fsstress,fsx: add io_uring test and do some fix

187 lines not on this page · raw message

index a4188e1c..0e7be6bb 100644
@@ -28,6 +28,7 @@
 #endif
 #ifdef AIO
 #include <libaio.h>
+#define AIO_ENTRIES	1
 io_context_t	io_ctx;
 #endif
 #ifdef URING

@@ -699,8 +700,8 @@ int main(int argc, char **argv)
 			}
 			procid = i;
 #ifdef AIO
-			if (io_setup(128, &io_ctx) != 0) {
-				fprintf(stderr, "io_setup failed");
+			if (io_setup(AIO_ENTRIES, &io_ctx) != 0) {
+				fprintf(stderr, "io_setup failed\n");
 				exit(1);
 			}
 #endif

index 0e7be6bb..4b255bf6 100644
@@ -2099,8 +2099,7 @@ do_aio_rw(int opno, long r, int flags)
 	if (!get_fname(FT_REGFILE, r, &f, NULL, NULL, &v)) {
 		if (v)
 			printf("%d/%d: do_aio_rw - no filename\n", procid, opno);
-		free_pathname(&f);
-		return;
+		goto aio_out3;
 	}
 	fd = open_path(&f, flags|O_DIRECT);
 	e = fd < 0 ? errno : 0;

@@ -2109,16 +2108,13 @@ do_aio_rw(int opno, long r, int flags)
 		if (v)
 			printf("%d/%d: do_aio_rw - open %s failed %d\n",
 			       procid, opno, f.path, e);
-		free_pathname(&f);
-		return;
+		goto aio_out3;
 	}
 	if (fstat64(fd, &stb) < 0) {
 		if (v)
 			printf("%d/%d: do_aio_rw - fstat64 %s failed %d\n",
 			       procid, opno, f.path, errno);
-		free_pathname(&f);
-		close(fd);
-		return;
+		goto aio_out2;
 	}
 	inode_info(st, sizeof(st), &stb, v);
 	if (!iswrite && stb.st_size == 0) {

@@ -2150,6 +2146,12 @@ do_aio_rw(int opno, long r, int flags)
 	else if (len > diob.d_maxiosz)
 		len = diob.d_maxiosz;
 	buf = memalign(diob.d_mem, len);
+	if (!buf) {
+		if (v)
+			printf("%d/%d: do_aio_rw - memalign failed\n",
+			       procid, opno);
+		goto aio_out2;
+	}
 
 	if (iswrite) {
 		off = (off64_t)(lr % MIN(stb.st_size + (1024 * 1024), MAXFSIZE));

@@ -2166,27 +2168,26 @@ do_aio_rw(int opno, long r, int flags)
 		if (v)
 			printf("%d/%d: %s - io_submit failed %d\n",
 			       procid, opno, iswrite ? "awrite" : "aread", e);
-		free_pathname(&f);
-		close(fd);
-		return;
+		goto aio_out1;
 	}
 	if ((e = io_getevents(io_ctx, 1, 1, &event, NULL)) != 1) {
 		if (v)
 			printf("%d/%d: %s - io_getevents failed %d\n",
 			       procid, opno, iswrite ? "awrite" : "aread", e);
-		free_pathname(&f);
-		close(fd);
-		return;
+		goto aio_out1;
 	}
 
 	e = event.res != len ? event.res2 : 0;
-	free(buf);
 	if (v)
 		printf("%d/%d: %s %s%s [%lld,%d] %d\n",
 		       procid, opno, iswrite ? "awrite" : "aread",
 		       f.path, st, (long long)off, (int)len, e);
-	free_pathname(&f);
+ aio_out1:
+	free(buf);
+ aio_out2:
 	close(fd);
+ aio_out3:
+	free_pathname(&f);
 }
 #endif
 

index 7c76655a..7475bf9d 100644
@@ -34,6 +34,9 @@
 #ifdef AIO
 #include <libaio.h>
 #endif
+#ifdef URING
+#include <liburing.h>
+#endif
 #include <sys/syscall.h>
 
