[PATCH 00/10] Limit usage of internal API in tests

[PATCH 00/10] Limit usage of internal API in tests

[David Marchand]

This series is aimed at 26.11.

We had a few bug reports related to internal (and experimental) symbols
issues during 26.07 development.
See for example https://bugs.dpdk.org/show_bug.cgi?id=1957 or more
recently https://bugs.dpdk.org/show_bug.cgi?id=1967.

To catch such issues earlier in the CI, this series proposes to run
the unit tests through meson with the ABI reference unit test binary
against the current ABI libraries and drivers.

For this to work, some unit tests must be skipped (since meson may
invoke the ABI reference code with tests that were unknown at the time).

A few unit tests were directly dereferencing internal structures and are
reworked so they use public APIs.

Additionally, unit tests were allowed to use any internal API which has
hidden a few issues (like a public API backed by internal symbols in the
hash library).
So disable the global ALLLOW_INTERNAL_API and move it to code explicitly
requiring internal API, with the hope it will push us to have better API.

The last patch contains a hack for now, where the cached ABI reference
is force regenerated with a branch of mine where I backported the
fixes of this series.


-- 
David Marchand

David Marchand (10):
  bbdev: fix stats aggregation from queues
  bbdev: add per-queue statistics API
  hash: fix GFNI stubs export
  test: uninline helper for forking
  test/bonding: get MAC address with public API
  test/devargs: check driver presence with public API
  test/vdev: find device with public API
  test: limit internal API usage
  ci: make ABI reference generation faster
  ci: run reference unit tests against ABI

 .ci/linux-build.sh                     |  35 +++-
 .github/workflows/build.yml            |  15 +-
 MAINTAINERS                            |   1 +
 app/test-bbdev/test_bbdev_perf.c       |  38 +---
 app/test/meson.build                   |   7 +-
 app/test/process.c                     | 231 +++++++++++++++++++++++++
 app/test/process.h                     | 228 +-----------------------
 app/test/test_devargs.c                |  12 +-
 app/test/test_external_mem.c           |   2 +
 app/test/test_link_bonding.c           |  35 ++--
 app/test/test_malloc.c                 |   2 +
 app/test/test_mempool.c                |   2 +
 app/test/test_pdump.c                  |   3 +
 app/test/test_pdump.h                  |   3 +
 app/test/test_vdev.c                   |  85 ++-------
 app/test/virtual_pmd.c                 |   2 +
 devtools/test-meson-builds.sh          |   5 +-
 doc/guides/rel_notes/release_26_07.rst |   5 +
 lib/bbdev/rte_bbdev.c                  |  26 ++-
 lib/bbdev/rte_bbdev.h                  |  20 +++
 lib/hash/rte_thash_gfni.c              |   4 +-
 lib/hash/rte_thash_gfni.h              |   2 -
 22 files changed, 405 insertions(+), 358 deletions(-)
 create mode 100644 app/test/process.c

-- 
2.54.0

[PATCH 01/10] bbdev: fix stats aggregation from queues

[David Marchand]

The device stats retrieval was missing aggregation of
enqueue_status_count, acc_offload_cycles, and enqueue_depth_avail
from per-queue statistics.

Fixes: 4f08028c5e24 ("bbdev: expose queue related warning and status")
Cc: stable@dpdk.org

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 lib/bbdev/rte_bbdev.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/lib/bbdev/rte_bbdev.c b/lib/bbdev/rte_bbdev.c
index ea644c1f9b..90095578cb 100644
--- a/lib/bbdev/rte_bbdev.c
+++ b/lib/bbdev/rte_bbdev.c
@@ -745,7 +745,7 @@ rte_bbdev_queue_stop(uint16_t dev_id, uint16_t queue_id)
 static void
 get_stats_from_queues(struct rte_bbdev *dev, struct rte_bbdev_stats *stats)
 {
-	unsigned int q_id;
+	unsigned int i, q_id;
 	for (q_id = 0; q_id < dev->data->num_queues; q_id++) {
 		struct rte_bbdev_stats *q_stats =
 				&dev->data->queues[q_id].queue_stats;
@@ -756,6 +756,10 @@ get_stats_from_queues(struct rte_bbdev *dev, struct rte_bbdev_stats *stats)
 		stats->dequeue_err_count += q_stats->dequeue_err_count;
 		stats->enqueue_warn_count += q_stats->enqueue_warn_count;
 		stats->dequeue_warn_count += q_stats->dequeue_warn_count;
+		for (i = 0; i < RTE_BBDEV_ENQ_STATUS_SIZE_MAX; i++)
+			stats->enqueue_status_count[i] += q_stats->enqueue_status_count[i];
+		stats->acc_offload_cycles += q_stats->acc_offload_cycles;
+		stats->enqueue_depth_avail += q_stats->enqueue_depth_avail;
 	}
 	rte_bbdev_log_debug("Got stats on %u", dev->data->dev_id);
 }
-- 
2.54.0

[PATCH 02/10] bbdev: add per-queue statistics API

[David Marchand]

Add rte_bbdev_queue_stats_get() to retrieve statistics for a specific
queue. This allows applications to get per-queue stats without
accessing internal data structures.

Update test-bbdev to use the new API.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 app/test-bbdev/test_bbdev_perf.c       | 38 +++++---------------------
 doc/guides/rel_notes/release_26_07.rst |  5 ++++
 lib/bbdev/rte_bbdev.c                  | 20 ++++++++++++++
 lib/bbdev/rte_bbdev.h                  | 20 ++++++++++++++
 4 files changed, 52 insertions(+), 31 deletions(-)

diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c
index 9fb0a25948..b75d8387bd 100644
--- a/app/test-bbdev/test_bbdev_perf.c
+++ b/app/test-bbdev/test_bbdev_perf.c
@@ -5849,30 +5849,6 @@ validation_test(struct active_device *ad, struct test_op_params *op_params)
 	return validation_latency_test(ad, op_params, false);
 }
 
-static int
-get_bbdev_queue_stats(uint16_t dev_id, uint16_t queue_id,
-		struct rte_bbdev_stats *stats)
-{
-	struct rte_bbdev *dev = &rte_bbdev_devices[dev_id];
-	struct rte_bbdev_stats *q_stats;
-
-	if (queue_id >= dev->data->num_queues)
-		return -1;
-
-	q_stats = &dev->data->queues[queue_id].queue_stats;
-
-	stats->enqueued_count = q_stats->enqueued_count;
-	stats->dequeued_count = q_stats->dequeued_count;
-	stats->enqueue_err_count = q_stats->enqueue_err_count;
-	stats->dequeue_err_count = q_stats->dequeue_err_count;
-	stats->enqueue_warn_count = q_stats->enqueue_warn_count;
-	stats->dequeue_warn_count = q_stats->dequeue_warn_count;
-	stats->acc_offload_cycles = q_stats->acc_offload_cycles;
-	stats->enqueue_depth_avail = q_stats->enqueue_depth_avail;
-
-	return 0;
-}
-
 static int
 offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs,
 		struct rte_bbdev_fft_op *ref_op, uint16_t dev_id,
@@ -5906,7 +5882,7 @@ offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs,
 					&ops_enq[enq], burst_sz - enq);
 		} while (unlikely(burst_sz != enq));
 
-		ret = get_bbdev_queue_stats(dev_id, queue_id, &stats);
+		ret = rte_bbdev_queue_stats_get(dev_id, queue_id, &stats);
 		TEST_ASSERT_SUCCESS(ret,
 				"Failed to get stats for queue (%u) of device (%u)",
 				queue_id, dev_id);
@@ -5988,7 +5964,7 @@ offload_latency_test_mldts(struct rte_mempool *mempool, struct test_buffers *buf
 					&ops_enq[enq], burst_sz - enq);
 		} while (unlikely(burst_sz != enq));
 
-		ret = get_bbdev_queue_stats(dev_id, queue_id, &stats);
+		ret = rte_bbdev_queue_stats_get(dev_id, queue_id, &stats);
 		TEST_ASSERT_SUCCESS(ret,
 				"Failed to get stats for queue (%u) of device (%u)",
 				queue_id, dev_id);
@@ -6069,7 +6045,7 @@ offload_latency_test_dec(struct rte_mempool *mempool, struct test_buffers *bufs,
 					&ops_enq[enq], burst_sz - enq);
 		} while (unlikely(burst_sz != enq));
 
-		ret = get_bbdev_queue_stats(dev_id, queue_id, &stats);
+		ret = rte_bbdev_queue_stats_get(dev_id, queue_id, &stats);
 		TEST_ASSERT_SUCCESS(ret,
 				"Failed to get stats for queue (%u) of device (%u)",
 				queue_id, dev_id);
@@ -6163,7 +6139,7 @@ offload_latency_test_ldpc_dec(struct rte_mempool *mempool,
 		} while (unlikely(burst_sz != enq));
 
 		enq_sw_last_time = rte_rdtsc_precise() - enq_start_time;
-		ret = get_bbdev_queue_stats(dev_id, queue_id, &stats);
+		ret = rte_bbdev_queue_stats_get(dev_id, queue_id, &stats);
 		TEST_ASSERT_SUCCESS(ret,
 				"Failed to get stats for queue (%u) of device (%u)",
 				queue_id, dev_id);
@@ -6251,7 +6227,7 @@ offload_latency_test_enc(struct rte_mempool *mempool, struct test_buffers *bufs,
 
 		enq_sw_last_time = rte_rdtsc_precise() - enq_start_time;
 
-		ret = get_bbdev_queue_stats(dev_id, queue_id, &stats);
+		ret = rte_bbdev_queue_stats_get(dev_id, queue_id, &stats);
 		TEST_ASSERT_SUCCESS(ret,
 				"Failed to get stats for queue (%u) of device (%u)",
 				queue_id, dev_id);
@@ -6332,7 +6308,7 @@ offload_latency_test_ldpc_enc(struct rte_mempool *mempool,
 		} while (unlikely(burst_sz != enq));
 
 		enq_sw_last_time = rte_rdtsc_precise() - enq_start_time;
-		ret = get_bbdev_queue_stats(dev_id, queue_id, &stats);
+		ret = rte_bbdev_queue_stats_get(dev_id, queue_id, &stats);
 		TEST_ASSERT_SUCCESS(ret,
 				"Failed to get stats for queue (%u) of device (%u)",
 				queue_id, dev_id);
@@ -6482,7 +6458,7 @@ offload_cost_test(struct active_device *ad,
 			rte_get_tsc_hz());
 
 	struct rte_bbdev_stats stats = {0};
-	ret = get_bbdev_queue_stats(ad->dev_id, queue_id, &stats);
+	ret = rte_bbdev_queue_stats_get(ad->dev_id, queue_id, &stats);
 	TEST_ASSERT_SUCCESS(ret,
 			"Failed to get stats for queue (%u) of device (%u)",
 			queue_id, ad->dev_id);
diff --git a/doc/guides/rel_notes/release_26_07.rst b/doc/guides/rel_notes/release_26_07.rst
index 6badd6d91b..dbad975d8c 100644
--- a/doc/guides/rel_notes/release_26_07.rst
+++ b/doc/guides/rel_notes/release_26_07.rst
@@ -190,6 +190,11 @@ New Features
   * Added ``eof`` devarg to use link state to signal end of receive file input.
   * Added unit test suite.
 
