringcb->qiov) {
173             goto end;
174         } else if (total_bytes == luringcb->qiov->size) {
175             ret = 0;
176         /* Only read/write */
177         } else {
178             /* Short Read/Write */
179             if (luringcb->is_read) {
180                 if (ret > 0) {
181                     luring_resubmit_short_read(s, luringcb, ret);
182                     continue;
183                 } else {
184                     /* Pad with zeroes */
185                     qemu_iovec_memset(luringcb->qiov, total_bytes, 0,
186                                       luringcb->qiov->size - total_bytes);
187                     ret = 0;
188                 }
189             } else {
190                 ret = -ENOSPC;
191             }
192         }
193 end:
194         luringcb->ret = ret;
195         qemu_iovec_destroy(&luringcb->resubmit_qiov);
196 
197         /*
198          * If the coroutine is already entered it must be in ioq_submit()
199          * and will notice luringcb->ret has been filled in when it
200          * eventually runs later. Coroutines cannot be entered recursively
201          * so avoid doing that!
202          */
203         if (!qemu_coroutine_entered(luringcb->co)) {
204             aio_co_wake(luringcb->co);
205         }
206     }
207     qemu_bh_cancel(s->completion_bh);
208 }
209 
210 static int ioq_submit(LuringState *s)
211 {
212     int ret = 0;
213     LuringAIOCB *luringcb, *luringcb_next;
214 
215     while (s->io_q.in_queue > 0) {
216         /*
217          * Try to fetch sqes from the ring for requests waiting in
218          * the overflow queue
219          */
220         QSIMPLEQ_FOREACH_SAFE(luringcb, &s->io_q.submit_queue, next,
221                               luringcb_next) {
222             struct io_uring_sqe *sqes = io_uring_get_sqe(&s->ring);
223             if (!sqes) {
224                 break;
225             }
226             /* Prep sqe for submission */
227             *sqes = luringcb->sqeq;
228             QSIMPLEQ_REMOVE_HEAD(&s->io_q.submit_queue, next);
229         }
230         ret = io_uring_submit(&s->ring);
231         trace_luring_io_uring_submit(s, ret);
232         /* Prevent infinite loop if submission is refused */
233         if (ret <= 0) {
234             if (ret == -EAGAIN) {
235                 continue;
236             }
237             break;
238         }
239         s->io_q.in_flight += ret;
240         s->io_q.in_queue  -= ret;
241     }
242     s->io_q.blocked = (s->io_q.in_queue > 0);
243 
244     if (s->io_q.in_flight) {
245         /*
246          * We can try to complete something just right away if there are
247          * still requests in-flight.
248          */
249         luring_process_completions(s);
250     }
251     return ret;
252 }
253 
254 static void luring_process_completions_and_submit(LuringState *s)
255 {
256     aio_context_acquire(s->aio_context);
257     luring_process_completions(s);
258 
259     if (!s->io_q.plugged && s->io_q.in_queue > 0) {
260         ioq_submit(s);
261     }
262     aio_context_release(s->aio_context);
263 }
264 
265 static void qemu_luring_completion_bh(void *opaque)
266 {
267     LuringState *s = opaque;
268     luring_process_completions_and_submit(s);
269 }
270 
271 static void qemu_luring_completion_cb(void *opaque)
272 {
273     LuringState *s = opaque;
274     luring_process_completions_and_submit(s);
275 }
276 
277 static bool qemu_luring_poll_cb(void *opaque)
278 {
279     LuringState *s = opaque;
280     struct io_uring_cqe *cqes;
281 
282     if (io_uring_peek_cqe(&s->ring, &cqes) == 0) {
283         if (cqes) {
284             luring_process_completions_and_submit(s);
285             return true;
286         }
287     }
288 
289     return false;
290 }
291 
292 static void ioq_init(LuringQueue *io_q)
293 {
294     QSIMPLEQ_INIT(&io_q->submit_queue);
295     io_q->plugged = 0;
296     io_q->in_queue = 0;
297     io_q->in_flight = 0;
298     io_q->blocked = false;
299 }
300 
301 void luring_io_plug(BlockDriverState *bs, LuringState *s)
302 {
303     trace_luring_io_plug(s);
304     s->io_q.plugged++;
305 }
306 
