common/mlx5: support auxiliary bus

Add auxiliary bus support for Sub-Function.

As a limitation of current driver, NUMA node of device is detected
from PCI bus of device sysfs symbol link.
It will be removed once NUMA node file will be available in sysfs.

Signed-off-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>

drivers/common/mlx5/linux/meson.build

@@ -46,12 +46,15 @@ if static_ibverbs
 endif
 
 sources += files('mlx5_nl.c')
+sources += files('mlx5_common_auxiliary.c')
 sources += files('mlx5_common_os.c')
 sources += files('mlx5_common_verbs.c')
 if not dlopen_ibverbs
     sources += files('mlx5_glue.c')
 endif
 
+deps += ['bus_auxiliary']
+
 # To maintain the compatibility with the make build system
 # mlx5_autoconf.h file is still generated.
 # input array for meson member search:

drivers/common/mlx5/linux/mlx5_common_auxiliary.c

@@ -0,0 +1,175 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2020 Mellanox Technologies Ltd
+ */
+
+#include <stdlib.h>
+#include <dirent.h>
+#include <libgen.h>
+
+#include <rte_malloc.h>
+#include <rte_errno.h>
+#include <rte_bus_auxiliary.h>
+#include <rte_common.h>
+#include "eal_filesystem.h"
+
+#include "mlx5_common_utils.h"
+#include "mlx5_common_private.h"
+
+#define AUXILIARY_SYSFS_PATH "/sys/bus/auxiliary/devices"
+#define MLX5_AUXILIARY_PREFIX "mlx5_core.sf."
+
+int
+mlx5_auxiliary_get_child_name(const char *dev, const char *node,
+			      char *child, size_t size)
+{
+	DIR *dir;
+	struct dirent *dent;
+	MKSTR(path, "%s/%s%s", AUXILIARY_SYSFS_PATH, dev, node);
+
+	dir = opendir(path);
+	if (dir == NULL) {
+		rte_errno = errno;
+		return -rte_errno;
+	}
+	/* Get the first file name. */
+	while ((dent = readdir(dir)) != NULL) {
+		if (dent->d_name[0] != '.')
+			break;
+	}
+	closedir(dir);
+	if (dent == NULL) {
+		rte_errno = ENOENT;
+		return -rte_errno;
+	}
+	if (rte_strscpy(child, dent->d_name, size) < 0)
+		return -rte_errno;
+	return 0;
+}
+
+static int
+mlx5_auxiliary_get_pci_path(const struct rte_auxiliary_device *dev,
+			    char *sysfs_pci, size_t size)
+{
+	char sysfs_real[PATH_MAX] = { 0 };
+	MKSTR(sysfs_aux, "%s/%s", AUXILIARY_SYSFS_PATH, dev->name);
+	char *dir;
+
+	if (realpath(sysfs_aux, sysfs_real) == NULL) {
+		rte_errno = errno;
+		return -rte_errno;
+	}
+	dir = dirname(sysfs_real);
+	if (dir == NULL) {
+		rte_errno = errno;
+		return -rte_errno;
+	}
+	if (rte_strscpy(sysfs_pci, dir, size) < 0)
+		return -rte_errno;
+	return 0;
+}
+
+static int
+mlx5_auxiliary_get_numa(const struct rte_auxiliary_device *dev)
+{
+	unsigned long numa;
+	char numa_path[PATH_MAX];
+
+	if (mlx5_auxiliary_get_pci_path(dev, numa_path, sizeof(numa_path)) != 0)
+		return SOCKET_ID_ANY;
+	if (strcat(numa_path, "/numa_node") == NULL) {
+		rte_errno = ENAMETOOLONG;
+		return SOCKET_ID_ANY;
+	}
+	if (eal_parse_sysfs_value(numa_path, &numa) != 0) {
+		rte_errno = EINVAL;
+		return SOCKET_ID_ANY;
+	}
+	return (int)numa;
+}
+
+struct ibv_device *
+mlx5_get_aux_ibv_device(const struct rte_auxiliary_device *dev)
+{
+	int n;
+	char ib_name[64] = { 0 };
+	struct ibv_device **ibv_list = mlx5_glue->get_device_list(&n);
+	struct ibv_device *ibv_match = NULL;
+
+	if (!ibv_list) {
+		rte_errno = ENOSYS;
+		return NULL;
+	}
+	if (mlx5_auxiliary_get_child_name(dev->name, "/infiniband",
+					  ib_name, sizeof(ib_name)) != 0)
+		goto out;
+	while (n-- > 0) {
+		if (strcmp(ibv_list[n]->name, ib_name) != 0)
+			continue;
+		ibv_match = ibv_list[n];
+		break;
+	}
+	if (ibv_match == NULL)
+		rte_errno = ENOENT;
+out:
+	mlx5_glue->free_device_list(ibv_list);
+	return ibv_match;
+}
+
+static bool
+mlx5_common_auxiliary_match(const char *name)
+{
+	return strncmp(name, MLX5_AUXILIARY_PREFIX,
+		       strlen(MLX5_AUXILIARY_PREFIX)) == 0;
+}
+
+static int
+mlx5_common_auxiliary_probe(struct rte_auxiliary_driver *drv __rte_unused,
+			    struct rte_auxiliary_device *dev)
+{
+	dev->device.numa_node = mlx5_auxiliary_get_numa(dev);
+	return mlx5_common_dev_probe(&dev->device);
+}
+
+static int
+mlx5_common_auxiliary_remove(struct rte_auxiliary_device *auxiliary_dev)
+{
+	return mlx5_common_dev_remove(&auxiliary_dev->device);
+}
+
+static int
+mlx5_common_auxiliary_dma_map(struct rte_auxiliary_device *auxiliary_dev,
+			      void *addr, uint64_t iova, size_t len)
+{
+	return mlx5_common_dev_dma_map(&auxiliary_dev->device, addr, iova, len);
+}
+
+static int
+mlx5_common_auxiliary_dma_unmap(struct rte_auxiliary_device *auxiliary_dev,
+				void *addr, uint64_t iova, size_t len)
+{
+	return mlx5_common_dev_dma_unmap(&auxiliary_dev->device, addr, iova,
+					 len);
+}
+
+static struct rte_auxiliary_driver mlx5_auxiliary_driver = {
+	.driver = {
+		   .name = MLX5_AUXILIARY_DRIVER_NAME,
+	},
+	.match = mlx5_common_auxiliary_match,
+	.probe = mlx5_common_auxiliary_probe,
+	.remove = mlx5_common_auxiliary_remove,
+	.dma_map = mlx5_common_auxiliary_dma_map,
+	.dma_unmap = mlx5_common_auxiliary_dma_unmap,
+};
+
+void mlx5_common_auxiliary_init(void)
+{
+	if (mlx5_auxiliary_driver.bus == NULL)
+		rte_auxiliary_register(&mlx5_auxiliary_driver);
+}
+
+RTE_FINI(mlx5_common_auxiliary_driver_finish)
+{
+	if (mlx5_auxiliary_driver.bus != NULL)
+		rte_auxiliary_unregister(&mlx5_auxiliary_driver);
+}