+* **Added per-queue statistics API in bbdev.**
+
+  Added ``rte_bbdev_queue_stats_get()`` function to retrieve statistics
+  for a specific queue, complementing the existing device-level statistics API.
+
 * **Updated Marvell cnxk crypto driver.**
 
   * Added support for ML-KEM and ML-DSA on CN20K platform.
diff --git a/lib/bbdev/rte_bbdev.c b/lib/bbdev/rte_bbdev.c
index 90095578cb..3f032a8e32 100644
--- a/lib/bbdev/rte_bbdev.c
+++ b/lib/bbdev/rte_bbdev.c
@@ -819,6 +819,26 @@ rte_bbdev_stats_reset(uint16_t dev_id)
 	return 0;
 }
 
+RTE_EXPORT_EXPERIMENTAL_SYMBOL(rte_bbdev_queue_stats_get, 26.07)
+int rte_bbdev_queue_stats_get(uint16_t dev_id, uint16_t queue_id, struct rte_bbdev_stats *stats)
+{
+	struct rte_bbdev *dev = get_dev(dev_id);
+
+	VALID_DEV_OR_RET_ERR(dev, dev_id);
+	VALID_QUEUE_OR_RET_ERR(queue_id, dev);
+
+	if (stats == NULL) {
+		rte_bbdev_log(ERR, "NULL stats structure");
+		return -EINVAL;
+	}
+
+	*stats = dev->data->queues[queue_id].queue_stats;
+
+	rte_bbdev_log_debug("Retrieved stats for queue %u of device %u",
+			    queue_id, dev_id);
+	return 0;
+}
+
 RTE_EXPORT_SYMBOL(rte_bbdev_info_get)
 int
 rte_bbdev_info_get(uint16_t dev_id, struct rte_bbdev_info *dev_info)
diff --git a/lib/bbdev/rte_bbdev.h b/lib/bbdev/rte_bbdev.h
index 36dabf8fa6..9fcfbf4c69 100644
--- a/lib/bbdev/rte_bbdev.h
+++ b/lib/bbdev/rte_bbdev.h
@@ -316,6 +316,26 @@ rte_bbdev_stats_get(uint16_t dev_id, struct rte_bbdev_stats *stats);
 int
 rte_bbdev_stats_reset(uint16_t dev_id);
 
+/**
+ * Retrieve the statistics of a specific queue.
+ *
+ * @param dev_id
+ *   The identifier of the device.
+ * @param queue_id
+ *   The index of the queue.
+ * @param stats
+ *   Pointer to structure to where statistics will be copied. On error, this
+ *   location may or may not have been modified.
+ *
+ * @return
+ *   - 0 on success
+ *   - -ENODEV if dev_id is invalid
+ *   - -EINVAL if stats is NULL
+ *   - -ERANGE if queue_id is out of range
+ */
+__rte_experimental
+int rte_bbdev_queue_stats_get(uint16_t dev_id, uint16_t queue_id, struct rte_bbdev_stats *stats);
+
 /** Device information supplied by the device's driver */
 
 /* Structure rte_bbdev_driver_info 8< */
-- 
2.54.0

[PATCH 03/10] hash: fix GFNI stubs export

[David Marchand]

Internal symbols *can not* be called as part of public ABI.
The GFNI stubs must be exported so that an application may call them
through the (inlined) GFNI API without requiring ALLOW_INTERNAL_API.

Fixes: 944a03a5cfc1 ("hash: fix MSVC link on GFNI stubs")

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 lib/hash/rte_thash_gfni.c | 4 ++--
 lib/hash/rte_thash_gfni.h | 2 --
 2 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/lib/hash/rte_thash_gfni.c b/lib/hash/rte_thash_gfni.c
index 2003c7b3db..ae55f7f711 100644
--- a/lib/hash/rte_thash_gfni.c
+++ b/lib/hash/rte_thash_gfni.c
@@ -13,7 +13,7 @@ RTE_LOG_REGISTER_SUFFIX(hash_gfni_logtype, gfni, INFO);
 #define HASH_LOG(level, ...) \
 	RTE_LOG_LINE(level, HASH, "" __VA_ARGS__)
 
-RTE_EXPORT_INTERNAL_SYMBOL(rte_thash_gfni_stub)
+RTE_EXPORT_SYMBOL(rte_thash_gfni_stub)
 uint32_t
 rte_thash_gfni_stub(const uint64_t *mtrx __rte_unused,
 	const uint8_t *key __rte_unused, int len __rte_unused)
@@ -29,7 +29,7 @@ rte_thash_gfni_stub(const uint64_t *mtrx __rte_unused,
 	return 0;
 }
 
-RTE_EXPORT_INTERNAL_SYMBOL(rte_thash_gfni_bulk_stub)
+RTE_EXPORT_SYMBOL(rte_thash_gfni_bulk_stub)
 void
 rte_thash_gfni_bulk_stub(const uint64_t *mtrx __rte_unused,
 	int len __rte_unused, uint8_t *tuple[] __rte_unused,
diff --git a/lib/hash/rte_thash_gfni.h b/lib/hash/rte_thash_gfni.h
index e82378933c..8a72fa9ec6 100644
--- a/lib/hash/rte_thash_gfni.h
+++ b/lib/hash/rte_thash_gfni.h
@@ -22,11 +22,9 @@ extern "C" {
  * @internal
  * Stubs only used when GFNI is not available.
  */
-__rte_internal
 uint32_t
 rte_thash_gfni_stub(const uint64_t *mtrx, const uint8_t *key, int len);
 
-__rte_internal
 void
 rte_thash_gfni_bulk_stub(const uint64_t *mtrx, int len, uint8_t *tuple[],
 	uint32_t val[], uint32_t num);
-- 
2.54.0

[PATCH 04/10] test: uninline helper for forking

[David Marchand]

There is no reason to keep those helpers inlined.

Besides, this code relies on internal API.
So any consumer of process.h requires internal API.

Move this code in process.c.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 MAINTAINERS           |   1 +
 app/test/meson.build  |   4 +
 app/test/process.c    | 229 ++++++++++++++++++++++++++++++++++++++++++
 app/test/process.h    | 228 ++---------------------------------------
 app/test/test_pdump.h |   3 +
 5 files changed, 246 insertions(+), 219 deletions(-)
 create mode 100644 app/test/process.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 84c528437f..176f58f64c 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1950,6 +1950,7 @@ F: app/test/commands.c
 F: app/test/packet_burst_generator.c
 F: app/test/packet_burst_generator.h
 F: app/test/process.h
+F: app/test/process.c
 F: app/test/test.c
 F: app/test/test.h
 F: app/test/test_pmd_perf.c
diff --git a/app/test/meson.build b/app/test/meson.build
index 51abeeb732..d8efa93050 100644
--- a/app/test/meson.build
+++ b/app/test/meson.build
@@ -5,6 +5,10 @@
 deps += ['cmdline', 'ring', 'mempool', 'mbuf']
 sources += files('commands.c', 'test.c')
 
