Context Navigation

io.cfa @ 6502a2b

ADTarm-ehast-experimentalenumforall-pointer-decayjacob/cs343-translationnew-astnew-ast-unique-exprpthread-emulationqualifiedEnum

Last change on this file since 6502a2b was 6502a2b, checked in by Thierry Delisle <tdelisle@…>, 4 years ago
Tentative fix to the deadlock in cluster shutdown
Property mode set to `100644`
File size: 29.5 KB

Line
1	//
2	// Cforall Version 1.0.0 Copyright (C) 2020 University of Waterloo
3	//
4	// The contents of this file are covered under the licence agreement in the
5	// file "LICENCE" distributed with Cforall.
6	//
7	// io.cfa --
8	//
9	// Author : Thierry Delisle
10	// Created On : Thu Apr 23 17:31:00 2020
11	// Last Modified By :
12	// Last Modified On :
13	// Update Count :
14	//
15
16	// #define __CFA_DEBUG_PRINT_IO__
17	// #define __CFA_DEBUG_PRINT_IO_CORE__
18
19	#include "kernel.hfa"
20
21	#if !defined(HAVE_LINUX_IO_URING_H)
22	void __kernel_io_startup( cluster & ) {
23	// Nothing to do without io_uring
24	}
25
26	void __kernel_io_start_thrd( cluster & ) {
27	// Nothing to do without io_uring
28	}
29
30	void __kernel_io_stop_thrd ( cluster & ) {
31	// Nothing to do without io_uring
32	}
33
34	void __kernel_io_shutdown( cluster & ) {
35	// Nothing to do without io_uring
36	}
37
38	#else
39	extern "C" {
40	#define _GNU_SOURCE /* See feature_test_macros(7) */
41	#include <errno.h>
42	#include <stdint.h>
43	#include <string.h>
44	#include <unistd.h>
45	#include <sys/mman.h>
46	#include <sys/syscall.h>
47
48	#include <linux/io_uring.h>
49	}
50
51	#include "bits/signal.hfa"
52	#include "kernel_private.hfa"
53	#include "thread.hfa"
54
55	uint32_t entries_per_cluster() {
56	return 256;
57	}
58
59	static void * __io_poller_slow( void * arg );
60
61	// Weirdly, some systems that do support io_uring don't actually define these
62	#ifdef __alpha__
63	/*
64	* alpha is the only exception, all other architectures
65	* have common numbers for new system calls.
66	*/
67	#ifndef __NR_io_uring_setup
68	#define __NR_io_uring_setup 535
69	#endif
70	#ifndef __NR_io_uring_enter
71	#define __NR_io_uring_enter 536
72	#endif
73	#ifndef __NR_io_uring_register
74	#define __NR_io_uring_register 537
75	#endif
76	#else /* !__alpha__ */
77	#ifndef __NR_io_uring_setup
78	#define __NR_io_uring_setup 425
79	#endif
80	#ifndef __NR_io_uring_enter
81	#define __NR_io_uring_enter 426
82	#endif
83	#ifndef __NR_io_uring_register
84	#define __NR_io_uring_register 427
85	#endif
86	#endif
87
88	#if defined(__CFA_IO_POLLING_USER__)
89	void ?{}( __io_poller_fast & this, struct cluster & cltr ) {
90	this.ring = &cltr.io;
91	this.waiting = true;
92	(this.thrd){ "Fast I/O Poller", cltr };
93	}
94	void ^?{}( __io_poller_fast & mutex this );
95	void main( __io_poller_fast & this );
96	static inline $thread * get_thread( __io_poller_fast & this ) { return &this.thrd; }
97	void ^?{}( __io_poller_fast & mutex this ) {}
98	#endif
99
100	//=============================================================================================
101	// I/O Startup / Shutdown logic
102	//=============================================================================================
103	void __kernel_io_startup( cluster & this, bool main_cluster ) {
104	// Step 1 : call to setup
105	struct io_uring_params params;
106	memset(&params, 0, sizeof(params));
107
108	uint32_t nentries = entries_per_cluster();
109
110	int fd = syscall(__NR_io_uring_setup, nentries, &params );
111	if(fd < 0) {
112	abort("KERNEL ERROR: IO_URING SETUP - %s\n", strerror(errno));
113	}
114
115	// Step 2 : mmap result
116	memset(&this.io, 0, sizeof(struct io_ring));
117	struct io_uring_sq & sq = this.io.submit_q;
118	struct io_uring_cq & cq = this.io.completion_q;
119
120	// calculate the right ring size
121	sq.ring_sz = params.sq_off.array + (params.sq_entries * sizeof(unsigned) );
122	cq.ring_sz = params.cq_off.cqes + (params.cq_entries * sizeof(struct io_uring_cqe));
123
124	// Requires features
125	#if defined(IORING_FEAT_SINGLE_MMAP)
126	// adjust the size according to the parameters
127	if ((params.features & IORING_FEAT_SINGLE_MMAP) != 0) {
128	cq->ring_sz = sq->ring_sz = max(cq->ring_sz, sq->ring_sz);
129	}
130	#endif
131
132	// mmap the Submit Queue into existence
133	sq.ring_ptr = mmap(0, sq.ring_sz, PROT_READ \| PROT_WRITE, MAP_SHARED \| MAP_POPULATE, fd, IORING_OFF_SQ_RING);
134	if (sq.ring_ptr == (void*)MAP_FAILED) {