307 void luring_io_unplug(BlockDriverState *bs, LuringState *s)
308 {
309     assert(s->io_q.plugged);
310     trace_luring_io_unplug(s, s->io_q.blocked, s->io_q.plugged,
311                            s->io_q.in_queue, s->io_q.in_flight);
312     if (--s->io_q.plugged == 0 &&
313         !s->io_q.blocked && s->io_q.in_queue > 0) {
314         ioq_submit(s);
315     }
316 }
317 
318 /**
319  * luring_do_submit:
320  * @fd: file descriptor for I/O
321  * @luringcb: AIO control block
322  * @s: AIO state
323  * @offset: offset for request
324  * @type: type of request
325  *
326  * Fetches sqes from ring, adds to pending queue and preps them
327  *
328  */
329 static int luring_do_submit(int fd, LuringAIOCB *luringcb, LuringState *s,
330                             uint64_t offset, int type)
331 {
332     int ret;
333     struct io_uring_sqe *sqes = &luringcb->sqeq;
334 
335     switch (type) {
336     case QEMU_AIO_WRITE:
337         io_uring_prep_writev(sqes, fd, luringcb->qiov->iov,
338                              luringcb->qiov->niov, offset);
339         break;
340     case QEMU_AIO_READ:
341         io_uring_prep_readv(sqes, fd, luringcb->qiov->iov,
342                             luringcb->qiov->niov, offset);
343         break;
344     case QEMU_AIO_FLUSH:
345         io_uring_prep_fsync(sqes, fd, IORING_FSYNC_DATASYNC);
346         break;
347     default:
348         fprintf(stderr, "%s: invalid AIO request type, aborting 0x%x.\n",
349                         __func__, type);
350         abort();
351     }
352     io_uring_sqe_set_data(sqes, luringcb);
353 
354     QSIMPLEQ_INSERT_TAIL(&s->io_q.submit_queue, luringcb, next);
355     s->io_q.in_queue++;
356     trace_luring_do_submit(s, s->io_q.blocked, s->io_q.plugged,
357                            s->io_q.in_queue, s->io_q.in_flight);
358     if (!s->io_q.blocked &&
359         (!s->io_q.plugged ||
360          s->io_q.in_flight + s->io_q.in_queue >= MAX_ENTRIES)) {
361         ret = ioq_submit(s);
362         trace_luring_do_submit_done(s, ret);
363         return ret;
364     }
365     return 0;
366 }
367 
368 int coroutine_fn luring_co_submit(BlockDriverState *bs, LuringState *s, int fd,
369                                   uint64_t offset, QEMUIOVector *qiov, int type)
370 {
371     int ret;
372     LuringAIOCB luringcb = {
373         .co         = qemu_coroutine_self(),
374         .ret        = -EINPROGRESS,
375         .qiov       = qiov,
376         .is_read    = (type == QEMU_AIO_READ),
377     };
378     trace_luring_co_submit(bs, s, &luringcb, fd, offset, qiov ? qiov->size : 0,
379                            type);
380     ret = luring_do_submit(fd, &luringcb, s, offset, type);
381 
382     if (ret < 0) {
383         return ret;
384     }
385 
386     if (luringcb.ret == -EINPROGRESS) {
387         qemu_coroutine_yield();
388     }
389     return luringcb.ret;
390 }
391 
392 void luring_detach_aio_context(LuringState *s, AioContext *old_context)
393 {
394     aio_set_fd_handler(old_context, s->ring.ring_fd, false, NULL, NULL, NULL,
395                        s);
396     qemu_bh_delete(s->completion_bh);
397     s->aio_context = NULL;
398 }
399 
400 void luring_attach_aio_context(LuringState *s, AioContext *new_context)
401 {
402     s->aio_context = new_context;
403     s->completion_bh = aio_bh_new(new_context, qemu_luring_completion_bh, s);
404     aio_set_fd_handler(s->aio_context, s->ring.ring_fd, false,
405                        qemu_luring_completion_cb, NULL, qemu_luring_poll_cb, s);
406 }
407 
408 LuringState *luring_init(Error **errp)
409 {
410     int rc;
411     LuringState *s = g_new0(LuringState, 1);
412     struct io_uring *ring = &s->ring;
413 
414     trace_luring_init_state(s, sizeof(*s));
415 
416     rc = io_uring_queue_init(MAX_ENTRIES, ring, 0);
417     if (rc < 0) {
418         error_setg_errno(errp, errno, "failed to init linux io_uring ring");
419         g_free(s);
420         return NULL;
421     }
422 
423     ioq_init(&s->io_q);
424     return s;
425 
426 }
427 
428 void luring_cleanup(LuringState *s)
429 {
430     io_uring_queue_exit(&s->ring);
431     g_free(s);
432     trace_luring_cleanup_state(s);
433 }
434