+if not is_windows
+    sources += files('process.c')
+endif
+
 # optional dependencies: some files may use these - and so we should link them in -
 # but do not explicitly require them so they are not listed in the per-file lists below
 optional_deps = ['crypto_scheduler', 'lpm']
diff --git a/app/test/process.c b/app/test/process.c
new file mode 100644
index 0000000000..2b6d7ae84b
--- /dev/null
+++ b/app/test/process.c
@@ -0,0 +1,229 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#include <errno.h>  /* errno */
+#include <limits.h> /* PATH_MAX */
+#ifndef RTE_EXEC_ENV_WINDOWS
+#include <sys/wait.h>
+#endif
+#include <stdlib.h> /* NULL */
+#include <string.h> /* strerror */
+#include <unistd.h>
+#include <dirent.h>
+
+#include <rte_string_fns.h> /* strlcpy */
+#include <rte_devargs.h>
+#include <rte_eal.h>
+
+#ifdef RTE_LIB_PDUMP
+#ifdef RTE_NET_RING
+#include <rte_thread.h>
+#include "test_pdump.h"
+#endif
+#endif
+
+#include "test.h"
+#include "process.h"
+
+#define PREFIX_ALLOW "--allow="
+#define PREFIX_DRIVER_PATH "--driver-path="
+
+static int
+add_parameter_allow(char **argv, int max_capacity)
+{
+	struct rte_devargs *devargs;
+	int count = 0;
+
+	RTE_EAL_DEVARGS_FOREACH(NULL, devargs) {
+		if (strlen(devargs->name) == 0)
+			continue;
+
+		if (devargs->data == NULL || strlen(devargs->data) == 0) {
+			if (asprintf(&argv[count], PREFIX_ALLOW"%s", devargs->name) < 0)
+				break;
+		} else {
+			if (asprintf(&argv[count], PREFIX_ALLOW"%s,%s",
+					 devargs->name, devargs->data) < 0)
+				break;
+		}
+
+		if (++count == max_capacity)
+			break;
+	}
+
+	return count;
+}
+
+static int
+add_parameter_driver_path(char **argv, int max_capacity)
+{
+	const char *driver_path;
+	int count = 0;
+
+	RTE_EAL_DRIVER_PATH_FOREACH(driver_path, true) {
+		if (asprintf(&argv[count], PREFIX_DRIVER_PATH"%s", driver_path) < 0)
+			break;
+
+		if (++count == max_capacity)
+			break;
+	}
+
+	return count;
+}
+
+int
+process_dup(const char *const argv[], int numargs, const char *env_value)
+{
+	int num = 0;
+	char **argv_cpy;
+	int allow_num;
+	int driver_path_num;
+	int argv_num;
+	int i, status;
+#ifdef RTE_LIB_PDUMP
+#ifdef RTE_NET_RING
+	rte_thread_t thread;
+	int rc;
+#endif
+#endif
+
+	pid_t pid = fork();
+	if (pid < 0)
+		return -1;
+	else if (pid == 0) {
+		allow_num = rte_devargs_type_count(RTE_DEVTYPE_ALLOWED);
+		driver_path_num = rte_eal_driver_path_count(true);
+		argv_num = numargs + allow_num + driver_path_num + 1;
+		argv_cpy = calloc(argv_num, sizeof(char *));
+		if (!argv_cpy)
+			rte_panic("Memory allocation failed\n");
+
+		/* make a copy of the arguments to be passed to exec */
+		for (i = 0; i < numargs; i++) {
+			argv_cpy[i] = strdup(argv[i]);
+			if (argv_cpy[i] == NULL)
+				rte_panic("Error dup args\n");
+		}
+		if (allow_num > 0)
+			num = add_parameter_allow(&argv_cpy[i], allow_num);
+		num += numargs;
+
+		if (driver_path_num > 0) {
+			int added = add_parameter_driver_path(&argv_cpy[num], driver_path_num);
+			num += added;
+		}
+
+#ifdef RTE_EXEC_ENV_LINUX
+		{
+			const char *procdir = "/proc/self/fd/";
+			struct dirent *dirent;
+			char *endptr;
+			int fd, fdir;
+			DIR *dir;
+
+			/* close all open file descriptors, check /proc/self/fd
+			 * to only call close on open fds. Exclude fds 0, 1 and
+			 * 2
+			 */
+			dir = opendir(procdir);
+			if (dir == NULL) {
+				rte_panic("Error opening %s: %s\n", procdir,
+						strerror(errno));
+			}
+
+			fdir = dirfd(dir);
+			if (fdir < 0) {
+				status = errno;
+				closedir(dir);
+				rte_panic("Error %d obtaining fd for dir %s: %s\n",
+						fdir, procdir,
+						strerror(status));
+			}
+
+			while ((dirent = readdir(dir)) != NULL) {
+
+				if (strcmp(dirent->d_name, ".") == 0 ||
+					strcmp(dirent->d_name, "..") == 0)
+					continue;
+
+				errno = 0;
+				fd = strtol(dirent->d_name, &endptr, 10);
+				if (errno != 0 || endptr[0] != '\0') {
+					printf("Error converting name fd %d %s:\n",
+						fd, dirent->d_name);
+					continue;
+				}
+
+				if (fd == fdir || fd <= 2)
+					continue;
+
+				close(fd);
+			}
+			closedir(dir);
+		}
+#endif
+		printf("Running binary with argv[]:");
+		for (i = 0; i < num; i++)
+			printf("'%s' ", argv_cpy[i]);
+		printf("\n");
+		fflush(stdout);
+
+		/* set the environment variable */
+		if (setenv(RECURSIVE_ENV_VAR, env_value, 1) != 0)
+			rte_panic("Cannot export environment variable\n");
+		if (execv(argv_cpy[0], argv_cpy) < 0)
+			rte_panic("Cannot exec: %s\n", strerror(errno));
+	}
+	/* parent process does a wait */
+#ifdef RTE_LIB_PDUMP
+#ifdef RTE_NET_RING
+	if ((strcmp(env_value, "run_pdump_server_tests") == 0)) {
+		rc = rte_thread_create(&thread, NULL, send_pkts, NULL);
+		if (rc != 0) {
+			rte_panic("Cannot start send pkts thread: %s\n",
+				  strerror(rc));
+		}
+	}
+#endif
+#endif
+
+	while (wait(&status) != pid)
+		;
+#ifdef RTE_LIB_PDUMP
+#ifdef RTE_NET_RING
+	if ((strcmp(env_value, "run_pdump_server_tests") == 0)) {
+		flag_for_send_pkts = 0;
+		rte_thread_join(thread, NULL);
+	}
+#endif
+#endif
+	return status;
+}
+
+#ifndef RTE_EXEC_ENV_LINUX
+const char *
+file_prefix_arg(void)
+{
+	return "";
+}
+#else /* RTE_EXEC_ENV_LINUX */
+char *
+get_current_prefix(char *prefix, int size)
+{
+	rte_basename(rte_eal_get_runtime_dir(), prefix, size);
+	return prefix;
+}
+
+/* Return a --file-prefix=XXXX argument */
+const char *
+file_prefix_arg(void)
+{
+	static char prefix[NAME_MAX + sizeof("--file-prefix=")];
+	char tmp[NAME_MAX];
+
+	snprintf(prefix, sizeof(prefix), "--file-prefix=%s",
+			get_current_prefix(tmp, sizeof(tmp)));
+	return prefix;
+}
+#endif /* RTE_EXEC_ENV_LINUX */
diff --git a/app/test/process.h b/app/test/process.h
index 3ee899dbc8..cf2f95f035 100644
--- a/app/test/process.h
+++ b/app/test/process.h
@@ -5,73 +5,7 @@
 #ifndef _PROCESS_H_
 #define _PROCESS_H_
 
-#include <errno.h>  /* errno */
-#include <limits.h> /* PATH_MAX */
-#ifndef RTE_EXEC_ENV_WINDOWS
-#include <sys/wait.h>
-#endif
-#include <stdlib.h> /* NULL */
-#include <string.h> /* strerror */
-#include <unistd.h>
-#include <dirent.h>
-
-#include <rte_string_fns.h> /* strlcpy */
-#include <rte_devargs.h>
-#include <rte_eal.h>
-
-#ifdef RTE_LIB_PDUMP
-#ifdef RTE_NET_RING
-#include <rte_thread.h>
-extern uint32_t send_pkts(void *empty);
-extern uint16_t flag_for_send_pkts;
-#endif
-#endif
-
-#define PREFIX_ALLOW "--allow="
-#define PREFIX_DRIVER_PATH "--driver-path="
-
-static int
-add_parameter_allow(char **argv, int max_capacity)
-{
-	struct rte_devargs *devargs;
-	int count = 0;
-
-	RTE_EAL_DEVARGS_FOREACH(NULL, devargs) {
-		if (strlen(devargs->name) == 0)
-			continue;
-
-		if (devargs->data == NULL || strlen(devargs->data) == 0) {
-			if (asprintf(&argv[count], PREFIX_ALLOW"%s", devargs->name) < 0)
-				break;
-		} else {
-			if (asprintf(&argv[count], PREFIX_ALLOW"%s,%s",
-					 devargs->name, devargs->data) < 0)
-				break;
-		}
-
-		if (++count == max_capacity)
-			break;
-	}
-
-	return count;
-}
-
-static int
-add_parameter_driver_path(char **argv, int max_capacity)
-{
-	const char *driver_path;
-	int count = 0;
-
-	RTE_EAL_DRIVER_PATH_FOREACH(driver_path, true) {
-		if (asprintf(&argv[count], PREFIX_DRIVER_PATH"%s", driver_path) < 0)
-			break;
-
-		if (++count == max_capacity)
-			break;
-	}
-
-	return count;
-}
+#include <stdint.h>
 
 /*
  * launches a second copy of the test process using the given argv parameters,
@@ -79,161 +13,17 @@ add_parameter_driver_path(char **argv, int max_capacity)
  * subprocess the source of the call, the env_value parameter is set in the
  * environment as $RTE_TEST
  */
-static inline int
-process_dup(const char *const argv[], int numargs, const char *env_value)
-{
-	int num = 0;
-	char **argv_cpy;
-	int allow_num;
-	int driver_path_num;
-	int argv_num;
-	int i, status;
-#ifdef RTE_LIB_PDUMP
-#ifdef RTE_NET_RING
-	rte_thread_t thread;
-	int rc;
-#endif
-#endif
-
-	pid_t pid = fork();
-	if (pid < 0)
-		return -1;
-	else if (pid == 0) {
-		allow_num = rte_devargs_type_count(RTE_DEVTYPE_ALLOWED);
-		driver_path_num = rte_eal_driver_path_count(true);
-		argv_num = numargs + allow_num + driver_path_num + 1;
-		argv_cpy = calloc(argv_num, sizeof(char *));
-		if (!argv_cpy)
-			rte_panic("Memory allocation failed\n");
-
-		/* make a copy of the arguments to be passed to exec */
-		for (i = 0; i < numargs; i++) {
-			argv_cpy[i] = strdup(argv[i]);
-			if (argv_cpy[i] == NULL)
-				rte_panic("Error dup args\n");
-		}
-		if (allow_num > 0)
-			num = add_parameter_allow(&argv_cpy[i], allow_num);
-		num += numargs;
+int process_dup(const char *const argv[], int numargs, const char *env_value);
 
-		if (driver_path_num > 0) {
-			int added = add_parameter_driver_path(&argv_cpy[num], driver_path_num);
-			num += added;
-		}
+/*
+ * Return a --file-prefix=XXXX argument
+ * Note: only Linux supports file prefixes.
+ */
+const char *file_prefix_arg(void);
 