135	abort("KERNEL ERROR: IO_URING MMAP1 - %s\n", strerror(errno));
136	}
137
138	// Requires features
139	#if defined(IORING_FEAT_SINGLE_MMAP)
140	// mmap the Completion Queue into existence (may or may not be needed)
141	if ((params.features & IORING_FEAT_SINGLE_MMAP) != 0) {
142	cq->ring_ptr = sq->ring_ptr;
143	}
144	else
145	#endif
146	{
147	// We need multiple call to MMAP
148	cq.ring_ptr = mmap(0, cq.ring_sz, PROT_READ \| PROT_WRITE, MAP_SHARED \| MAP_POPULATE, fd, IORING_OFF_CQ_RING);
149	if (cq.ring_ptr == (void*)MAP_FAILED) {
150	munmap(sq.ring_ptr, sq.ring_sz);
151	abort("KERNEL ERROR: IO_URING MMAP2 - %s\n", strerror(errno));
152	}
153	}
154
155	// mmap the submit queue entries
156	size_t size = params.sq_entries * sizeof(struct io_uring_sqe);
157	sq.sqes = (struct io_uring_sqe *)mmap(0, size, PROT_READ \| PROT_WRITE, MAP_SHARED \| MAP_POPULATE, fd, IORING_OFF_SQES);
158	if (sq.sqes == (struct io_uring_sqe *)MAP_FAILED) {
159	munmap(sq.ring_ptr, sq.ring_sz);
160	if (cq.ring_ptr != sq.ring_ptr) munmap(cq.ring_ptr, cq.ring_sz);
161	abort("KERNEL ERROR: IO_URING MMAP3 - %s\n", strerror(errno));
162	}
163
164	// Get the pointers from the kernel to fill the structure
165	// submit queue
166	sq.head = (volatile uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.head);
167	sq.tail = (volatile uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.tail);
168	sq.mask = ( const uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.ring_mask);
169	sq.num = ( const uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.ring_entries);
170	sq.flags = ( uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.flags);
171	sq.dropped = ( uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.dropped);
172	sq.array = ( uint32_t *)(((intptr_t)sq.ring_ptr) + params.sq_off.array);
173	sq.alloc = *sq.tail;
174	sq.ready = *sq.tail;
175
176	// completion queue
177	cq.head = (volatile uint32_t *)(((intptr_t)cq.ring_ptr) + params.cq_off.head);
178	cq.tail = (volatile uint32_t *)(((intptr_t)cq.ring_ptr) + params.cq_off.tail);
179	cq.mask = ( const uint32_t *)(((intptr_t)cq.ring_ptr) + params.cq_off.ring_mask);
180	cq.num = ( const uint32_t *)(((intptr_t)cq.ring_ptr) + params.cq_off.ring_entries);
181	cq.overflow = ( uint32_t *)(((intptr_t)cq.ring_ptr) + params.cq_off.overflow);
182	cq.cqes = (struct io_uring_cqe *)(((intptr_t)cq.ring_ptr) + params.cq_off.cqes);
183
184	// some paranoid checks
185	/* paranoid / verifyf( (cq.mask) == ((cq.num) - 1ul32), "IO_URING Expected mask to be %u (%u entries), was %u", (cq.num) - 1ul32, cq.num, cq.mask );
186	/* paranoid / verifyf( (cq.num) >= nentries, "IO_URING Expected %u entries, got %u", nentries, *cq.num );
187	/* paranoid / verifyf( (cq.head) == 0, "IO_URING Expected head to be 0, got %u", *cq.head );
188	/* paranoid / verifyf( (cq.tail) == 0, "IO_URING Expected tail to be 0, got %u", *cq.tail );
189
190	/* paranoid / verifyf( (sq.mask) == ((sq.num) - 1ul32), "IO_URING Expected mask to be %u (%u entries), was %u", (sq.num) - 1ul32, sq.num, sq.mask );
191	/* paranoid / verifyf( (sq.num) >= nentries, "IO_URING Expected %u entries, got %u", nentries, *sq.num );
192	/* paranoid / verifyf( (sq.head) == 0, "IO_URING Expected head to be 0, got %u", *sq.head );
193	/* paranoid / verifyf( (sq.tail) == 0, "IO_URING Expected tail to be 0, got %u", *sq.tail );
194
195	// Update the global ring info
196	this.io.flags = params.flags;
197	this.io.fd = fd;
198	this.io.done = false;
199	(this.io.submit){ min(sq.num, cq.num) };
200
201	// Initialize statistics
202	#if !defined(__CFA_NO_STATISTICS__)
203	this.io.submit_q.stats.submit_avg.val = 0;
204	this.io.submit_q.stats.submit_avg.cnt = 0;
205	this.io.completion_q.stats.completed_avg.val = 0;
206	this.io.completion_q.stats.completed_avg.cnt = 0;
207	#endif
208
209	if(!main_cluster) {
210	__kernel_io_finish_start( this );
211	}
212	}
213
214	void __kernel_io_finish_start( cluster & this ) {
215	#if defined(__CFA_IO_POLLING_USER__)
216	__cfadbg_print_safe(io_core, "Kernel I/O : Creating fast poller for cluter %p\n", &this);
217	(this.io.poller.fast){ this };
218	__thrd_start( this.io.poller.fast, main );
219	#endif
220
221	// Create the poller thread
222	__cfadbg_print_safe(io_core, "Kernel I/O : Creating slow poller for cluter %p\n", &this);
223	this.io.poller.slow.stack = __create_pthread( &this.io.poller.slow.kthrd, __io_poller_slow, &this );
