[PATCH liburing v3] test: add test cases for hugepage registered buffers

[PATCH liburing v3] test: add test cases for hugepage registered buffers

[Chenliang Li]

Add a test file for hugepage registered buffers, to make sure the
fixed buffer coalescing feature works safe and soundly.

Testcases include read/write with single/multiple/unaligned/non-2MB
hugepage fixed buffers, and also a should-not coalesce case where
buffer is a mixture of different size'd pages.

Signed-off-by: Chenliang Li <cliang01.li@samsung.com>
---
Changes since v2:
- Return T_EXIT_SKIP for mmap/posix_memalign failures;
- Rebased to the newest commit;
- Code style issues.
v2: https://lore.kernel.org/io-uring/20240531014131.1441446-1-cliang01.li@samsung.com/T/#t

Changes since v1:
- Added unaligned/non-2MB hugepage/page mixture testcases.
- Rearranged the code.
v1: https://lore.kernel.org/io-uring/20240514051343.582556-1-cliang01.li@samsung.com/T/#u

 test/Makefile         |   1 +
 test/fixed-hugepage.c | 389 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 390 insertions(+)
 create mode 100644 test/fixed-hugepage.c

diff --git a/test/Makefile b/test/Makefile
index fcf6554..c40d590 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -94,6 +94,7 @@ test_srcs := \
 	file-verify.c \
 	fixed-buf-iter.c \
 	fixed-buf-merge.c \
+	fixed-hugepage.c \
 	fixed-link.c \
 	fixed-reuse.c \
 	fpos.c \
diff --git a/test/fixed-hugepage.c b/test/fixed-hugepage.c
new file mode 100644
index 0000000..396fe43
--- /dev/null
+++ b/test/fixed-hugepage.c
@@ -0,0 +1,389 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Test fixed buffers consisting of hugepages.
+ */
+#include <stdio.h>
+#include <string.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <sys/mman.h>
+#include <linux/mman.h>
+
+#include "liburing.h"
+#include "helpers.h"
+
+/*
+ * Before testing
+ * echo (>=4) > /proc/sys/vm/nr_hugepages
+ * echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
+ * echo always > /sys/kernel/mm/transparent_hugepage/hugepages-16kB/enabled
+ *
+ * Not 100% guaranteed to get THP-backed memory, but in general it does.
+ */
+#define MTHP_16KB	(16UL * 1024)
+#define HUGEPAGE_SIZE	(2UL * 1024 * 1024)
+#define NR_BUFS		1
+#define IN_FD		"/dev/urandom"
+#define OUT_FD		"/dev/zero"
+
+static int open_files(char *fname_in, int *fd_in, int *fd_out)
+{
+	*fd_in = open(fname_in, O_RDONLY, 0644);
+	if (*fd_in < 0) {
+		printf("open %s failed\n", fname_in);
+		return -1;
+	}
+
+	*fd_out = open(OUT_FD, O_RDWR, 0644);
+	if (*fd_out < 0) {
+		printf("open %s failed\n", OUT_FD);
+		return -1;
+	}
+
+	return 0;
+}
+
+static void unmap(struct iovec *iov, int nr_bufs, size_t offset)
+{
+	int i;
+
+	for (i = 0; i < nr_bufs; i++)
+		munmap(iov[i].iov_base - offset, iov[i].iov_len + offset);
+}
+
+static int mmap_hugebufs(struct iovec *iov, int nr_bufs, size_t buf_size, size_t offset)