 #ifdef RTE_EXEC_ENV_LINUX
-		{
-			const char *procdir = "/proc/self/fd/";
-			struct dirent *dirent;
-			char *endptr;
-			int fd, fdir;
-			DIR *dir;
-
-			/* close all open file descriptors, check /proc/self/fd
-			 * to only call close on open fds. Exclude fds 0, 1 and
-			 * 2
-			 */
-			dir = opendir(procdir);
-			if (dir == NULL) {
-				rte_panic("Error opening %s: %s\n", procdir,
-						strerror(errno));
-			}
-
-			fdir = dirfd(dir);
-			if (fdir < 0) {
-				status = errno;
-				closedir(dir);
-				rte_panic("Error %d obtaining fd for dir %s: %s\n",
-						fdir, procdir,
-						strerror(status));
-			}
-
-			while ((dirent = readdir(dir)) != NULL) {
-
-				if (strcmp(dirent->d_name, ".") == 0 ||
-					strcmp(dirent->d_name, "..") == 0)
-					continue;
-
-				errno = 0;
-				fd = strtol(dirent->d_name, &endptr, 10);
-				if (errno != 0 || endptr[0] != '\0') {
-					printf("Error converting name fd %d %s:\n",
-						fd, dirent->d_name);
-					continue;
-				}
-
-				if (fd == fdir || fd <= 2)
-					continue;
-
-				close(fd);
-			}
-			closedir(dir);
-		}
+/* Get current hugepage file prefix */
+char *get_current_prefix(char *prefix, int size);
 #endif
-		printf("Running binary with argv[]:");
-		for (i = 0; i < num; i++)
-			printf("'%s' ", argv_cpy[i]);
-		printf("\n");
-		fflush(stdout);
-
-		/* set the environment variable */
-		if (setenv(RECURSIVE_ENV_VAR, env_value, 1) != 0)
-			rte_panic("Cannot export environment variable\n");
-		if (execv(argv_cpy[0], argv_cpy) < 0)
-			rte_panic("Cannot exec: %s\n", strerror(errno));
-	}
-	/* parent process does a wait */
-#ifdef RTE_LIB_PDUMP
-#ifdef RTE_NET_RING
-	if ((strcmp(env_value, "run_pdump_server_tests") == 0)) {
-		rc = rte_thread_create(&thread, NULL, send_pkts, NULL);
-		if (rc != 0) {
-			rte_panic("Cannot start send pkts thread: %s\n",
-				  strerror(rc));
-		}
-	}
-#endif
-#endif
-
-	while (wait(&status) != pid)
-		;
-#ifdef RTE_LIB_PDUMP
-#ifdef RTE_NET_RING
-	if ((strcmp(env_value, "run_pdump_server_tests") == 0)) {
-		flag_for_send_pkts = 0;
-		rte_thread_join(thread, NULL);
-	}
-#endif
-#endif
-	return status;
-}
-
-/* Only Linux supports file prefixes. */
-#ifndef RTE_EXEC_ENV_LINUX
-static inline const char *
-file_prefix_arg(void)
-{
-	return "";
-}
-#else /* RTE_EXEC_ENV_LINUX */
-static inline char *
-get_current_prefix(char *prefix, int size)
-{
-	rte_basename(rte_eal_get_runtime_dir(), prefix, size);
-	return prefix;
-}
-
-/* Return a --file-prefix=XXXX argument */
-static inline const char *
-file_prefix_arg(void)
-{
-	static char prefix[NAME_MAX + sizeof("--file-prefix=")];
-	char tmp[NAME_MAX];
-
-	snprintf(prefix, sizeof(prefix), "--file-prefix=%s",
-			get_current_prefix(tmp, sizeof(tmp)));
-	return prefix;
-}
-#endif /* RTE_EXEC_ENV_LINUX */
 
 #endif /* _PROCESS_H_ */
diff --git a/app/test/test_pdump.h b/app/test/test_pdump.h
index 8746d61269..b026dcc6fb 100644
--- a/app/test/test_pdump.h
+++ b/app/test/test_pdump.h
@@ -11,6 +11,9 @@
 /* sample test to send packets to the pdump client recursively */
 uint32_t send_pkts(void *empty);
 
+/* flag to stop send_pkts thread */
+extern uint16_t flag_for_send_pkts;
+
 /* Sample test to create setup for the pdump server tests */
 int test_pdump_init(void);
 
-- 
2.54.0

[PATCH 05/10] test/bonding: get MAC address with public API

[David Marchand]

Replace direct rte_eth_devices[].data->mac_addrs access with
rte_eth_macaddr_get() public API. This removes the dependency
on the internal ethdev_driver.h header.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 app/test/test_link_bonding.c | 35 +++++++++++++----------------------
 1 file changed, 13 insertions(+), 22 deletions(-)

diff --git a/app/test/test_link_bonding.c b/app/test/test_link_bonding.c
index 19b064771a..b9c6718e36 100644
--- a/app/test/test_link_bonding.c
+++ b/app/test/test_link_bonding.c
@@ -19,7 +19,6 @@
 #include <rte_common.h>
 #include <rte_debug.h>
 #include <rte_ethdev.h>
-#include <ethdev_driver.h>
 #include <rte_log.h>
 #include <rte_lcore.h>
 #include <rte_memory.h>
@@ -4763,7 +4762,7 @@ test_alb_change_mac_in_reply_sent(void)
 	int retval = 0;
 
 	struct rte_ether_addr bond_mac, client_mac;
-	struct rte_ether_addr *member_mac1, *member_mac2;
+	struct rte_ether_addr member_mac1, member_mac2;
 
 	TEST_ASSERT_SUCCESS(
 			initialize_bonding_device_with_members(BONDING_MODE_ALB,
@@ -4779,9 +4778,7 @@ test_alb_change_mac_in_reply_sent(void)
 				MAX_PKT_BURST);
 	}
 
-	rte_ether_addr_copy(
-			rte_eth_devices[test_params->bonding_port_id].data->mac_addrs,
-			&bond_mac);
+	rte_eth_macaddr_get(test_params->bonding_port_id, &bond_mac);
 
 	/*
 	 * Generating four packets with different mac and ip addresses and sending
@@ -4831,10 +4828,8 @@ test_alb_change_mac_in_reply_sent(void)
 			RTE_ARP_OP_REPLY);
 	rte_eth_tx_burst(test_params->bonding_port_id, 0, &pkt, 1);
 
-	member_mac1 =
-			rte_eth_devices[test_params->member_port_ids[0]].data->mac_addrs;
-	member_mac2 =
-			rte_eth_devices[test_params->member_port_ids[1]].data->mac_addrs;
+	rte_eth_macaddr_get(test_params->member_port_ids[0], &member_mac1);
+	rte_eth_macaddr_get(test_params->member_port_ids[1], &member_mac2);
 
 	/*
 	 * Checking if packets are properly distributed on bonding ports. Packets
@@ -4852,13 +4847,13 @@ test_alb_change_mac_in_reply_sent(void)
 						sizeof(struct rte_ether_hdr));
 
 			if (member_idx%2 == 0) {
-				if (!rte_is_same_ether_addr(member_mac1,
+				if (!rte_is_same_ether_addr(&member_mac1,
 						&arp_pkt->arp_data.arp_sha)) {
 					retval = -1;
 					goto test_end;
 				}
 			} else {
-				if (!rte_is_same_ether_addr(member_mac2,
+				if (!rte_is_same_ether_addr(&member_mac2,
 						&arp_pkt->arp_data.arp_sha)) {
 					retval = -1;
 					goto test_end;
@@ -4885,7 +4880,7 @@ test_alb_reply_from_client(void)
 	int retval = 0;
 
 	struct rte_ether_addr bond_mac, client_mac;
-	struct rte_ether_addr *member_mac1, *member_mac2;
+	struct rte_ether_addr member_mac1, member_mac2;
 
 	TEST_ASSERT_SUCCESS(
 			initialize_bonding_device_with_members(BONDING_MODE_ALB,
@@ -4900,9 +4895,7 @@ test_alb_reply_from_client(void)
 				MAX_PKT_BURST);
 	}
 
-	rte_ether_addr_copy(
-			rte_eth_devices[test_params->bonding_port_id].data->mac_addrs,
-			&bond_mac);
+	rte_eth_macaddr_get(test_params->bonding_port_id, &bond_mac);
 
 	/*
 	 * Generating four packets with different mac and ip addresses and placing
@@ -4963,8 +4956,8 @@ test_alb_reply_from_client(void)
 	rte_eth_rx_burst(test_params->bonding_port_id, 0, pkts_sent, MAX_PKT_BURST);
 	rte_eth_tx_burst(test_params->bonding_port_id, 0, NULL, 0);
 
-	member_mac1 = rte_eth_devices[test_params->member_port_ids[0]].data->mac_addrs;
-	member_mac2 = rte_eth_devices[test_params->member_port_ids[1]].data->mac_addrs;
+	rte_eth_macaddr_get(test_params->member_port_ids[0], &member_mac1);
+	rte_eth_macaddr_get(test_params->member_port_ids[1], &member_mac2);
 
 	/*
 	 * Checking if update ARP packets were properly send on member ports.
@@ -4981,13 +4974,13 @@ test_alb_reply_from_client(void)
 						sizeof(struct rte_ether_hdr));
 
 			if (member_idx%2 == 0) {
-				if (!rte_is_same_ether_addr(member_mac1,
+				if (!rte_is_same_ether_addr(&member_mac1,
 						&arp_pkt->arp_data.arp_sha)) {
 					retval = -1;
 					goto test_end;
 				}
 			} else {
-				if (!rte_is_same_ether_addr(member_mac2,
+				if (!rte_is_same_ether_addr(&member_mac2,
 						&arp_pkt->arp_data.arp_sha)) {
 					retval = -1;
 					goto test_end;
@@ -5035,9 +5028,7 @@ test_alb_receive_vlan_reply(void)
 				MAX_PKT_BURST);
 	}
 
-	rte_ether_addr_copy(
-			rte_eth_devices[test_params->bonding_port_id].data->mac_addrs,
-			&bond_mac);
+	rte_eth_macaddr_get(test_params->bonding_port_id, &bond_mac);
 
 	/*
 	 * Generating packet with double VLAN header and placing it in the rx queue.
-- 
2.54.0

[PATCH 06/10] test/devargs: check driver presence with public API

[David Marchand]

Let's avoid calling internal drivers API.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 app/test/test_devargs.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
index 0bd14c24ec..6c194ecfc6 100644
--- a/app/test/test_devargs.c
+++ b/app/test/test_devargs.c
@@ -10,7 +10,7 @@
 #include <rte_common.h>
 #include <rte_devargs.h>
 #include <rte_kvargs.h>
-#include <bus_driver.h>
+#include <rte_bus.h>
 #include <rte_class.h>
 