224	}
225
226	void __kernel_io_prepare_stop( cluster & this ) {
227	__cfadbg_print_safe(io_core, "Kernel I/O : Stopping pollers for cluster\n", &this);
228	// Notify the poller thread of the shutdown
229	__atomic_store_n(&this.io.done, true, __ATOMIC_SEQ_CST);
230
231	// Stop the IO Poller
232	sigval val = { 1 };
233	pthread_sigqueue( this.io.poller.slow.kthrd, SIGUSR1, val );
234	#if defined(__CFA_IO_POLLING_USER__)
235	post( this.io.poller.sem );
236	#endif
237
238	// Wait for the poller thread to finish
239	pthread_join( this.io.poller.slow.kthrd, 0p );
240	free( this.io.poller.slow.stack );
241
242	__cfadbg_print_safe(io_core, "Kernel I/O : Slow poller stopped for cluster\n", &this);
243
244	#if defined(__CFA_IO_POLLING_USER__)
245	verify( this.io.poller.fast.waiting );
246	verify( this.io.poller.fast.thrd.state == Blocked );
247
248	this.io.poller.fast.thrd.curr_cluster = mainCluster;
249
250	// unpark the fast io_poller
251	unpark( &this.io.poller.fast.thrd __cfaabi_dbg_ctx2 );
252
253	^(this.io.poller.fast){};
254
255	__cfadbg_print_safe(io_core, "Kernel I/O : Fast poller stopped for cluster\n", &this);
256	#endif
257	}
258
259	void __kernel_io_shutdown( cluster & this, bool main_cluster ) {
260	if(!main_cluster) {
261	__kernel_io_prepare_stop( this );
262	}
263
264	// print statistics
265	#if !defined(__CFA_NO_STATISTICS__)
266	if(this.print_stats) {
267	__cfaabi_bits_print_safe( STDERR_FILENO,
268	"----- I/O uRing Stats -----\n"
269	"- total submit calls : %llu\n"
270	"- avg submit : %lf\n"
271	"- total wait calls : %llu\n"
272	"- avg completion/wait : %lf\n",
273	this.io.submit_q.stats.submit_avg.cnt,
274	((double)this.io.submit_q.stats.submit_avg.val) / this.io.submit_q.stats.submit_avg.cnt,
275	this.io.completion_q.stats.completed_avg.cnt,
276	((double)this.io.completion_q.stats.completed_avg.val) / this.io.completion_q.stats.completed_avg.cnt
277	);
278	}
279	#endif
280
281	// Shutdown the io rings
282	struct io_uring_sq & sq = this.io.submit_q;
283	struct io_uring_cq & cq = this.io.completion_q;
284
285	// unmap the submit queue entries
286	munmap(sq.sqes, (sq.num) sizeof(struct io_uring_sqe));
287
288	// unmap the Submit Queue ring
289	munmap(sq.ring_ptr, sq.ring_sz);
290
291	// unmap the Completion Queue ring, if it is different
292	if (cq.ring_ptr != sq.ring_ptr) {
293	munmap(cq.ring_ptr, cq.ring_sz);
294	}
295
296	// close the file descriptor
297	close(this.io.fd);
298	}
299
300	//=============================================================================================
301	// I/O Polling
302	//=============================================================================================
303	struct io_user_data {
304	int32_t result;
305	$thread * thrd;
306	};
307
308	// Process a single completion message from the io_uring
309	// This is NOT thread-safe
310	static int __drain_io( struct io_ring & ring, sigset_t * mask, int waitcnt, bool in_kernel ) {
311	int ret = syscall( __NR_io_uring_enter, ring.fd, 0, waitcnt, IORING_ENTER_GETEVENTS, mask, _NSIG / 8);
312	if( ret < 0 ) {
313	switch((int)errno) {
314	case EAGAIN:
315	case EINTR:
316	return -EAGAIN;
317	default:
318	abort( "KERNEL ERROR: IO_URING WAIT - %s\n", strerror(errno) );
319	}
320	}
321
322	// Drain the queue
323	unsigned head = *ring.completion_q.head;
324	unsigned tail = __atomic_load_n(ring.completion_q.tail, __ATOMIC_ACQUIRE);
325
326	// Nothing was new return 0
327	if (head == tail) {
328	#if !defined(__CFA_NO_STATISTICS__)
329	ring.completion_q.stats.completed_avg.cnt += 1;
330	#endif
331	return 0;
332	}
333
334	uint32_t count = tail - head;
335	for(i; count) {
336	unsigned idx = (head + i) & (*ring.completion_q.mask);
337	struct io_uring_cqe & cqe = ring.completion_q.cqes[idx];
338
339	/* paranoid */ verify(&cqe);
340
341	struct io_user_data * data = (struct io_user_data *)cqe.user_data;
342	__cfadbg_print_safe( io, "Kernel I/O : Performed reading io cqe %p, result %d for %p\n", data, cqe.res, data->thrd );
343
344	data->result = cqe.res;
345	if(!in_kernel) { unpark( data->thrd __cfaabi_dbg_ctx2 ); }
346	else { __unpark( data->thrd __cfaabi_dbg_ctx2 ); }
347	}
348
349	// Allow new submissions to happen
350	V(ring.submit, count);
351
352	// Mark to the kernel that the cqe has been seen
353	// Ensure that the kernel only sees the new value of the head index after the CQEs have been read.