+{
+	int i;
+
+	for (i = 0; i < nr_bufs; i++) {
+		void *base = NULL;
+
+		base = mmap(NULL, buf_size, PROT_READ | PROT_WRITE,
+				MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB, -1, 0);
+		if (base == MAP_FAILED) {
+			printf("Unable to map hugetlb page. Try increasing the "
+				"value in /proc/sys/vm/nr_hugepages\n");
+			unmap(iov, i, offset);
+			return -1;
+		}
+
+		memset(base, 0, buf_size);
+		iov[i].iov_base = base + offset;
+		iov[i].iov_len = buf_size - offset;
+	}
+
+	return 0;
+}
+
+/* map a hugepage and smaller page to a contiguous memory */
+static int mmap_mixture(struct iovec *iov, int nr_bufs, size_t buf_size)
+{
+	int i;
+	void *small_base = NULL, *huge_base = NULL, *start = NULL;
+	size_t small_size = buf_size - HUGEPAGE_SIZE;
+	size_t seg_size = ((buf_size / HUGEPAGE_SIZE) + 1) * HUGEPAGE_SIZE;
+
+	start = mmap(NULL, seg_size * nr_bufs, PROT_NONE,
+			MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0);
+	if (start == MAP_FAILED) {
+		printf("Unable to preserve the page mixture memory. "
+			"Try increasing the RLIMIT_MEMLOCK resource limit\n");
+		return -1;
+	}
+
+	for (i = 0; i < nr_bufs; i++) {
+		huge_base = mmap(start, HUGEPAGE_SIZE, PROT_READ | PROT_WRITE,
+				MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB | MAP_FIXED, -1, 0);
+		if (huge_base == MAP_FAILED) {
+			printf("Unable to map hugetlb page in the page mixture. "
+				"Try increasing the value in /proc/sys/vm/nr_hugepages\n");
+			unmap(iov, nr_bufs, 0);
+			return -1;
+		}
+
+		small_base = mmap(start + HUGEPAGE_SIZE, small_size, PROT_READ | PROT_WRITE,
+				MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
+		if (small_base == MAP_FAILED) {
+			printf("Unable to map small page in the page mixture. "
+				"Try increasing the RLIMIT_MEMLOCK resource limit\n");
+			unmap(iov, nr_bufs, 0);
+			return -1;
+		}
+
+		memset(huge_base, 0, buf_size);
+		iov[i].iov_base = huge_base;
+		iov[i].iov_len = buf_size;
+		start += seg_size;
+	}
+
+	return 0;
+}
+
+static void free_bufs(struct iovec *iov, int nr_bufs, size_t offset)
+{
+	int i;
+
+	for (i = 0; i < nr_bufs; i++)
+		free(iov[i].iov_base - offset);
+}
+
+static int get_mthp_bufs(struct iovec *iov, int nr_bufs, size_t buf_size,
+		size_t alignment, size_t offset)
+{
+	int i;
+
+	for (i = 0; i < nr_bufs; i++) {
+		void *base = NULL;
+
+		if (posix_memalign(&base, alignment, buf_size)) {
+			printf("Unable to allocate mthp pages. "
+				"Try increasing the RLIMIT_MEMLOCK resource limit\n");
+			free_bufs(iov, i, offset);
+			return -1;
+		}
+
+		memset(base, 0, buf_size);
+		iov[i].iov_base = base + offset;
+		iov[i].iov_len = buf_size - offset;
+	}
+
+	return 0;
+}
+
+static int do_read(struct io_uring *ring, int fd, struct iovec *iov, int nr_bufs)
+{
+	struct io_uring_sqe *sqe;
+	struct io_uring_cqe *cqe;