 #include "test.h"
@@ -167,14 +167,18 @@ test_valid_devargs(void)
 		{ "net_ring0,iface=test,path=/class/bus/,queues=1",
 		  0, 0, 3, "vdev", "net_ring0", NULL },
 	};
-	struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
+	struct rte_devargs da;
 	int ret;
 
 	ret = test_valid_devargs_cases(list, RTE_DIM(list));
-	if (vdev_bus != NULL && vdev_bus->parse("net_ring0", NULL) == 0)
+
+	memset(&da, 0, sizeof(da));
+	if (rte_devargs_parse(&da, "net_ring0") == 0)
 		/* Ring vdev driver enabled. */
 		ret |= test_valid_devargs_cases(legacy_ring_list,
 						RTE_DIM(legacy_ring_list));
+	rte_devargs_reset(&da);
+
 	return ret;
 }
 
-- 
2.54.0

[PATCH 07/10] test/vdev: find device with public API

[David Marchand]

Checking internal API has little sense as users are not supposed to call
those.
Use public API to achieve similar coverage.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 app/test/test_vdev.c | 85 ++++++++------------------------------------
 1 file changed, 15 insertions(+), 70 deletions(-)

diff --git a/app/test/test_vdev.c b/app/test/test_vdev.c
index c300976ace..6951ef7cc3 100644
--- a/app/test/test_vdev.c
+++ b/app/test/test_vdev.c
@@ -2,69 +2,26 @@
  * Copyright 2021 6WIND S.A.
  */
 
-#include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
 
-#include <rte_common.h>
 #include <rte_dev.h>
-#include <rte_kvargs.h>
-#include <bus_driver.h>
+#include <rte_bus.h>
 #include <rte_bus_vdev.h>
 
 #include "test.h"
 
-#define TEST_VDEV_KEY_NAME "name"
-
-static const char * const valid_keys[] = {
-	TEST_VDEV_KEY_NAME,
-	NULL,
-};
-
-static int
-cmp_dev_name(const struct rte_device *dev, const void *name)
-{
-	return strcmp(rte_dev_name(dev), name);
-}
-
-static int
-cmp_dev_match(const struct rte_device *dev, const void *_kvlist)
-{
-	const struct rte_kvargs *kvlist = _kvlist;
-	const char *key = TEST_VDEV_KEY_NAME;
-	const char *name;
-
-	/* no kvlist arg, all devices match */
-	if (kvlist == NULL)
-		return 0;
-
-	/* if key is present in kvlist and does not match, filter device */
-	name = rte_kvargs_get(kvlist, key);
-	if (name != NULL && strcmp(name, rte_dev_name(dev)) != 0)
-		return -1;
-
-	return 0;
-}
-
 static struct rte_device *
-get_matching_vdev(const char *match_str)
+find_vdev_by_name(const char *name)
 {
-	struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
-	struct rte_kvargs *kvargs = NULL;
+	struct rte_dev_iterator it = { 0 };
 	struct rte_device *dev;
 
-	if (match_str != NULL) {
-		kvargs = rte_kvargs_parse(match_str, valid_keys);
-		if (kvargs == NULL) {
-			printf("Failed to parse match string\n");
-			return NULL;
-		}
+	RTE_DEV_FOREACH(dev, "bus=vdev", &it) {
+		if (strcmp(rte_dev_name(dev), name) == 0)
+			return dev;
 	}
-
-	dev = vdev_bus->find_device(vdev_bus, NULL, cmp_dev_match, kvargs);
-	rte_kvargs_free(kvargs);
-
-	return dev;
+	return NULL;
 }
 
 static int
@@ -83,7 +40,7 @@ test_vdev_bus(void)
 		printf("Failed to create vdev net_null_test0\n");
 		goto fail;
 	}
-	dev0 = vdev_bus->find_device(vdev_bus, NULL, cmp_dev_name, "net_null_test0");
+	dev0 = find_vdev_by_name("net_null_test0");
 	if (dev0 == NULL) {
 		printf("Cannot find net_null_test0 vdev\n");
 		goto fail;
@@ -94,44 +51,32 @@ test_vdev_bus(void)
 		printf("Failed to create vdev net_null_test1\n");
 		goto fail;
 	}
-	dev1 = vdev_bus->find_device(vdev_bus, NULL, cmp_dev_name, "net_null_test1");
+	dev1 = find_vdev_by_name("net_null_test1");
 	if (dev1 == NULL) {
 		printf("Cannot find net_null_test1 vdev\n");
 		goto fail;
 	}
 
-	/* try to match vdevs */
-	dev = get_matching_vdev("name=net_null_test0");
+	/* try to find vdevs */
+	dev = find_vdev_by_name("net_null_test0");
 	if (dev != dev0) {
 		printf("Cannot match net_null_test0 vdev\n");
 		goto fail;
 	}
 
-	dev = get_matching_vdev("name=net_null_test1");
+	dev = find_vdev_by_name("net_null_test1");
 	if (dev != dev1) {
 		printf("Cannot match net_null_test1 vdev\n");
 		goto fail;
 	}
 
-	dev = get_matching_vdev("name=unexistant");
+	dev = find_vdev_by_name("nonexistent");
 	if (dev != NULL) {
-		printf("Unexistant vdev should not match\n");
-		goto fail;
-	}
-
-	dev = get_matching_vdev("");
-	if (dev == NULL || dev == dev1) {
-		printf("Cannot match any vdev with empty match string\n");
-		goto fail;
-	}
-
-	dev = get_matching_vdev(NULL);
-	if (dev == NULL || dev == dev1) {
-		printf("Cannot match any vdev with NULL match string\n");
+		printf("Nonexistent vdev should not match\n");
 		goto fail;
 	}
 
-	/* iterate all vdevs, and ensure we find vdev0 and vdev1 */
+	/* iterate all vdevs, and ensure we find dev0 and dev1 */
 	RTE_DEV_FOREACH(dev, "bus=vdev", &dev_iter) {
 		if (dev == dev0)
 			dev0 = NULL;
-- 
2.54.0

[PATCH 08/10] test: limit internal API usage

[David Marchand]

Internal APIs may be used for validating the internals of some
component, but in general, it should be discouraged.

Let's not allow globally internal API in the unit tests.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 app/test/meson.build         | 3 ---
 app/test/process.c           | 2 ++
 app/test/test_devargs.c      | 2 ++
 app/test/test_external_mem.c | 2 ++
 app/test/test_malloc.c       | 2 ++
 app/test/test_mempool.c      | 2 ++
 app/test/test_pdump.c        | 3 +++
 app/test/virtual_pmd.c       | 2 ++
 8 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/app/test/meson.build b/app/test/meson.build
index d8efa93050..e3dfa1d24c 100644
--- a/app/test/meson.build
+++ b/app/test/meson.build
@@ -286,9 +286,6 @@ foreach arg: extra_flags
     endif
 endforeach
 
-# Enable using internal APIs in unit tests
-cflags += '-DALLOW_INTERNAL_API'
-
 # create a symlink in the app/test directory for the binary, for backward compatibility
 if not is_windows
     custom_target('test_symlink',
diff --git a/app/test/process.c b/app/test/process.c
index 2b6d7ae84b..92c8af702b 100644
--- a/app/test/process.c
+++ b/app/test/process.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
+#define ALLOW_INTERNAL_API
+
 #include <errno.h>  /* errno */
 #include <limits.h> /* PATH_MAX */
 #ifndef RTE_EXEC_ENV_WINDOWS
diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
index 6c194ecfc6..66cf046463 100644
--- a/app/test/test_devargs.c
+++ b/app/test/test_devargs.c
@@ -2,6 +2,8 @@
  * Copyright (c) 2021 NVIDIA Corporation & Affiliates
  */
 
+#define ALLOW_INTERNAL_API
+
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
diff --git a/app/test/test_external_mem.c b/app/test/test_external_mem.c
index 53300983ed..2c8d0d8e00 100644
--- a/app/test/test_external_mem.c
+++ b/app/test/test_external_mem.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2018 Intel Corporation
  */
 
+#define ALLOW_INTERNAL_API
+
 #include "test.h"
 
 #include <errno.h>
diff --git a/app/test/test_malloc.c b/app/test/test_malloc.c
index da868c8091..151d1a3f2c 100644
--- a/app/test/test_malloc.c
+++ b/app/test/test_malloc.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2010-2019 Intel Corporation
  */
 
+#define ALLOW_INTERNAL_API
+
 #include "test.h"
 
 #include <stdio.h>
diff --git a/app/test/test_mempool.c b/app/test/test_mempool.c
index e54249ce61..f1e8adeef9 100644
--- a/app/test/test_mempool.c
+++ b/app/test/test_mempool.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
+#define ALLOW_INTERNAL_API
+
 #include <string.h>
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/app/test/test_pdump.c b/app/test/test_pdump.c
index 64f1bbf21c..f1c9bd9d7e 100644
--- a/app/test/test_pdump.c
+++ b/app/test/test_pdump.c
@@ -1,6 +1,9 @@
 /* SPDX-License-Identifier: BSD-3-Clause
  * Copyright(c) 2018 Intel Corporation
  */
+
+#define ALLOW_INTERNAL_API
+
 #include <stdio.h>
 #include <unistd.h>
 #include <stdint.h>
diff --git a/app/test/virtual_pmd.c b/app/test/virtual_pmd.c
index aa37e4073c..74bb3516ab 100644
--- a/app/test/virtual_pmd.c
+++ b/app/test/virtual_pmd.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
+#define ALLOW_INTERNAL_API
+
 #include <rte_mbuf.h>
 #include <rte_ethdev.h>
 #include <ethdev_driver.h>
-- 
2.54.0

[PATCH 09/10] ci: make ABI reference generation faster

[David Marchand]

Regenerating ABI is not something we do often, but when working on ABI
topics, a lot of time may be wasted on waiting for the reference being
generated.