drivers/common/mlx5/linux/mlx5_common_verbs.c

@@ -12,6 +12,7 @@
 
 #include <rte_errno.h>
 #include <rte_bus_pci.h>
+#include <rte_bus_auxiliary.h>
 
 #include "mlx5_common_utils.h"
 #include "mlx5_common_log.h"
@@ -24,10 +25,12 @@
 struct ibv_device *
 mlx5_os_get_ibv_dev(const struct rte_device *dev)
 {
-	struct ibv_device *ibv = NULL;
+	struct ibv_device *ibv;
 
 	if (mlx5_dev_is_pci(dev))
 		ibv = mlx5_os_get_ibv_device(&RTE_DEV_TO_PCI_CONST(dev)->addr);
+	else
+		ibv = mlx5_get_aux_ibv_device(RTE_DEV_TO_AUXILIARY_CONST(dev));
 	if (ibv == NULL) {
 		rte_errno = ENODEV;
 		DRV_LOG(ERR, "Verbs device not found: %s", dev->name);

drivers/common/mlx5/meson.build

@@ -7,7 +7,7 @@ if not (is_linux or (is_windows and is_ms_linker))
     subdir_done()
 endif
 
-deps += ['hash', 'pci', 'bus_pci', 'net', 'eal', 'kvargs']
+deps += ['hash', 'pci', 'bus_pci', 'bus_auxiliary', 'net', 'eal', 'kvargs']
 sources += files(
         'mlx5_devx_cmds.c',
         'mlx5_common.c',

drivers/common/mlx5/mlx5_common.c

@@ -398,6 +398,9 @@ mlx5_class_driver_register(struct mlx5_class_driver *driver)
 static void mlx5_common_driver_init(void)
 {
 	mlx5_common_pci_init();
+#ifdef RTE_EXEC_ENV_LINUX
+	mlx5_common_auxiliary_init();
+#endif
 }
 
 static bool mlx5_common_initialized;

drivers/common/mlx5/mlx5_common.h

@@ -24,6 +24,7 @@
 
 /* Reported driver name. */
 #define MLX5_PCI_DRIVER_NAME "mlx5_pci"
+#define MLX5_AUXILIARY_DRIVER_NAME "mlx5_auxiliary"
 
 /* Bit-field manipulation. */
 #define BITFIELD_DECLARE(bf, type, size) \
@@ -109,6 +110,7 @@ pmd_drv_log_basename(const char *s)
 	int mkstr_size_##name = snprintf(NULL, 0, "" __VA_ARGS__); \
 	char name[mkstr_size_##name + 1]; \
 	\
+	memset(name, 0, mkstr_size_##name + 1); \
 	snprintf(name, sizeof(name), "" __VA_ARGS__)
 
 enum {
@@ -235,6 +237,9 @@ int mlx5_dev_to_pci_addr(const char *dev_path, struct rte_pci_addr *pci_addr);
 __rte_internal
 int mlx5_get_ifname_sysfs(const char *ibdev_path, char *ifname);
 
+__rte_internal
+int mlx5_auxiliary_get_child_name(const char *dev, const char *node,
+				  char *child, size_t size);
 
 enum mlx5_class {
 	MLX5_CLASS_INVALID,

drivers/common/mlx5/mlx5_common_private.h

@@ -6,6 +6,7 @@
 #define MLX5_COMMON_PRIVATE_H
 
 #include <rte_pci.h>
+#include <rte_bus_auxiliary.h>
 
 #include "mlx5_common.h"
 
@@ -34,6 +35,11 @@ void mlx5_common_driver_on_register_pci(struct mlx5_class_driver *driver);
 bool mlx5_dev_pci_match(const struct mlx5_class_driver *drv,
 			const struct rte_device *dev);
 
+/* Common auxiliary bus driver: */
+void mlx5_common_auxiliary_init(void);
+struct ibv_device *mlx5_get_aux_ibv_device(
+		const struct rte_auxiliary_device *dev);
+
 #ifdef __cplusplus
 }
 #endif /* __cplusplus */

drivers/common/mlx5/version.map

@@ -3,6 +3,8 @@ INTERNAL {
 
 	haswell_broadwell_cpu;
 
+	mlx5_auxiliary_get_child_name; # WINDOWS_NO_EXPORT
+
 	mlx5_class_driver_register;
 
 	mlx5_common_init;