354	__atomic_fetch_add( ring.completion_q.head, count, __ATOMIC_RELAXED );
355
356	// Update statistics
357	#if !defined(__CFA_NO_STATISTICS__)
358	ring.completion_q.stats.completed_avg.val += count;
359	ring.completion_q.stats.completed_avg.cnt += 1;
360	#endif
361
362	return count;
363	}
364
365	static void * __io_poller_slow( void * arg ) {
366	cluster * cltr = (cluster *)arg;
367	struct io_ring & ring = cltr->io;
368
369	sigset_t mask;
370	sigfillset(&mask);
371	if ( pthread_sigmask( SIG_BLOCK, &mask, 0p ) == -1 ) {
372	abort( "KERNEL ERROR: IO_URING - pthread_sigmask" );
373	}
374
375	sigdelset( &mask, SIGUSR1 );
376
377	verify( (ring.submit_q.head) == (ring.submit_q.tail) );
378	verify( (ring.completion_q.head) == (ring.completion_q.tail) );
379
380	__cfadbg_print_safe(io_core, "Kernel I/O : Slow poller for ring %p ready\n", &ring);
381
382	while(!__atomic_load_n(&ring.done, __ATOMIC_SEQ_CST)) {
383	#if defined(__CFA_IO_POLLING_USER__)
384
385	// In the user-thread approach drain and if anything was drained,
386	// batton pass to the user-thread
387	int count = __drain_io( ring, &mask, 1, true );
388	if(count > 0) {
389	__cfadbg_print_safe(io_core, "Kernel I/O : Moving to ring %p to fast poller\n", &ring);
390	__unpark( &ring.poller.fast.thrd __cfaabi_dbg_ctx2 );
391	wait( ring.poller.sem );
392	}
393
394	#else
395
396	//In the naive approach, just poll the io completion queue directly
397	__drain_io( ring, &mask, 1, true );
398
399	#endif
400	}
401
402	__cfadbg_print_safe(io_core, "Kernel I/O : Slow poller for ring %p stopping\n", &ring);
403
404	return 0p;
405	}
406
407	#if defined(__CFA_IO_POLLING_USER__)
408	void main( __io_poller_fast & this ) {
409	// Start parked
410	park( __cfaabi_dbg_ctx );
411
412	__cfadbg_print_safe(io_core, "Kernel I/O : Fast poller for ring %p ready\n", &this.ring);
413
414	// Then loop until we need to start
415	while(!__atomic_load_n(&this.ring->done, __ATOMIC_SEQ_CST)) {
416	// Drain the io
417	this.waiting = false;
418	int ret = __drain_io( *this.ring, 0p, 0, false );
419	this.waiting = true;
420	if(0 > ret) {
421	// If we got something, just yield and check again
422	yield();
423	}
424	else {
425	// We didn't get anything baton pass to the slow poller
426	__cfadbg_print_safe(io_core, "Kernel I/O : Moving to ring %p to slow poller\n", &this.ring);
427	post( this.ring->poller.sem );
428	park( __cfaabi_dbg_ctx );
429	}
430	}
431
432	__cfadbg_print_safe(io_core, "Kernel I/O : Fast poller for ring %p stopping\n", &this.ring);
433	}
434	#endif
435
436	//=============================================================================================
437	// I/O Submissions
438	//=============================================================================================
439
440	// Submition steps :
441	// 1 - We need to make sure we don't overflow any of the buffer, P(ring.submit) to make sure
442	// entries are available. The semaphore make sure that there is no more operations in
443	// progress then the number of entries in the buffer. This probably limits concurrency
444	// more than necessary since submitted but not completed operations don't need any
445	// entries in user space. However, I don't know what happens if we overflow the buffers
446	// because too many requests completed at once. This is a safe approach in all cases.
