patch 'common/mlx5: add device duplication function' has been queued to stable release 24.11.2

[Kevin Traynor]

Hi,

FYI, your patch has been queued to stable release 24.11.2

Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
It will be pushed if I get no objections before 03/28/25. So please
shout if anyone has objections.

Also note that after the patch there's a diff of the upstream commit vs the
patch applied to the branch. This will indicate if there was any rebasing
needed to apply to the stable branch. If there were code changes for rebasing
(ie: not only metadata diffs), please double check that the rebase was
correctly done.

Queued patches are on a temporary branch at:
https://github.com/kevintraynor/dpdk-stable

This queued commit can be viewed at:
https://github.com/kevintraynor/dpdk-stable/commit/b3c8347553fb62b70ec4fff07b71c1667a59acf9

Thanks.

Kevin

---
>From b3c8347553fb62b70ec4fff07b71c1667a59acf9 Mon Sep 17 00:00:00 2001
From: Michael Baum <michaelba at nvidia.com>
Date: Thu, 13 Mar 2025 10:33:50 +0200
Subject: [PATCH] common/mlx5: add device duplication function

[ upstream commit 786cd5b0fac36386695627a102d58deb7b3552f8 ]

Add function for creating additional CTX for same device base on
existing CTX.

Signed-off-by: Michael Baum <michaelba at nvidia.com>
Acked-by: Matan Azrad <matan at nvidia.com>
---
 drivers/common/mlx5/linux/mlx5_common_os.c   | 52 ++++++++++++++++++++
 drivers/common/mlx5/mlx5_common.h            |  4 ++
 drivers/common/mlx5/version.map              |  1 +
 drivers/common/mlx5/windows/mlx5_common_os.c | 31 ++++++++++++
 4 files changed, 88 insertions(+)

diff --git a/drivers/common/mlx5/linux/mlx5_common_os.c b/drivers/common/mlx5/linux/mlx5_common_os.c
index 41345e1597..2f5032f0da 100644
--- a/drivers/common/mlx5/linux/mlx5_common_os.c
+++ b/drivers/common/mlx5/linux/mlx5_common_os.c
@@ -874,4 +874,56 @@ mlx5_os_open_device(struct mlx5_common_device *cdev, uint32_t classes)
 }
 
+/**
+ * API function to obtain a new InfiniBand (IB) context for a given common device.
+ *
+ * This function provides a port-agnostic IB context for a physical device, enabling the
+ * device to create and manage resources that can be initialized when a port starts and
+ * released when another port stops.
+ *
+ * For Linux, it imports new context from the existing context.
+ *
+ * @param cdev
+ *   Pointer to the mlx5 device structure.
+ *
+ * @return
+ *   Pointer to an `ibv_context` on success, or NULL on failure, with `rte_errno` set.
+ */
+void *
+mlx5_os_get_physical_device_ctx(struct mlx5_common_device *cdev)
+{
+	struct ibv_context *ctx = NULL;
+	int cmd_fd = ((struct ibv_context *)cdev->ctx)->cmd_fd;
+	int new_cmd_fd;
+
+	/*
+	 * Duplicate the command FD to pass it as input to the import device function.
+	 * If the import function succeeds, the new device context takes ownership of
+	 * this FD, which will be freed when the new device is closed.
+	 * If the import function fails, we are responsible for closing this FD.
+	 */
+	new_cmd_fd = dup(cmd_fd);
+	if (new_cmd_fd < 0) {
+		DRV_LOG(ERR,
+			"Failed to duplicate FD %d for IB device \"%s\": %s",
+			cmd_fd, mlx5_os_get_ctx_device_name(cdev->ctx),
+			rte_strerror(errno));
+		rte_errno = errno;
+		return NULL;
+	}
+	/* Attempt to import the duplicated FD to create a new device context. */
+	ctx = mlx5_glue->import_device(new_cmd_fd);
+	if (!ctx) {
+		DRV_LOG(ERR, "Failed to import IB device \"%s\": %s",
+			mlx5_os_get_ctx_device_name(cdev->ctx),
+			rte_strerror(errno));
+		close(new_cmd_fd);
+		rte_errno = errno;
+		return NULL;
+	}
+	DRV_LOG(INFO, "IB device \"%s\" successfully imported, old_fd=%d, new_fd=%d",
+		mlx5_os_get_ctx_device_name(cdev->ctx), cmd_fd, new_cmd_fd);
+	return (void *)ctx;
+}
+
 int
 mlx5_get_device_guid(const struct rte_pci_addr *dev, uint8_t *guid, size_t len)
diff --git a/drivers/common/mlx5/mlx5_common.h b/drivers/common/mlx5/mlx5_common.h
index 1abd1e8239..304cabcc7e 100644
--- a/drivers/common/mlx5/mlx5_common.h
+++ b/drivers/common/mlx5/mlx5_common.h
@@ -636,4 +636,8 @@ mlx5_devx_uar_release(struct mlx5_uar *uar);
 /* mlx5_common_os.c */
 
