[PATCH liburing] test/io_uring_passthrough: enhance vectored I/O test coverage

[PATCH liburing] test/io_uring_passthrough: enhance vectored I/O test coverage

[Anuj Gupta]

This patch improves the vectored io test coverage by ensuring that we
exercise all three kinds of iovec imports:
1. Single segment iovec
2. Multi segment iovec, below dynamic allocation threshold
3. Multi segment iovec, above dynamic allocation threshold

To support this we adjust the test logic to vary iovcnt appropriately
across submissions. For fixed vectored I/O support (case 2 and 3), we
register a single large buffer and slice it into vecs[]. This ensures
that all iovecs map to valid regions within the registered fixed buffer.
Additionally buffer allocation is adjusted accordingly, while
maintaining compatibility with existing non-vectored  tests.

Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Anuj Gupta <anuj20.g@samsung.com>
---
 test/io_uring_passthrough.c | 40 +++++++++++++++++++++++++++----------
 1 file changed, 30 insertions(+), 10 deletions(-)

diff --git a/test/io_uring_passthrough.c b/test/io_uring_passthrough.c
index 74bbfe0..5efe2a0 100644
--- a/test/io_uring_passthrough.c
+++ b/test/io_uring_passthrough.c
@@ -13,12 +13,14 @@
 #include "../src/syscall.h"
 #include "nvme.h"
 
+#define min(a, b)	((a) < (b) ? (a) : (b))
+
 #define FILE_SIZE	(256 * 1024)
 #define BS		8192
 #define BUFFERS		(FILE_SIZE / BS)
 
 static void *meta_mem;
-static struct iovec *vecs;
+static struct iovec *vecs, *backing_vec;
 static int no_pt;
 static bool vec_fixed_supported = true;
 
@@ -75,7 +77,7 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 	struct nvme_uring_cmd *cmd;
 	int open_flags;
 	int do_fixed;
-	int i, ret, fd = -1, use_fd = -1;
+	int i, ret, fd = -1, use_fd = -1, submit_count = 0;
 	off_t offset;
 	__u64 slba;
 	__u32 nlb;
@@ -86,7 +88,7 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 		open_flags = O_WRONLY;
 
 	if (fixed) {
-		ret = t_register_buffers(ring, vecs, BUFFERS);
+		ret = t_register_buffers(ring, backing_vec, 1);
 		if (ret == T_SETUP_SKIP)
 			return 0;
 		if (ret != T_SETUP_OK) {
@@ -116,6 +118,9 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 
 	offset = 0;
 	for (i = 0; i < BUFFERS; i++) {
+		unsigned int iovcnt = 1;
+		size_t total_len;
+
 		sqe = io_uring_get_sqe(ring);
 		if (!sqe) {
 			fprintf(stderr, "sqe get failed\n");
@@ -129,7 +134,7 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 		if (fixed && (i & 1))
 			do_fixed = 0;
 		if (do_fixed)
-			sqe->buf_index = i;
+			sqe->buf_index = 0;
 		if (nonvec)
 			sqe->cmd_op = NVME_URING_CMD_IO;
 		else
@@ -147,8 +152,14 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 
 		cmd->opcode = read ? nvme_cmd_read : nvme_cmd_write;
 
+		if (!nonvec) {
+			iovcnt = (submit_count % 3 == 0) ? 1 : ((submit_count % 3 == 1) ? 3 : 9);
+			iovcnt = min(iovcnt, BUFFERS - i);
+		}
+		total_len = BS * iovcnt;
+
 		slba = offset >> lba_shift;
-		nlb = (BS >> lba_shift) - 1;
+		nlb = (total_len >> lba_shift) - 1;
 
 		/* cdw10 and cdw11 represent starting lba */
 		cmd->cdw10 = slba & 0xffffffff;
@@ -160,7 +171,7 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 			cmd->data_len = vecs[i].iov_len;
 		} else {
 			cmd->addr = (__u64)(uintptr_t)&vecs[i];
-			cmd->data_len = 1;
+			cmd->data_len = iovcnt;
 		}
 
 		if (meta_size) {
@@ -170,16 +181,19 @@ static int __test_io(const char *file, struct io_uring *ring, int tc, int read,
 		}
 		cmd->nsid = nsid;
 
-		offset += BS;
+		offset += total_len;
+		if (!nonvec)
+			i += iovcnt - 1;
+		submit_count++;
 	}
 
 	ret = io_uring_submit(ring);
-	if (ret != BUFFERS) {
+	if (ret != submit_count) {
 		fprintf(stderr, "submit got %d, wanted %d\n", ret, BUFFERS);
 		goto err;
 	}
 
-	for (i = 0; i < BUFFERS; i++) {
+	for (i = 0; i < submit_count; i++) {
 		ret = io_uring_wait_cqe(ring, &cqe);
 		if (ret) {
 			fprintf(stderr, "wait_cqe=%d\n", ret);
@@ -439,7 +453,13 @@ int main(int argc, char *argv[])
 	if (ret)
 		return T_EXIT_SKIP;
 
-	vecs = t_create_buffers(BUFFERS, BS);
+	vecs = t_malloc(BUFFERS * sizeof(struct iovec));
+	backing_vec = t_create_buffers(1, BUFFERS * BS);
+	/* Slice single large backing_vec into multiple smaller vecs */
+	for (int i = 0; i < BUFFERS; i++) {
+		vecs[i].iov_base = backing_vec[0].iov_base + i * BS;
+		vecs[i].iov_len = BS;
+	}
 	if (meta_size)
 		t_posix_memalign(&meta_mem, 0x1000,
 				 meta_size * BUFFERS * (BS >> lba_shift));
-- 
2.25.1

Re: [PATCH liburing] test/io_uring_passthrough: enhance vectored I/O test coverage

[Jens Axboe]

On Thu, 22 May 2025 16:59:48 +0530, Anuj Gupta wrote:
> This patch improves the vectored io test coverage by ensuring that we
> exercise all three kinds of iovec imports:
> 1. Single segment iovec
> 2. Multi segment iovec, below dynamic allocation threshold
> 3. Multi segment iovec, above dynamic allocation threshold
> 
> To support this we adjust the test logic to vary iovcnt appropriately
> across submissions. For fixed vectored I/O support (case 2 and 3), we
> register a single large buffer and slice it into vecs[]. This ensures
> that all iovecs map to valid regions within the registered fixed buffer.
> Additionally buffer allocation is adjusted accordingly, while
> maintaining compatibility with existing non-vectored  tests.
> 
> [...]

Applied, thanks!

[1/1] test/io_uring_passthrough: enhance vectored I/O test coverage
      commit: 296dd55813bb91656752a91dd333825eb759b35e

Best regards,
-- 
Jens Axboe