447	// Furthermore, with hundreds of entries, this may be okay.
448	//
449	// 2 - Allocate a queue entry. The ring already has memory for all entries but only the ones
450	// listed in sq.array are visible by the kernel. For those not listed, the kernel does not
451	// offer any assurance that an entry is not being filled by multiple flags. Therefore, we
452	// need to write an allocator that allows allocating concurrently.
453	//
454	// 3 - Actually fill the submit entry, this is the only simple and straightforward step.
455	//
456	// 4 - Append the entry index to the array and adjust the tail accordingly. This operation
457	// needs to arrive to two concensus at the same time:
458	// A - The order in which entries are listed in the array: no two threads must pick the
459	// same index for their entries
460	// B - When can the tail be update for the kernel. EVERY entries in the array between
461	// head and tail must be fully filled and shouldn't ever be touched again.
462	//
463
464	static inline [* struct io_uring_sqe, uint32_t] __submit_alloc( struct io_ring & ring ) {
465	// Wait for a spot to be available
466	P(ring.submit);
467
468	// Allocate the sqe
469	uint32_t idx = __atomic_fetch_add(&ring.submit_q.alloc, 1ul32, __ATOMIC_SEQ_CST);
470
471	// Validate that we didn't overflow anything
472	// Check that nothing overflowed
473	/* paranoid */ verify( true );
474
475	// Check that it goes head -> tail -> alloc and never head -> alloc -> tail
476	/* paranoid */ verify( true );
477
478	// Return the sqe
479	return [&ring.submit_q.sqes[ idx & (*ring.submit_q.mask)], idx];
480	}
481
482	static inline void __submit( struct io_ring & ring, uint32_t idx ) {
483	// get mutual exclusion
484	lock(ring.submit_q.lock __cfaabi_dbg_ctx2);
485
486	// Append to the list of ready entries
487	uint32_t * tail = ring.submit_q.tail;
488	const uint32_t mask = *ring.submit_q.mask;
489
490	ring.submit_q.array[ (*tail) & mask ] = idx & mask;
491	__atomic_fetch_add(tail, 1ul32, __ATOMIC_SEQ_CST);
492
493	// Submit however, many entries need to be submitted
494	int ret = syscall( __NR_io_uring_enter, ring.fd, 1, 0, 0, 0p, 0);
495	if( ret < 0 ) {
496	switch((int)errno) {
497	default:
498	abort( "KERNEL ERROR: IO_URING SUBMIT - %s\n", strerror(errno) );
499	}
500	}
501
502	// update statistics
503	#if !defined(__CFA_NO_STATISTICS__)
504	ring.submit_q.stats.submit_avg.val += 1;
505	ring.submit_q.stats.submit_avg.cnt += 1;
506	#endif
507
508	unlock(ring.submit_q.lock);
509	// Make sure that idx was submitted
510	// Be careful to not get false positive if we cycled the entire list or that someone else submitted for us
511	__cfadbg_print_safe( io, "Kernel I/O : Performed io_submit for %p, returned %d\n", active_thread(), ret );
512	}
513
514	static inline void ?{}(struct io_uring_sqe & this, uint8_t opcode, int fd) {
515	this.opcode = opcode;
516	#if !defined(IOSQE_ASYNC)
517	this.flags = 0;
518	#else
519	this.flags = IOSQE_ASYNC;
520	#endif
521	this.ioprio = 0;
522	this.fd = fd;
523	this.off = 0;
524	this.addr = 0;
525	this.len = 0;
526	this.rw_flags = 0;
527	this.__pad2[0] = this.__pad2[1] = this.__pad2[2] = 0;
528	}
529
530	static inline void ?{}(struct io_uring_sqe & this, uint8_t opcode, int fd, void * addr, uint32_t len, uint64_t off ) {
531	(this){ opcode, fd };
532	this.off = off;
533	this.addr = (uint64_t)addr;
534	this.len = len;
535	}
536
537
538	//=============================================================================================
539	// I/O Interface
540	//=============================================================================================
541
542	#define __submit_prelude \
543	struct io_ring & ring = active_cluster()->io; \
544	struct io_uring_sqe * sqe; \
545	uint32_t idx; \
546	[sqe, idx] = __submit_alloc( ring );
547
548	#define __submit_wait \
549	io_user_data data = { 0, active_thread() }; \
550	/__cfaabi_bits_print_safe( STDERR_FILENO, "Preparing user data %p for %p\n", &data, data.thrd );/ \