We only care about shared libraries, avoid building final applications
or examples.
And we don't need to double check headers in the ABI reference.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 .ci/linux-build.sh            | 11 ++++++++++-
 devtools/test-meson-builds.sh |  5 ++++-
 2 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/.ci/linux-build.sh b/.ci/linux-build.sh
index e0b914a142..a5e602ed83 100755
--- a/.ci/linux-build.sh
+++ b/.ci/linux-build.sh
@@ -201,9 +201,18 @@ if [ "$ABI_CHECKS" = "true" ]; then
     fi
 
     if [ ! -d reference ]; then
+        REF_OPTS="$OPTS"
+        # We only care about the drivers and libs, disable developer checks and
+        # don't try to link apps.
+        REF_OPTS="$REF_OPTS -Dcheck_includes=false"
+        REF_OPTS="$REF_OPTS -Ddeveloper_mode=disabled"
+        REF_OPTS="$REF_OPTS -Ddisable_apps=*"
+        REF_OPTS="$REF_OPTS -Denable_docs=false"
+        REF_OPTS="$REF_OPTS -Dexamples="
+        REF_OPTS="$REF_OPTS -Dtests=false"
         refsrcdir=$(readlink -f $(pwd)/../dpdk-$REF_GIT_TAG)
         git clone --single-branch -b "$REF_GIT_TAG" $REF_GIT_REPO $refsrcdir
-        meson setup $OPTS -Dexamples= $refsrcdir $refsrcdir/build
+        meson setup $REF_OPTS $refsrcdir $refsrcdir/build
         ninja -C $refsrcdir/build
         DESTDIR=$(pwd)/reference meson install -C $refsrcdir/build
         find reference/usr/local -name '*.a' -delete
diff --git a/devtools/test-meson-builds.sh b/devtools/test-meson-builds.sh
index 11e4be3f88..8ca7091b47 100755
--- a/devtools/test-meson-builds.sh
+++ b/devtools/test-meson-builds.sh
@@ -206,8 +206,11 @@ build () # <directory> <target cc | cross file> <ABI check> [meson options]
 			fi
 
 			rm -rf $abirefdir/build
+			# We only care about the drivers and libs, disable
+			# developer checks and don't try to link apps.
 			config $abirefdir/src $abirefdir/build ABI $cross \
-				-Dexamples= $*
+				$* -Dcheck_includes=false -Ddeveloper_mode=disabled \
+				-Ddisable_apps=* -Dexamples= -Dtests=false
 			compile $abirefdir/build
 			install_target $abirefdir/build $abirefdir/$targetdir
 
-- 
2.54.0

[PATCH 10/10] ci: run reference unit tests against ABI

[David Marchand]

Let's check if unit tests from the reference build can run with the
current release libraries/drivers.

This gives another level of ABI checking, like catching regressions on
experimental or internal ABI symbols that get promoted to stable.

Skip unit tests that were added since the reference (iow v26.03).
$ git diff v26.03..origin/main |
  sed -ne 's/^[+]REGISTER_FAST_TEST(\([^,]*\).*$/\1/p' |
  sort
af_packet_pmd_autotest
bpf_exec_wrong_flags_autotest
bpf_exec_wrong_nb_prog_arg_autotest
bpf_load_null_autotest
bpf_validate_autotest
eal_flags_pagesz_mem_autotest
feature_autotest
pcap_pmd_autotest
reassembly_autotest

As reported in https://bugs.dpdk.org/show_bug.cgi?id=1967, devargs
and vdev autotests are broken due to drivers ABI changes since
those unit tests are directly accessing internal objects.
FIXME: for now, use a branch of mine with backports on 26.03.

Signed-off-by: David Marchand <david.marchand@redhat.com>
---
 .ci/linux-build.sh          | 28 +++++++++++++++++++++++++---
 .github/workflows/build.yml | 15 +++++++++++++--
 2 files changed, 38 insertions(+), 5 deletions(-)

diff --git a/.ci/linux-build.sh b/.ci/linux-build.sh
index a5e602ed83..f4fb398cd9 100755
--- a/.ci/linux-build.sh
+++ b/.ci/linux-build.sh
@@ -200,23 +200,28 @@ if [ "$ABI_CHECKS" = "true" ]; then
         rm -rf reference
     fi
 
+    # FIXME: force regeneration in CI for this patch
+    if [ ! -e reference/usr/local/bin/dpdk-test ]; then
+        rm -rf reference
+    fi
+
     if [ ! -d reference ]; then
         REF_OPTS="$OPTS"
         # We only care about the drivers and libs, disable developer checks and
         # don't try to link apps.
         REF_OPTS="$REF_OPTS -Dcheck_includes=false"
         REF_OPTS="$REF_OPTS -Ddeveloper_mode=disabled"
-        REF_OPTS="$REF_OPTS -Ddisable_apps=*"
+        REF_OPTS="$REF_OPTS -Denable_apps=test"
         REF_OPTS="$REF_OPTS -Denable_docs=false"
         REF_OPTS="$REF_OPTS -Dexamples="
-        REF_OPTS="$REF_OPTS -Dtests=false"
         refsrcdir=$(readlink -f $(pwd)/../dpdk-$REF_GIT_TAG)
         git clone --single-branch -b "$REF_GIT_TAG" $REF_GIT_REPO $refsrcdir
         meson setup $REF_OPTS $refsrcdir $refsrcdir/build
         ninja -C $refsrcdir/build
         DESTDIR=$(pwd)/reference meson install -C $refsrcdir/build
         find reference/usr/local -name '*.a' -delete
-        rm -rf reference/usr/local/bin
+        rm -rf reference/usr/local/bin/*
+        cp $refsrcdir/build/app/dpdk-test reference/usr/local/bin/dpdk-test
         rm -rf reference/usr/local/share
         echo $REF_GIT_TAG > reference/VERSION
     fi
@@ -233,6 +238,23 @@ if [ "$RUN_TESTS" = "true" ]; then
     catch_ubsan DPDK:fast-tests build/meson-logs/testlog.txt
     check_traces
     [ "$failed" != "true" ]
+
+    if [ "$ABI_CHECKS" = "true" ]; then
+        failed=
+        configure_coredump
+        mv -f build/app/dpdk-test build/app/dpdk-test.ori
+        cp reference/usr/local/bin/dpdk-test build/app/dpdk-test
+        for t in ${ABI_SKIP_TESTS}; do
+            DPDK_TEST_SKIP=${DPDK_TEST_SKIP+$DPDK_TEST_SKIP,}$t
+        done
+        sudo env DPDK_TEST_SKIP="$DPDK_TEST_SKIP" \
+            meson test -C build --suite fast-tests -t 3 --no-stdsplit --print-errorlogs || failed="true"
+        catch_coredump
+        catch_ubsan DPDK:fast-tests build/meson-logs/testlog.txt
+        check_traces
+        mv -f build/app/dpdk-test.ori build/app/dpdk-test
+        [ "$failed" != "true" ]
+    fi
 fi
 
 # Test examples compilation with an installed dpdk
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index f0ef39d34f..b311ee26f9 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -11,8 +11,9 @@ defaults:
 
 env:
   REF_GIT_BRANCH: main
-  REF_GIT_REPO: https://github.com/DPDK/dpdk
-  REF_GIT_TAG: v26.03
+  # FIXME
+  REF_GIT_REPO: https://github.com/david-marchand/dpdk
+  REF_GIT_TAG: 26.03
 
 jobs:
   checkpatch:
@@ -39,6 +40,16 @@ jobs:
     env:
       AARCH64: ${{ matrix.config.cross == 'aarch64' }}
       ABI_CHECKS: ${{ contains(matrix.config.checks, 'abi') }}
+      ABI_SKIP_TESTS: |
+        af_packet_pmd_autotest
+        bpf_exec_wrong_flags_autotest
+        bpf_exec_wrong_nb_prog_arg_autotest
+        bpf_load_null_autotest
+        bpf_validate_autotest
+        eal_flags_pagesz_mem_autotest
+        feature_autotest
+        pcap_pmd_autotest
+        reassembly_autotest
       ASAN: ${{ contains(matrix.config.checks, 'asan') }}
       BUILD_32BIT: ${{ matrix.config.cross == 'i386' }}
       BUILD_DEBUG: ${{ contains(matrix.config.checks, 'debug') }}
-- 
2.54.0

Re: [PATCH 00/10] Limit usage of internal API in tests

[David Marchand]

On Fri, 17 Jul 2026 at 11:30, David Marchand <david.marchand@redhat.com> wrote:
>
> This series is aimed at 26.11.
>
> We had a few bug reports related to internal (and experimental) symbols
> issues during 26.07 development.
> See for example https://bugs.dpdk.org/show_bug.cgi?id=1957 or more
> recently https://bugs.dpdk.org/show_bug.cgi?id=1967.
>
> To catch such issues earlier in the CI, this series proposes to run
> the unit tests through meson with the ABI reference unit test binary
> against the current ABI libraries and drivers.

I could have called test-null.sh for checking testpmd too.
For a next revision if any.

>
> For this to work, some unit tests must be skipped (since meson may
> invoke the ABI reference code with tests that were unknown at the time).
>
> A few unit tests were directly dereferencing internal structures and are
> reworked so they use public APIs.
>
> Additionally, unit tests were allowed to use any internal API which has
> hidden a few issues (like a public API backed by internal symbols in the
> hash library).
> So disable the global ALLLOW_INTERNAL_API and move it to code explicitly
> requiring internal API, with the hope it will push us to have better API.
>
> The last patch contains a hack for now, where the cached ABI reference
> is force regenerated with a branch of mine where I backported the
> fixes of this series.