+	int i, ret;
+
+	for (i = 0; i < nr_bufs; i++) {
+		sqe = io_uring_get_sqe(ring);
+		if (!sqe) {
+			fprintf(stderr, "Could not get SQE.\n");
+			return -1;
+		}
+
+		io_uring_prep_read_fixed(sqe, fd, iov[i].iov_base, iov[i].iov_len, 0, i);
+		io_uring_submit(ring);
+
+		ret = io_uring_wait_cqe(ring, &cqe);
+		if (ret < 0) {
+			fprintf(stderr, "Error waiting for completion: %s\n", strerror(-ret));
+			return -1;
+		}
+
+		if (cqe->res < 0) {
+			fprintf(stderr, "Error in async read operation: %s\n", strerror(-cqe->res));
+			return -1;
+		}
+		if (cqe->res != iov[i].iov_len) {
+			fprintf(stderr, "cqe res: %d, expected: %lu\n", cqe->res, iov[i].iov_len);
+			return -1;
+		}
+
+		io_uring_cqe_seen(ring, cqe);
+	}
+
+	return 0;
+}
+
+static int do_write(struct io_uring *ring, int fd, struct iovec *iov, int nr_bufs)
+{
+	struct io_uring_sqe *sqe;
+	struct io_uring_cqe *cqe;
+	int i, ret;
+
+	for (i = 0; i < nr_bufs; i++) {
+		sqe = io_uring_get_sqe(ring);
+		if (!sqe) {
+			fprintf(stderr, "Could not get SQE.\n");
+			return -1;
+		}
+
+		io_uring_prep_write_fixed(sqe, fd, iov[i].iov_base, iov[i].iov_len, 0, i);
+		io_uring_submit(ring);
+
+		ret = io_uring_wait_cqe(ring, &cqe);
+		if (ret < 0) {
+			fprintf(stderr, "Error waiting for completion: %s\n", strerror(-ret));
+			return -1;
+		}
+
+		if (cqe->res < 0) {
+			fprintf(stderr, "Error in async write operation: %s\n", strerror(-cqe->res));
+			return -1;
+		}
+		if (cqe->res != iov[i].iov_len) {
+			fprintf(stderr, "cqe res: %d, expected: %lu\n", cqe->res, iov[i].iov_len);
+			return -1;
+		}
+
+		io_uring_cqe_seen(ring, cqe);
+	}
+
+	return 0;
+}
+
+static int register_submit(struct io_uring *ring, struct iovec *iov,
+						int nr_bufs, int fd_in, int fd_out)
+{
+	int ret;
+
+	ret = io_uring_register_buffers(ring, iov, nr_bufs);
+	if (ret) {
+		fprintf(stderr, "Error registering buffers: %s\n", strerror(-ret));
+		return ret;
+	}
+
+	ret = do_read(ring, fd_in, iov, nr_bufs);
+	if (ret) {
+		fprintf(stderr, "Read test failed\n");
+		return ret;
+	}
+
+	ret = do_write(ring, fd_out, iov, nr_bufs);
+	if (ret) {
+		fprintf(stderr, "Write test failed\n");
+		return ret;
+	}
+
+	ret = io_uring_unregister_buffers(ring);
+	if (ret) {
+		fprintf(stderr, "Error unregistering buffers for one hugepage test: %s", strerror(-ret));
+		return ret;
+	}
+
+	return 0;
+}
+
+static int test_one_hugepage(struct io_uring *ring, int fd_in, int fd_out)
+{
+	struct iovec iov[NR_BUFS];
+	size_t buf_size = HUGEPAGE_SIZE;
+	int ret;
+
+	if (mmap_hugebufs(iov, NR_BUFS, buf_size, 0))
+		return T_EXIT_SKIP;
+
+	ret = register_submit(ring, iov, NR_BUFS, fd_in, fd_out);
+	unmap(iov, NR_BUFS, 0);
+	return ret ? T_EXIT_FAIL : T_EXIT_PASS;
+}
+
+static int test_multi_hugepages(struct io_uring *ring, int fd_in, int fd_out)
+{
+	struct iovec iov[NR_BUFS];