551	sqe->user_data = (uint64_t)&data; \
552	__submit( ring, idx ); \
553	park( __cfaabi_dbg_ctx ); \
554	return data.result;
555	#endif
556
557	// Some forward declarations
558	extern "C" {
559	#include <unistd.h>
560	#include <sys/types.h>
561	#include <sys/socket.h>
562	#include <sys/syscall.h>
563	struct iovec;
564	extern ssize_t preadv2 (int fd, const struct iovec *iov, int iovcnt, off_t offset, int flags);
565	extern ssize_t pwritev2(int fd, const struct iovec *iov, int iovcnt, off_t offset, int flags);
566
567	extern int fsync(int fd);
568	extern int sync_file_range(int fd, int64_t offset, int64_t nbytes, unsigned int flags);
569
570	struct msghdr;
571	struct sockaddr;
572	extern ssize_t sendmsg(int sockfd, const struct msghdr *msg, int flags);
573	extern ssize_t recvmsg(int sockfd, struct msghdr *msg, int flags);
574	extern ssize_t send(int sockfd, const void *buf, size_t len, int flags);
575	extern ssize_t recv(int sockfd, void *buf, size_t len, int flags);
576	extern int accept4(int sockfd, struct sockaddr addr, socklen_t addrlen, int flags);
577	extern int connect(int sockfd, const struct sockaddr *addr, socklen_t addrlen);
578
579	extern int fallocate(int fd, int mode, uint64_t offset, uint64_t len);
580	extern int posix_fadvise(int fd, uint64_t offset, uint64_t len, int advice);
581	extern int madvise(void *addr, size_t length, int advice);
582
583	extern int openat(int dirfd, const char *pathname, int flags, mode_t mode);
584	extern int close(int fd);
585
586	struct statx;
587	extern int statx(int dirfd, const char pathname, int flags, unsigned int mask, struct statx statxbuf);
588
589	extern ssize_t read (int fd, void *buf, size_t count);
590	}
591
592	//-----------------------------------------------------------------------------
593	// Asynchronous operations
594	ssize_t cfa_preadv2(int fd, const struct iovec *iov, int iovcnt, off_t offset, int flags) {
595	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_READV)
596	return preadv2(fd, iov, iovcnt, offset, flags);
597	#else
598	__submit_prelude
599
600	(*sqe){ IORING_OP_READV, fd, iov, iovcnt, offset };
601
602	__submit_wait
603	#endif
604	}
605
606	ssize_t cfa_pwritev2(int fd, const struct iovec *iov, int iovcnt, off_t offset, int flags) {
607	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_WRITEV)
608	return pwritev2(fd, iov, iovcnt, offset, flags);
609	#else
610	__submit_prelude
611
612	(*sqe){ IORING_OP_WRITEV, fd, iov, iovcnt, offset };
613
614	__submit_wait
615	#endif
616	}
617
618	int cfa_fsync(int fd) {
619	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_FSYNC)
620	return fsync(fd);
621	#else
622	__submit_prelude
623
624	(*sqe){ IORING_OP_FSYNC, fd };
625
626	__submit_wait
627	#endif
628	}
629
630	int cfa_sync_file_range(int fd, int64_t offset, int64_t nbytes, unsigned int flags) {
631	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_SYNC_FILE_RANGE)
632	return sync_file_range(fd, offset, nbytes, flags);
633	#else
634	__submit_prelude
635
636	(*sqe){ IORING_OP_SYNC_FILE_RANGE, fd };
637	sqe->off = offset;
638	sqe->len = nbytes;
639	sqe->sync_range_flags = flags;
640
641	__submit_wait
642	#endif
643	}
644
645
646	ssize_t cfa_sendmsg(int sockfd, const struct msghdr *msg, int flags) {
647	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_SENDMSG)
648	return sendmsg(sockfd, msg, flags);
649	#else
650	__submit_prelude
651
652	(*sqe){ IORING_OP_SENDMSG, sockfd, msg, 1, 0 };
653	sqe->msg_flags = flags;
654
655	__submit_wait
656	#endif
657	}
658
659	ssize_t cfa_recvmsg(int sockfd, struct msghdr *msg, int flags) {
660	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_RECVMSG)
661	return recvmsg(sockfd, msg, flags);
662	#else
663	__submit_prelude
664
665	(*sqe){ IORING_OP_RECVMSG, sockfd, msg, 1, 0 };
666	sqe->msg_flags = flags;
667
668	__submit_wait
669	#endif
670	}
671
672	ssize_t cfa_send(int sockfd, const void *buf, size_t len, int flags) {
673	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_SEND)