 #ifndef MAP_FILE

@@ -176,21 +179,17 @@ int	integrity = 0;			/* -i flag */
 int	fsxgoodfd = 0;
 int	o_direct;			/* -Z */
 int	aio = 0;
+int	uring = 0;
 int	mark_nr = 0;
 
 int page_size;
 int page_mask;
 int mmap_mask;
-#ifdef AIO
-int aio_rw(int rw, int fd, char *buf, unsigned len, unsigned offset);
+int fsx_rw(int rw, int fd, char *buf, unsigned len, unsigned offset);
 #define READ 0
 #define WRITE 1
-#define fsxread(a,b,c,d)	aio_rw(READ, a,b,c,d)
-#define fsxwrite(a,b,c,d)	aio_rw(WRITE, a,b,c,d)
-#else
-#define fsxread(a,b,c,d)	read(a,b,c)
-#define fsxwrite(a,b,c,d)	write(a,b,c)
-#endif
+#define fsxread(a,b,c,d)	fsx_rw(READ, a,b,c,d)
+#define fsxwrite(a,b,c,d)	fsx_rw(WRITE, a,b,c,d)
 
 const char *replayops = NULL;
 const char *recordops = NULL;

@@ -2242,7 +2241,7 @@ void
 usage(void)
 {
 	fprintf(stdout, "usage: %s",
-		"fsx [-dknqxABEFJLOWZ] [-b opnum] [-c Prob] [-g filldata] [-i logdev] [-j logid] [-l flen] [-m start:end] [-o oplen] [-p progressinterval] [-r readbdy] [-s style] [-t truncbdy] [-w writebdy] [-D startingop] [-N numops] [-P dirpath] [-S seed] fname\n\
+		"fsx [-dknqxBEFJLOWZ][-A|-U] [-b opnum] [-c Prob] [-g filldata] [-i logdev] [-j logid] [-l flen] [-m start:end] [-o oplen] [-p progressinterval] [-r readbdy] [-s style] [-t truncbdy] [-w writebdy] [-D startingop] [-N numops] [-P dirpath] [-S seed] fname\n\
 	-b opnum: beginning operation number (default 1)\n\
 	-c P: 1 in P chance of file close+open at each op (default infinity)\n\
 	-d: debug output for all operations\n\

@@ -2265,7 +2264,10 @@ usage(void)
 	-y synchronize changes to a file\n"
 
 #ifdef AIO
-"	-A: Use the AIO system calls\n"
+"	-A: Use the AIO system calls, -A excludes -U\n"
+#endif
+#ifdef URING
+"	-U: Use the IO_URING system calls, -U excludes -A\n"
 #endif
 "	-D startingop: debug output starting at specified operation\n"
 #ifdef HAVE_LINUX_FALLOC_H

@@ -2425,13 +2427,131 @@ out_error:
 	errno = -ret;
 	return -1;
 }
+#endif
+
+#ifdef URING
+struct io_uring ring;
+#define URING_ENTRIES	1024
+int
+uring_setup()
+{
+	int ret;
+
+	ret = io_uring_queue_init(URING_ENTRIES, &ring, 0);
+	if (ret != 0) {
+		fprintf(stderr, "uring_setup: io_uring_queue_init failed: %s\n",
+                        strerror(ret));
+                return -1;
+        }
+        return 0;
+}
 