-- 
David Marchand

Re: [PATCH 03/10] hash: fix GFNI stubs export

[Bruce Richardson]

On Fri, Jul 17, 2026 at 11:29:58AM +0200, David Marchand wrote:
> Internal symbols *can not* be called as part of public ABI.
> The GFNI stubs must be exported so that an application may call them
> through the (inlined) GFNI API without requiring ALLOW_INTERNAL_API.
> 
> Fixes: 944a03a5cfc1 ("hash: fix MSVC link on GFNI stubs")
> 
> Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
>  lib/hash/rte_thash_gfni.c | 4 ++--
>  lib/hash/rte_thash_gfni.h | 2 --
>  2 files changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/lib/hash/rte_thash_gfni.c b/lib/hash/rte_thash_gfni.c
> index 2003c7b3db..ae55f7f711 100644
> --- a/lib/hash/rte_thash_gfni.c
> +++ b/lib/hash/rte_thash_gfni.c
> @@ -13,7 +13,7 @@ RTE_LOG_REGISTER_SUFFIX(hash_gfni_logtype, gfni, INFO);
>  #define HASH_LOG(level, ...) \
>  	RTE_LOG_LINE(level, HASH, "" __VA_ARGS__)
>  
> -RTE_EXPORT_INTERNAL_SYMBOL(rte_thash_gfni_stub)
> +RTE_EXPORT_SYMBOL(rte_thash_gfni_stub)
>  uint32_t
>  rte_thash_gfni_stub(const uint64_t *mtrx __rte_unused,
>  	const uint8_t *key __rte_unused, int len __rte_unused)
> @@ -29,7 +29,7 @@ rte_thash_gfni_stub(const uint64_t *mtrx __rte_unused,
>  	return 0;
>  }
>  
> -RTE_EXPORT_INTERNAL_SYMBOL(rte_thash_gfni_bulk_stub)
> +RTE_EXPORT_SYMBOL(rte_thash_gfni_bulk_stub)
>  void
>  rte_thash_gfni_bulk_stub(const uint64_t *mtrx __rte_unused,
>  	int len __rte_unused, uint8_t *tuple[] __rte_unused,
> diff --git a/lib/hash/rte_thash_gfni.h b/lib/hash/rte_thash_gfni.h
> index e82378933c..8a72fa9ec6 100644
> --- a/lib/hash/rte_thash_gfni.h
> +++ b/lib/hash/rte_thash_gfni.h
> @@ -22,11 +22,9 @@ extern "C" {
>   * @internal
>   * Stubs only used when GFNI is not available.
>   */
> -__rte_internal
>  uint32_t
>  rte_thash_gfni_stub(const uint64_t *mtrx, const uint8_t *key, int len);
>  
> -__rte_internal
>  void
>  rte_thash_gfni_bulk_stub(const uint64_t *mtrx, int len, uint8_t *tuple[],
>  	uint32_t val[], uint32_t num);
> -- 
> 2.54.0
> 

Re: [PATCH 04/10] test: uninline helper for forking

[Bruce Richardson]

On Fri, Jul 17, 2026 at 11:29:59AM +0200, David Marchand wrote:
> There is no reason to keep those helpers inlined.
> 
> Besides, this code relies on internal API.
> So any consumer of process.h requires internal API.
> 
> Move this code in process.c.
> 
> Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

RE: [PATCH 03/10] hash: fix GFNI stubs export

[Konstantin Ananyev]

> 
> Internal symbols *can not* be called as part of public ABI.
> The GFNI stubs must be exported so that an application may call them
> through the (inlined) GFNI API without requiring ALLOW_INTERNAL_API.
> 
> Fixes: 944a03a5cfc1 ("hash: fix MSVC link on GFNI stubs")
> 
> Signed-off-by: David Marchand <david.marchand@redhat.com>
> ---
>  lib/hash/rte_thash_gfni.c | 4 ++--
>  lib/hash/rte_thash_gfni.h | 2 --
>  2 files changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/lib/hash/rte_thash_gfni.c b/lib/hash/rte_thash_gfni.c
> index 2003c7b3db..ae55f7f711 100644
> --- a/lib/hash/rte_thash_gfni.c
> +++ b/lib/hash/rte_thash_gfni.c
> @@ -13,7 +13,7 @@ RTE_LOG_REGISTER_SUFFIX(hash_gfni_logtype, gfni, INFO);
>  #define HASH_LOG(level, ...) \
>  	RTE_LOG_LINE(level, HASH, "" __VA_ARGS__)
> 
> -RTE_EXPORT_INTERNAL_SYMBOL(rte_thash_gfni_stub)
> +RTE_EXPORT_SYMBOL(rte_thash_gfni_stub)
>  uint32_t
>  rte_thash_gfni_stub(const uint64_t *mtrx __rte_unused,
>  	const uint8_t *key __rte_unused, int len __rte_unused)
> @@ -29,7 +29,7 @@ rte_thash_gfni_stub(const uint64_t *mtrx __rte_unused,
>  	return 0;
>  }
> 
> -RTE_EXPORT_INTERNAL_SYMBOL(rte_thash_gfni_bulk_stub)
> +RTE_EXPORT_SYMBOL(rte_thash_gfni_bulk_stub)
>  void
>  rte_thash_gfni_bulk_stub(const uint64_t *mtrx __rte_unused,
>  	int len __rte_unused, uint8_t *tuple[] __rte_unused,
> diff --git a/lib/hash/rte_thash_gfni.h b/lib/hash/rte_thash_gfni.h
> index e82378933c..8a72fa9ec6 100644
> --- a/lib/hash/rte_thash_gfni.h
> +++ b/lib/hash/rte_thash_gfni.h
> @@ -22,11 +22,9 @@ extern "C" {
>   * @internal
>   * Stubs only used when GFNI is not available.
>   */
> -__rte_internal
>  uint32_t
>  rte_thash_gfni_stub(const uint64_t *mtrx, const uint8_t *key, int len);
> 
> -__rte_internal
>  void
>  rte_thash_gfni_bulk_stub(const uint64_t *mtrx, int len, uint8_t *tuple[],
>  	uint32_t val[], uint32_t num);
> --

Acked-by: Konstantin Ananyev <konstantin.ananyev@huawei.com>

> 2.54.0

Re: [PATCH 06/10] test/devargs: check driver presence with public API

[Bruce Richardson]

On Fri, Jul 17, 2026 at 11:30:01AM +0200, David Marchand wrote:
> Let's avoid calling internal drivers API.
> 
> Signed-off-by: David Marchand <david.marchand@redhat.com>
> ---
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

One suggestion inline below.

>  app/test/test_devargs.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
> index 0bd14c24ec..6c194ecfc6 100644
> --- a/app/test/test_devargs.c
> +++ b/app/test/test_devargs.c
> @@ -10,7 +10,7 @@
>  #include <rte_common.h>
>  #include <rte_devargs.h>
>  #include <rte_kvargs.h>
> -#include <bus_driver.h>
> +#include <rte_bus.h>
>  #include <rte_class.h>
>  
>  #include "test.h"
> @@ -167,14 +167,18 @@ test_valid_devargs(void)
>  		{ "net_ring0,iface=test,path=/class/bus/,queues=1",
>  		  0, 0, 3, "vdev", "net_ring0", NULL },
>  	};
> -	struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
> +	struct rte_devargs da;
>  	int ret;
>  
>  	ret = test_valid_devargs_cases(list, RTE_DIM(list));
> -	if (vdev_bus != NULL && vdev_bus->parse("net_ring0", NULL) == 0)
> +
> +	memset(&da, 0, sizeof(da));
> +	if (rte_devargs_parse(&da, "net_ring0") == 0)

From what I/AI can see, there is nothing in this test case that requires
the ring PMD specifically. I think it would be good if it were updated to
use net/null, so that we standardize on our basic unit tests only relying
upon that PMD, rather than requiring a variety of drivers. WDYT?

>  		/* Ring vdev driver enabled. */
>  		ret |= test_valid_devargs_cases(legacy_ring_list,
>  						RTE_DIM(legacy_ring_list));
> +	rte_devargs_reset(&da);
> +
>  	return ret;
>  }
>  
> -- 
> 2.54.0
> 

Re: [PATCH 06/10] test/devargs: check driver presence with public API

[David Marchand]

On Fri, 17 Jul 2026 at 11:56, Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> On Fri, Jul 17, 2026 at 11:30:01AM +0200, David Marchand wrote:
> > Let's avoid calling internal drivers API.
> >
> > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > ---
> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
>
> One suggestion inline below.
>
> >  app/test/test_devargs.c | 10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> >
> > diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
> > index 0bd14c24ec..6c194ecfc6 100644
> > --- a/app/test/test_devargs.c
> > +++ b/app/test/test_devargs.c
> > @@ -10,7 +10,7 @@
> >  #include <rte_common.h>
> >  #include <rte_devargs.h>
> >  #include <rte_kvargs.h>
> > -#include <bus_driver.h>
> > +#include <rte_bus.h>
> >  #include <rte_class.h>
> >
> >  #include "test.h"
> > @@ -167,14 +167,18 @@ test_valid_devargs(void)
> >               { "net_ring0,iface=test,path=/class/bus/,queues=1",
> >                 0, 0, 3, "vdev", "net_ring0", NULL },
> >       };
> > -     struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
> > +     struct rte_devargs da;
> >       int ret;
> >
> >       ret = test_valid_devargs_cases(list, RTE_DIM(list));
> > -     if (vdev_bus != NULL && vdev_bus->parse("net_ring0", NULL) == 0)
> > +
> > +     memset(&da, 0, sizeof(da));
> > +     if (rte_devargs_parse(&da, "net_ring0") == 0)
>
> From what I/AI can see, there is nothing in this test case that requires

(you could write it as "From what A?I can see")

> the ring PMD specifically. I think it would be good if it were updated to
> use net/null, so that we standardize on our basic unit tests only relying
> upon that PMD, rather than requiring a variety of drivers. WDYT?

There may be a hidden (capillotracté) reason.
Passing path=/class/bus/ seems to be intended at catching errors in
devargs parsing.
No argument in net/null could accept such string.