674	return send( sockfd, buf, len, flags );
675	#else
676	__submit_prelude
677
678	(*sqe){ IORING_OP_SEND, sockfd };
679	sqe->addr = (uint64_t)buf;
680	sqe->len = len;
681	sqe->msg_flags = flags;
682
683	__submit_wait
684	#endif
685	}
686
687	ssize_t cfa_recv(int sockfd, void *buf, size_t len, int flags) {
688	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_RECV)
689	return recv( sockfd, buf, len, flags );
690	#else
691	__submit_prelude
692
693	(*sqe){ IORING_OP_RECV, sockfd };
694	sqe->addr = (uint64_t)buf;
695	sqe->len = len;
696	sqe->msg_flags = flags;
697
698	__submit_wait
699	#endif
700	}
701
702	int cfa_accept4(int sockfd, struct sockaddr addr, socklen_t addrlen, int flags) {
703	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_ACCEPT)
704	return accept4( sockfd, addr, addrlen, flags );
705	#else
706	__submit_prelude
707
708	(*sqe){ IORING_OP_ACCEPT, sockfd };
709	sqe->addr = addr;
710	sqe->addr2 = addrlen;
711	sqe->accept_flags = flags;
712
713	__submit_wait
714	#endif
715	}
716
717	int cfa_connect(int sockfd, const struct sockaddr *addr, socklen_t addrlen) {
718	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_CONNECT)
719	return connect( sockfd, addr, addrlen );
720	#else
721	__submit_prelude
722
723	(*sqe){ IORING_OP_CONNECT, sockfd };
724	sqe->addr = (uint64_t)addr;
725	sqe->off = addrlen;
726
727	__submit_wait
728	#endif
729	}
730
731	int cfa_fallocate(int fd, int mode, uint64_t offset, uint64_t len) {
732	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_FALLOCATE)
733	return fallocate( fd, mode, offset, len );
734	#else
735	__submit_prelude
736
737	(*sqe){ IORING_OP_FALLOCATE, fd };
738	sqe->off = offset;
739	sqe->len = length;
740	sqe->mode = mode;
741
742	__submit_wait
743	#endif
744	}
745
746	int cfa_fadvise(int fd, uint64_t offset, uint64_t len, int advice) {
747	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_FADVISE)
748	return posix_fadvise( fd, offset, len, advice );
749	#else
750	__submit_prelude
751
752	(*sqe){ IORING_OP_FADVISE, fd };
753	sqe->off = (uint64_t)offset;
754	sqe->len = length;
755	sqe->fadvise_advice = advice;
756
757	__submit_wait
758	#endif
759	}
760
761	int cfa_madvise(void *addr, size_t length, int advice) {
762	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_MADVISE)
763	return madvise( addr, length, advice );
764	#else
765	__submit_prelude
766
767	(*sqe){ IORING_OP_MADVISE, 0 };
768	sqe->addr = (uint64_t)addr;
769	sqe->len = length;
770	sqe->fadvise_advice = advice;
771
772	__submit_wait
773	#endif
774	}
775
776	int cfa_openat(int dirfd, const char *pathname, int flags, mode_t mode) {
777	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_OPENAT)
778	return openat( dirfd, pathname, flags, mode );
779	#else
780	__submit_prelude
781
782	(*sqe){ IORING_OP_OPENAT, dirfd };
783	sqe->addr = (uint64_t)pathname;
784	sqe->open_flags = flags;
785	sqe->mode = mode;
786
787	__submit_wait
788	#endif
789	}
790
791	int cfa_close(int fd) {
792	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_CLOSE)
793	return close( fd );
794	#else
795	__submit_prelude
796
797	(*sqe){ IORING_OP_CLOSE, fd };
798
799	__submit_wait
800	#endif
801	}
802
803	int cfa_statx(int dirfd, const char pathname, int flags, unsigned int mask, struct statx statxbuf) {
804	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_STATX)
805	//return statx( dirfd, pathname, flags, mask, statxbuf );
806	return syscall( __NR_statx, dirfd, pathname, flags, mask, statxbuf );
807	#else
808	__submit_prelude
809
810	(*sqe){ IORING_OP_STATX, dirfd };
811	sqe->addr = (uint64_t)pathname;
812	sqe->statx_flags = flags;
813	sqe->len = mask;
814	sqe->off = (uint64_t)statxbuf;
815
816	__submit_wait
817	#endif
818	}
819
820
821	ssize_t cfa_read(int fd, void *buf, size_t count) {
822	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_READ)
823	return read( fd, buf, count );
824	#else
825	__submit_prelude
826
827	(*sqe){ IORING_OP_READ, fd, buf, count, 0 };
828
829	__submit_wait
830	#endif
831	}
832