+__rte_internal
+void *
+mlx5_os_get_physical_device_ctx(struct mlx5_common_device *cdev);
+
 int mlx5_os_open_device(struct mlx5_common_device *cdev, uint32_t classes);
 int mlx5_os_pd_prepare(struct mlx5_common_device *cdev);
diff --git a/drivers/common/mlx5/version.map b/drivers/common/mlx5/version.map
index a2f72ef46a..6311b27c8a 100644
--- a/drivers/common/mlx5/version.map
+++ b/drivers/common/mlx5/version.map
@@ -148,4 +148,5 @@ INTERNAL {
 	mlx5_nl_vlan_vmwa_delete; # WINDOWS_NO_EXPORT
 
+	mlx5_os_get_physical_device_ctx;
 	mlx5_os_umem_dereg;
 	mlx5_os_umem_reg;
diff --git a/drivers/common/mlx5/windows/mlx5_common_os.c b/drivers/common/mlx5/windows/mlx5_common_os.c
index 2b6058eb3e..68f1981193 100644
--- a/drivers/common/mlx5/windows/mlx5_common_os.c
+++ b/drivers/common/mlx5/windows/mlx5_common_os.c
@@ -267,4 +267,35 @@ error:
 }
 
+/**
+ * API function to obtain a new MLX5 context for a given common device.
+ *
+ * This function provides a port-agnostic context for a physical device, enabling the
+ * device to create and manage resources that can be initialized when a port starts and
+ * released when another port stops.
+ *
+ * For Windows, it creates a new context for the device regardless to existing context.
+ *
+ * @param cdev
+ *   Pointer to the mlx5 device structure.
+ *
+ * @return
+ *   Pointer to an `ibv_context` on success, or NULL on failure, with `rte_errno` set.
+ */
+void *
+mlx5_os_get_physical_device_ctx(struct mlx5_common_device *cdev)
+{
+	struct mlx5_common_device temp = {
+		.dev = cdev->dev,
+	};
+
+	if (mlx5_os_open_device(&temp, MLX5_CLASS_ETH) < 0) {
+		DRV_LOG(ERR, "Failed to duplicate DevX device \"%s\": %s",
+			mlx5_os_get_ctx_device_name(cdev->ctx),
+			rte_strerror(rte_errno));
+		return NULL;
+	}
+	return (void *)temp.ctx;
+}
+
 /**
  * Register umem.
-- 
2.48.1

---
  Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- -	2025-03-24 16:15:15.639304331 +0000
+++ 0022-common-mlx5-add-device-duplication-function.patch	2025-03-24 16:15:14.847735896 +0000
@@ -1 +1 @@
-From 786cd5b0fac36386695627a102d58deb7b3552f8 Mon Sep 17 00:00:00 2001
+From b3c8347553fb62b70ec4fff07b71c1667a59acf9 Mon Sep 17 00:00:00 2001
@@ -5,0 +6,2 @@
+[ upstream commit 786cd5b0fac36386695627a102d58deb7b3552f8 ]
+
@@ -80 +82 @@
-index e7bd4c6ec4..bea1382911 100644
+index 1abd1e8239..304cabcc7e 100644
@@ -83 +85 @@
-@@ -652,4 +652,8 @@ mlx5_devx_uar_release(struct mlx5_uar *uar);
+@@ -636,4 +636,8 @@ mlx5_devx_uar_release(struct mlx5_uar *uar);
@@ -93 +95 @@
-index 8301485839..819e6b96cb 100644
+index a2f72ef46a..6311b27c8a 100644
@@ -96,2 +98,2 @@
-@@ -151,4 +151,5 @@ INTERNAL {
- 	mlx5_nl_rdma_monitor_cap_get; # WINDOWS_NO_EXPORT
+@@ -148,4 +148,5 @@ INTERNAL {
+ 	mlx5_nl_vlan_vmwa_delete; # WINDOWS_NO_EXPORT