+	size_t buf_size = 4 * HUGEPAGE_SIZE;
+	int ret;
+
+	if (mmap_hugebufs(iov, NR_BUFS, buf_size, 0))
+		return T_EXIT_SKIP;
+
+	ret = register_submit(ring, iov, NR_BUFS, fd_in, fd_out);
+	unmap(iov, NR_BUFS, 0);
+	return ret ? T_EXIT_FAIL : T_EXIT_PASS;
+}
+
+static int test_unaligned_hugepage(struct io_uring *ring, int fd_in, int fd_out)
+{
+	struct iovec iov[NR_BUFS];
+	size_t buf_size = 3 * HUGEPAGE_SIZE;
+	size_t offset = 0x1234;
+	int ret;
+
+	if (mmap_hugebufs(iov, NR_BUFS, buf_size, offset))
+		return T_EXIT_SKIP;
+
+	ret = register_submit(ring, iov, NR_BUFS, fd_in, fd_out);
+	unmap(iov, NR_BUFS, offset);
+	return ret ? T_EXIT_FAIL : T_EXIT_PASS;
+}
+
+static int test_multi_unaligned_mthps(struct io_uring *ring, int fd_in, int fd_out)
+{
+	struct iovec iov[NR_BUFS];
+	int ret;
+	size_t buf_size = 3 * MTHP_16KB;
+	size_t offset = 0x1234;
+
+	if (get_mthp_bufs(iov, NR_BUFS, buf_size, MTHP_16KB, offset))
+		return T_EXIT_SKIP;
+
+	ret = register_submit(ring, iov, NR_BUFS, fd_in, fd_out);
+	free_bufs(iov, NR_BUFS, offset);
+	return ret ? T_EXIT_FAIL : T_EXIT_PASS;
+}
+
+/* Should not coalesce */
+static int test_page_mixture(struct io_uring *ring, int fd_in, int fd_out)
+{
+	struct iovec iov[NR_BUFS];
+	size_t buf_size = HUGEPAGE_SIZE + MTHP_16KB;
+	int ret;
+
+	if (mmap_mixture(iov, NR_BUFS, buf_size))
+		return T_EXIT_SKIP;
+
+	ret = register_submit(ring, iov, NR_BUFS, fd_in, fd_out);
+	unmap(iov, NR_BUFS, 0);
+	return ret ? T_EXIT_FAIL : T_EXIT_PASS;
+}
+
+int main(int argc, char *argv[])
+{
+	struct io_uring ring;
+	int ret, fd_in, fd_out;
+	char *fname_in;
+
+	if (argc > 1)
+		fname_in = argv[1];
+	else
+		fname_in = IN_FD;
+
+	if (open_files(fname_in, &fd_in, &fd_out))
+		return T_EXIT_SKIP;
+
+	ret = t_create_ring(8, &ring, 0);
+	if (ret == T_SETUP_SKIP)
+		return T_EXIT_SKIP;
+	else if (ret < 0)
+		return T_EXIT_FAIL;
+
+	ret = test_one_hugepage(&ring, fd_in, fd_out);
+	if (ret != T_EXIT_PASS) {
+		if (ret != T_EXIT_SKIP)
+			fprintf(stderr, "Test one hugepage failed.\n");
+		return ret;
+	}
+
+	ret = test_multi_hugepages(&ring, fd_in, fd_out);
+	if (ret != T_EXIT_PASS) {
+		if (ret != T_EXIT_SKIP)
+			fprintf(stderr, "Test multi hugepages failed.\n");
+		return ret;
+	}
+
+	ret = test_unaligned_hugepage(&ring, fd_in, fd_out);
+	if (ret != T_EXIT_PASS) {
+		if (ret != T_EXIT_SKIP)
+			fprintf(stderr, "Test unaligned hugepage failed.\n");
+		return ret;
+	}
+
+	ret = test_multi_unaligned_mthps(&ring, fd_in, fd_out);
+	if (ret != T_EXIT_PASS) {
+		if (ret != T_EXIT_SKIP)
+			fprintf(stderr, "Test unaligned multi-size'd THPs failed.\n");
+		return ret;
+	}
+
+	ret = test_page_mixture(&ring, fd_in, fd_out);
+	if (ret != T_EXIT_PASS) {
+		if (ret != T_EXIT_SKIP)
+			fprintf(stderr, "Test huge small page mixture failed.\n");
+		return ret;
+	}
+
+	io_uring_queue_exit(&ring);
+	return T_EXIT_PASS;
+}