833	ssize_t cfa_write(int fd, void *buf, size_t count) {
834	#if !defined(HAVE_LINUX_IO_URING_H) \|\| !defined(IORING_OP_WRITE)
835	return read( fd, buf, count );
836	#else
837	__submit_prelude
838
839	(*sqe){ IORING_OP_WRITE, fd, buf, count, 0 };
840
841	__submit_wait
842	#endif
843	}
844
845	//-----------------------------------------------------------------------------
846	// Check if a function is asynchronous
847
848	// Macro magic to reduce the size of the following switch case
849	#define IS_DEFINED_APPLY(f, ...) f(__VA_ARGS__)
850	#define IS_DEFINED_SECOND(first, second, ...) second
851	#define IS_DEFINED_TEST(expansion) _CFA_IO_FEATURE_##expansion
852	#define IS_DEFINED(macro) IS_DEFINED_APPLY( IS_DEFINED_SECOND,IS_DEFINED_TEST(macro) false, true)
853
854	bool has_user_level_blocking( fptr_t func ) {
855	#if defined(HAVE_LINUX_IO_URING_H)
856	if( /func == (fptr_t)preadv2 \|\| /
857	func == (fptr_t)cfa_preadv2 )
858	#define _CFA_IO_FEATURE_IORING_OP_READV ,
859	return IS_DEFINED(IORING_OP_READV);
860
861	if( /func == (fptr_t)pwritev2 \|\| /
862	func == (fptr_t)cfa_pwritev2 )
863	#define _CFA_IO_FEATURE_IORING_OP_WRITEV ,
864	return IS_DEFINED(IORING_OP_WRITEV);
865
866	if( /func == (fptr_t)fsync \|\| /
867	func == (fptr_t)cfa_fsync )
868	#define _CFA_IO_FEATURE_IORING_OP_FSYNC ,
869	return IS_DEFINED(IORING_OP_FSYNC);
870
871	if( /func == (fptr_t)ync_file_range \|\| /
872	func == (fptr_t)cfa_sync_file_range )
873	#define _CFA_IO_FEATURE_IORING_OP_SYNC_FILE_RANGE ,
874	return IS_DEFINED(IORING_OP_SYNC_FILE_RANGE);
875
876	if( /func == (fptr_t)sendmsg \|\| /
877	func == (fptr_t)cfa_sendmsg )
878	#define _CFA_IO_FEATURE_IORING_OP_SENDMSG ,
879	return IS_DEFINED(IORING_OP_SENDMSG);
880
881	if( /func == (fptr_t)recvmsg \|\| /
882	func == (fptr_t)cfa_recvmsg )
883	#define _CFA_IO_FEATURE_IORING_OP_RECVMSG ,
884	return IS_DEFINED(IORING_OP_RECVMSG);
885
886	if( /func == (fptr_t)send \|\| /
887	func == (fptr_t)cfa_send )
888	#define _CFA_IO_FEATURE_IORING_OP_SEND ,
889	return IS_DEFINED(IORING_OP_SEND);
890
891	if( /func == (fptr_t)recv \|\| /
892	func == (fptr_t)cfa_recv )
893	#define _CFA_IO_FEATURE_IORING_OP_RECV ,
894	return IS_DEFINED(IORING_OP_RECV);
895
896	if( /func == (fptr_t)accept4 \|\| /
897	func == (fptr_t)cfa_accept4 )
898	#define _CFA_IO_FEATURE_IORING_OP_ACCEPT ,
899	return IS_DEFINED(IORING_OP_ACCEPT);
900
901	if( /func == (fptr_t)connect \|\| /
902	func == (fptr_t)cfa_connect )
903	#define _CFA_IO_FEATURE_IORING_OP_CONNECT ,
904	return IS_DEFINED(IORING_OP_CONNECT);
905
906	if( /func == (fptr_t)fallocate \|\| /
907	func == (fptr_t)cfa_fallocate )
908	#define _CFA_IO_FEATURE_IORING_OP_FALLOCATE ,
909	return IS_DEFINED(IORING_OP_FALLOCATE);
910
911	if( /func == (fptr_t)posix_fadvise \|\| /
912	func == (fptr_t)cfa_fadvise )
913	#define _CFA_IO_FEATURE_IORING_OP_FADVISE ,
914	return IS_DEFINED(IORING_OP_FADVISE);
915
916	if( /func == (fptr_t)madvise \|\| /
917	func == (fptr_t)cfa_madvise )
918	#define _CFA_IO_FEATURE_IORING_OP_MADVISE ,
919	return IS_DEFINED(IORING_OP_MADVISE);
920
921	if( /func == (fptr_t)openat \|\| /
922	func == (fptr_t)cfa_openat )
923	#define _CFA_IO_FEATURE_IORING_OP_OPENAT ,
924	return IS_DEFINED(IORING_OP_OPENAT);
925
926	if( /func == (fptr_t)close \|\| /
927	func == (fptr_t)cfa_close )
928	#define _CFA_IO_FEATURE_IORING_OP_CLOSE ,
929	return IS_DEFINED(IORING_OP_CLOSE);
930
931	if( /func == (fptr_t)statx \|\| /
932	func == (fptr_t)cfa_statx )
933	#define _CFA_IO_FEATURE_IORING_OP_STATX ,
934	return IS_DEFINED(IORING_OP_STATX);
935
936	if( /func == (fptr_t)read \|\| /
937	func == (fptr_t)cfa_read )
938	#define _CFA_IO_FEATURE_IORING_OP_READ ,
939	return IS_DEFINED(IORING_OP_READ);
940
941	if( /func == (fptr_t)write \|\| /
942	func == (fptr_t)cfa_write )
943	#define _CFA_IO_FEATURE_IORING_OP_WRITE ,
944	return IS_DEFINED(IORING_OP_WRITE);
945	#endif
946
947	return false;
948	}

Note: See TracBrowser for help on using the repository browser.

Download in other formats:

Original Format