-- 
David Marchand

Re: [PATCH 06/10] test/devargs: check driver presence with public API

[Bruce Richardson]

On Fri, Jul 17, 2026 at 12:08:01PM +0200, David Marchand wrote:
> On Fri, 17 Jul 2026 at 11:56, Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > On Fri, Jul 17, 2026 at 11:30:01AM +0200, David Marchand wrote:
> > > Let's avoid calling internal drivers API.
> > >
> > > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > > ---
> > Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> >
> > One suggestion inline below.
> >
> > >  app/test/test_devargs.c | 10 +++++++---
> > >  1 file changed, 7 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
> > > index 0bd14c24ec..6c194ecfc6 100644
> > > --- a/app/test/test_devargs.c
> > > +++ b/app/test/test_devargs.c
> > > @@ -10,7 +10,7 @@
> > >  #include <rte_common.h>
> > >  #include <rte_devargs.h>
> > >  #include <rte_kvargs.h>
> > > -#include <bus_driver.h>
> > > +#include <rte_bus.h>
> > >  #include <rte_class.h>
> > >
> > >  #include "test.h"
> > > @@ -167,14 +167,18 @@ test_valid_devargs(void)
> > >               { "net_ring0,iface=test,path=/class/bus/,queues=1",
> > >                 0, 0, 3, "vdev", "net_ring0", NULL },
> > >       };
> > > -     struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
> > > +     struct rte_devargs da;
> > >       int ret;
> > >
> > >       ret = test_valid_devargs_cases(list, RTE_DIM(list));
> > > -     if (vdev_bus != NULL && vdev_bus->parse("net_ring0", NULL) == 0)
> > > +
> > > +     memset(&da, 0, sizeof(da));
> > > +     if (rte_devargs_parse(&da, "net_ring0") == 0)
> >
> > From what I/AI can see, there is nothing in this test case that requires
> 
> (you could write it as "From what A?I can see")
> 
> > the ring PMD specifically. I think it would be good if it were updated to
> > use net/null, so that we standardize on our basic unit tests only relying
> > upon that PMD, rather than requiring a variety of drivers. WDYT?
> 
> There may be a hidden (capillotracté) reason.
> Passing path=/class/bus/ seems to be intended at catching errors in
> devargs parsing.
> No argument in net/null could accept such string.
> 
Not a path string, no, though it can take other parameters. Do we
specifically want/need it to take a path string?

/Bruce

Re: [PATCH 06/10] test/devargs: check driver presence with public API

[David Marchand]

On Fri, 17 Jul 2026 at 12:15, Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> On Fri, Jul 17, 2026 at 12:08:01PM +0200, David Marchand wrote:
> > On Fri, 17 Jul 2026 at 11:56, Bruce Richardson
> > <bruce.richardson@intel.com> wrote:
> > >
> > > On Fri, Jul 17, 2026 at 11:30:01AM +0200, David Marchand wrote:
> > > > Let's avoid calling internal drivers API.
> > > >
> > > > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > > > ---
> > > Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> > >
> > > One suggestion inline below.
> > >
> > > >  app/test/test_devargs.c | 10 +++++++---
> > > >  1 file changed, 7 insertions(+), 3 deletions(-)
> > > >
> > > > diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
> > > > index 0bd14c24ec..6c194ecfc6 100644
> > > > --- a/app/test/test_devargs.c
> > > > +++ b/app/test/test_devargs.c
> > > > @@ -10,7 +10,7 @@
> > > >  #include <rte_common.h>
> > > >  #include <rte_devargs.h>
> > > >  #include <rte_kvargs.h>
> > > > -#include <bus_driver.h>
> > > > +#include <rte_bus.h>
> > > >  #include <rte_class.h>
> > > >
> > > >  #include "test.h"
> > > > @@ -167,14 +167,18 @@ test_valid_devargs(void)
> > > >               { "net_ring0,iface=test,path=/class/bus/,queues=1",
> > > >                 0, 0, 3, "vdev", "net_ring0", NULL },
> > > >       };
> > > > -     struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
> > > > +     struct rte_devargs da;
> > > >       int ret;
> > > >
> > > >       ret = test_valid_devargs_cases(list, RTE_DIM(list));
> > > > -     if (vdev_bus != NULL && vdev_bus->parse("net_ring0", NULL) == 0)
> > > > +
> > > > +     memset(&da, 0, sizeof(da));
> > > > +     if (rte_devargs_parse(&da, "net_ring0") == 0)
> > >
> > > From what I/AI can see, there is nothing in this test case that requires
> >
> > (you could write it as "From what A?I can see")
> >
> > > the ring PMD specifically. I think it would be good if it were updated to
> > > use net/null, so that we standardize on our basic unit tests only relying
> > > upon that PMD, rather than requiring a variety of drivers. WDYT?
> >
> > There may be a hidden (capillotracté) reason.
> > Passing path=/class/bus/ seems to be intended at catching errors in
> > devargs parsing.
> > No argument in net/null could accept such string.
> >
> Not a path string, no, though it can take other parameters. Do we
> specifically want/need it to take a path string?

This is some special case tested for "legacy".
I prefer asking Xueming and Thomas what the intention was.


-- 
David Marchand

Re: [PATCH 06/10] test/devargs: check driver presence with public API

[Thomas Monjalon]

17/07/2026 14:34, David Marchand:
> On Fri, 17 Jul 2026 at 12:15, Bruce Richardson
> <bruce.richardson@intel.com> wrote:
> >
> > On Fri, Jul 17, 2026 at 12:08:01PM +0200, David Marchand wrote:
> > > On Fri, 17 Jul 2026 at 11:56, Bruce Richardson
> > > <bruce.richardson@intel.com> wrote:
> > > >
> > > > On Fri, Jul 17, 2026 at 11:30:01AM +0200, David Marchand wrote:
> > > > > Let's avoid calling internal drivers API.
> > > > >
> > > > > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > > > > ---
> > > > Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> > > >
> > > > One suggestion inline below.
> > > >
> > > > >  app/test/test_devargs.c | 10 +++++++---
> > > > >  1 file changed, 7 insertions(+), 3 deletions(-)
> > > > >
> > > > > diff --git a/app/test/test_devargs.c b/app/test/test_devargs.c
> > > > > index 0bd14c24ec..6c194ecfc6 100644
> > > > > --- a/app/test/test_devargs.c
> > > > > +++ b/app/test/test_devargs.c
> > > > > @@ -10,7 +10,7 @@
> > > > >  #include <rte_common.h>
> > > > >  #include <rte_devargs.h>
> > > > >  #include <rte_kvargs.h>
> > > > > -#include <bus_driver.h>
> > > > > +#include <rte_bus.h>
> > > > >  #include <rte_class.h>
> > > > >
> > > > >  #include "test.h"
> > > > > @@ -167,14 +167,18 @@ test_valid_devargs(void)
> > > > >               { "net_ring0,iface=test,path=/class/bus/,queues=1",
> > > > >                 0, 0, 3, "vdev", "net_ring0", NULL },
> > > > >       };
> > > > > -     struct rte_bus *vdev_bus = rte_bus_find_by_name("vdev");
> > > > > +     struct rte_devargs da;
> > > > >       int ret;
> > > > >
> > > > >       ret = test_valid_devargs_cases(list, RTE_DIM(list));
> > > > > -     if (vdev_bus != NULL && vdev_bus->parse("net_ring0", NULL) == 0)
> > > > > +
> > > > > +     memset(&da, 0, sizeof(da));
> > > > > +     if (rte_devargs_parse(&da, "net_ring0") == 0)
> > > >
> > > > From what I/AI can see, there is nothing in this test case that requires
> > >
> > > (you could write it as "From what A?I can see")
> > >
> > > > the ring PMD specifically. I think it would be good if it were updated to
> > > > use net/null, so that we standardize on our basic unit tests only relying
> > > > upon that PMD, rather than requiring a variety of drivers. WDYT?
> > >
> > > There may be a hidden (capillotracté) reason.
> > > Passing path=/class/bus/ seems to be intended at catching errors in
> > > devargs parsing.
> > > No argument in net/null could accept such string.
> > >
> > Not a path string, no, though it can take other parameters. Do we
> > specifically want/need it to take a path string?
> 
> This is some special case tested for "legacy".
> I prefer asking Xueming and Thomas what the intention was.

I have no idea about the original intent.
If it's unclear, it means it can be removed :-)

RE: [PATCH 01/10] bbdev: fix stats aggregation from queues

[Chautru, Nicolas]

Hi David, 
We use the same structure for device and queue level stats for simplicity.
But really bringing these 2 at device level as an aggregate level do not really make much sense at system evel (ie sum of of queue depth is irrelevant). The other ones are relevant. 
I don't think we should change this, it would just create confusion with users. That's a nack to me, let me know if unclear. 
Thanks
Nic


> -----Original Message-----
> From: David Marchand <david.marchand@redhat.com>
> Sent: Friday, July 17, 2026 2:30 AM
> To: dev@dpdk.org
> Cc: stable@dpdk.org; Chautru, Nicolas <nicolas.chautru@intel.com>; Maxime
> Coquelin <maxime.coquelin@redhat.com>; Akhil Goyal <gakhil@marvell.com>
> Subject: [PATCH 01/10] bbdev: fix stats aggregation from queues
> 
> The device stats retrieval was missing aggregation of enqueue_status_count,
> acc_offload_cycles, and enqueue_depth_avail from per-queue statistics.
> 
> Fixes: 4f08028c5e24 ("bbdev: expose queue related warning and status")
> Cc: stable@dpdk.org
> 
> Signed-off-by: David Marchand <david.marchand@redhat.com>
> ---
>  lib/bbdev/rte_bbdev.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/bbdev/rte_bbdev.c b/lib/bbdev/rte_bbdev.c index
> ea644c1f9b..90095578cb 100644
> --- a/lib/bbdev/rte_bbdev.c
> +++ b/lib/bbdev/rte_bbdev.c
> @@ -745,7 +745,7 @@ rte_bbdev_queue_stop(uint16_t dev_id, uint16_t
> queue_id)  static void  get_stats_from_queues(struct rte_bbdev *dev, struct
> rte_bbdev_stats *stats)  {
> -	unsigned int q_id;
> +	unsigned int i, q_id;
>  	for (q_id = 0; q_id < dev->data->num_queues; q_id++) {
>  		struct rte_bbdev_stats *q_stats =
>  				&dev->data->queues[q_id].queue_stats;
> @@ -756,6 +756,10 @@ get_stats_from_queues(struct rte_bbdev *dev, struct
> rte_bbdev_stats *stats)
>  		stats->dequeue_err_count += q_stats->dequeue_err_count;
>  		stats->enqueue_warn_count += q_stats->enqueue_warn_count;
>  		stats->dequeue_warn_count += q_stats->dequeue_warn_count;
> +		for (i = 0; i < RTE_BBDEV_ENQ_STATUS_SIZE_MAX; i++)
> +			stats->enqueue_status_count[i] += q_stats-
> >enqueue_status_count[i];
> +		stats->acc_offload_cycles += q_stats->acc_offload_cycles;
> +		stats->enqueue_depth_avail += q_stats->enqueue_depth_avail;
>  	}
>  	rte_bbdev_log_debug("Got stats on %u", dev->data->dev_id);  }
> --
> 2.54.0