-int aio_rw(int rw, int fd, char *buf, unsigned len, unsigned offset)
+int
+__uring_rw(int rw, int fd, char *buf, unsigned len, unsigned offset)
 {
+	struct io_uring_sqe	*sqe;
+	struct io_uring_cqe	*cqe;
+	struct iovec		iovec;
 	int ret;
+	int res, res2 = 0;
+	char *p = buf;
+	unsigned l = len;
+	unsigned o = offset;
+
+
+	/*
+	 * Due to io_uring tries non-blocking IOs (especially read), that
+	 * always cause 'normal' short reading. To avoid this short read
+	 * fail, try to loop read/write (escpecilly read) data.
+	 */
+ uring_loop:
+	sqe = io_uring_get_sqe(&ring);
+	if (!sqe) {
+		fprintf(stderr, "uring_rw: io_uring_get_sqe failed: %s\n",
+		        strerror(errno));
+		return -1;
+        }
+
+	iovec.iov_base = p;
+	iovec.iov_len = l;
+	if (rw == READ) {
+		io_uring_prep_readv(sqe, fd, &iovec, 1, o);
+	} else {
+		io_uring_prep_writev(sqe, fd, &iovec, 1, o);
+	}
+
+	ret = io_uring_submit(&ring);
+	if (ret != 1) {
+		fprintf(stderr, "errcode=%d\n", -ret);
+		fprintf(stderr, "uring %s: io_uring_submit failed: %s\n",
+		        rw == READ ? "read":"write", strerror(-ret));
+		goto uring_error;
+	}
+
+	ret = io_uring_wait_cqe(&ring, &cqe);
+	if (ret < 0) {
+		if (ret == 0)
+			fprintf(stderr, "uring %s: no events available\n",
+			        rw == READ ? "read":"write");
+		else {
+			fprintf(stderr, "errcode=%d\n", -ret);
+			fprintf(stderr, "uring %s: io_uring_wait_cqe failed: %s\n",
+			        rw == READ ? "read":"write", strerror(-ret));
+		}
+		goto uring_error;
+	}
+	res = cqe->res;
+	io_uring_cqe_seen(&ring, cqe);
+
+	res2 += res;
+	if (len != res2) {
+		if (res > 0) {
+			o += res;
+			l -= res;
+			p += res;
+			if (l > 0)
+				goto uring_loop;
+		} else if (res < 0) {
+			ret = res;
+			fprintf(stderr, "errcode=%d\n", -ret);
+			fprintf(stderr, "uring %s: io_uring failed: %s\n",
+			        rw == READ ? "read":"write", strerror(-ret));
+			goto uring_error;
+		} else {
+			fprintf(stderr, "uring %s bad io length: %d instead of %u\n",
+			        rw == READ ? "read":"write", res2, len);
+		}
+	}
+	return res2;
+
+ uring_error:
+	/*
+	 * The caller expects error return in traditional libc
+	 * convention, i.e. -1 and the errno set to error.
+	 */
+	errno = -ret;
+	return -1;
+}
+#endif
+
+int fsx_rw(int rw, int fd, char *buf, unsigned len, unsigned offset)
+{
+	int ret = -1;
 
 	if (aio) {
+#ifdef AIO
 		ret = __aio_rw(rw, fd, buf, len, offset);
+#elif
+		fprintf(stderr, "io_rw: need AIO support!\n");
+		exit(111);
+#endif
+	} else if (uring) {
+#ifdef URING
+		ret = __uring_rw(rw, fd, buf, len, offset);
+#elif
+		fprintf(stderr, "io_rw: need IO_URING support!\n");
+		exit(111);
+#endif
 	} else {
 		if (rw == READ)
 			ret = read(fd, buf, len);

@@ -2441,8 +2561,6 @@ int aio_rw(int rw, int fd, char *buf, unsigned len, unsigned offset)
 	return ret;
 }
 
-#endif
-
 #define test_fallocate(mode) __test_fallocate(mode, #mode)
 
 int

@@ -2496,7 +2614,7 @@ main(int argc, char **argv)
 	setvbuf(stdout, (char *)0, _IOLBF, 0); /* line buffered stdout */
 
 	while ((ch = getopt_long(argc, argv,
-				 "b:c:dfg:i:j:kl:m:no:p:qr:s:t:w:xyABD:EFJKHzCILN:OP:RS:WXZ",
+				 "b:c:dfg:i:j:kl:m:no:p:qr:s:t:w:xyABD:EFJKHzCILN:OP:RS:UWXZ",
 				 longopts, NULL)) != EOF)
 		switch (ch) {
 		case 'b':

@@ -2604,6 +2722,9 @@ main(int argc, char **argv)
 		case 'A':
 		        aio = 1;
 			break;
+		case 'U':
+		        uring = 1;
+			break;
 		case 'D':
 			debugstart = getnum(optarg, &endp);
 			if (debugstart < 1)

@@ -2694,6 +2815,11 @@ main(int argc, char **argv)
 	if (argc != 1)
 		usage();
 
+	if (aio && uring) {
+		fprintf(stderr, "-A and -U shouldn't be used together\n");
+		usage();
+	}
+
 	if (integrity && !dirpath) {
 		fprintf(stderr, "option -i <logdev> requires -P <dirpath>\n");
 		usage();

@@ -2784,6 +2910,10 @@ main(int argc, char **argv)
 	if (aio) 
 		aio_setup();
 #endif
+#ifdef URING
+	if (uring)
+		uring_setup();
+#endif
 
 	if (!(o_flags & O_TRUNC)) {
 		off_t ret;