base-commit: 3df984a28a28014a060157d90f3e46e657c04314
-- 
2.34.1

Re: [PATCH liburing v3] test: add test cases for hugepage registered buffers

[Pavel Begunkov]

On 5/31/24 06:20, Chenliang Li wrote:
> Add a test file for hugepage registered buffers, to make sure the
> fixed buffer coalescing feature works safe and soundly.
> 
> Testcases include read/write with single/multiple/unaligned/non-2MB
> hugepage fixed buffers, and also a should-not coalesce case where
> buffer is a mixture of different size'd pages.

lgtm, would be even better if you can add another patch
testing adding a small buffer on the left size of a hugepage,
i.e. like mmap_mixutre() but the small buffer is on the other
side.

-- 
Pavel Begunkov

Re: [PATCH liburing v3] test: add test cases for hugepage registered buffers

[Chenliang Li]

On Thu, 1 Aug 2024 00:13:10 +0100, Pavel Begunkov wrote:
> On 5/31/24 06:20, Chenliang Li wrote:
>> Add a test file for hugepage registered buffers, to make sure the
>> fixed buffer coalescing feature works safe and soundly.
>> 
>> Testcases include read/write with single/multiple/unaligned/non-2MB
>> hugepage fixed buffers, and also a should-not coalesce case where
>> buffer is a mixture of different size'd pages.
>
> lgtm, would be even better if you can add another patch
> testing adding a small buffer on the left size of a hugepage,
> i.e. like mmap_mixutre() but the small buffer is on the other
> side.

Sure, will add that.

Thanks,
Chenliang Li

Re: [PATCH liburing v3] test: add test cases for hugepage registered buffers

[Jens Axboe]

On 7/31/24 7:01 PM, Chenliang Li wrote:
> On Thu, 1 Aug 2024 00:13:10 +0100, Pavel Begunkov wrote:
>> On 5/31/24 06:20, Chenliang Li wrote:
>>> Add a test file for hugepage registered buffers, to make sure the
>>> fixed buffer coalescing feature works safe and soundly.
>>>
>>> Testcases include read/write with single/multiple/unaligned/non-2MB
>>> hugepage fixed buffers, and also a should-not coalesce case where
>>> buffer is a mixture of different size'd pages.
>>
>> lgtm, would be even better if you can add another patch
>> testing adding a small buffer on the left size of a hugepage,
>> i.e. like mmap_mixutre() but the small buffer is on the other
>> side.
> 
> Sure, will add that.

I'll queue up this v3, but please do send that as an incremental
patch as I agree it's a good addition.

-- 
Jens Axboe

Re: [PATCH liburing v3] test: add test cases for hugepage registered buffers

[Pavel Begunkov]

On 8/1/24 21:51, Jens Axboe wrote:
> On 7/31/24 7:01 PM, Chenliang Li wrote:
>> On Thu, 1 Aug 2024 00:13:10 +0100, Pavel Begunkov wrote:
>>> On 5/31/24 06:20, Chenliang Li wrote:
>>>> Add a test file for hugepage registered buffers, to make sure the
>>>> fixed buffer coalescing feature works safe and soundly.
>>>>
>>>> Testcases include read/write with single/multiple/unaligned/non-2MB
>>>> hugepage fixed buffers, and also a should-not coalesce case where
>>>> buffer is a mixture of different size'd pages.
>>>
>>> lgtm, would be even better if you can add another patch
>>> testing adding a small buffer on the left size of a hugepage,
>>> i.e. like mmap_mixutre() but the small buffer is on the other
>>> side.
>>
>> Sure, will add that.
> 
> I'll queue up this v3, but please do send that as an incremental
> patch as I agree it's a good addition

I have to add that tests are already well made, the kernel code
is not symmetric however, so it'd give us more confidence and
help to not regress in the future.

-- 
Pavel Begunkov

Re: [PATCH liburing v3] test: add test cases for hugepage registered buffers

[Jens Axboe]

On Fri, 31 May 2024 13:20:23 +0800, Chenliang Li wrote:
> Add a test file for hugepage registered buffers, to make sure the
> fixed buffer coalescing feature works safe and soundly.
> 
> Testcases include read/write with single/multiple/unaligned/non-2MB
> hugepage fixed buffers, and also a should-not coalesce case where
> buffer is a mixture of different size'd pages.
> 
> [...]

Applied, thanks!

[1/1] test: add test cases for hugepage registered buffers
      (no commit info)

Best regards,
-- 
Jens Axboe