All of lore.kernel.org
 help / color / mirror / Atom feed
* [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus
@ 2020-05-04 16:43 Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 2/9] ice: Create and register virtual bus for RDMA Jeff Kirsher
                   ` (7 more replies)
  0 siblings, 8 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

This is the initial implementation of the Virtual Bus,
virtbus_device and virtbus_driver.  The virtual bus is
a software based bus intended to support registering
virtbus_devices and virtbus_drivers and provide matching
between them and probing of the registered drivers.

The bus will support probe/remove shutdown and
suspend/resume callbacks.

Kconfig and Makefile alterations are included

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 Documentation/driver-api/virtual_bus.rst |  88 ++++++++
 drivers/bus/Kconfig                      |  10 +
 drivers/bus/Makefile                     |   2 +
 drivers/bus/virtual_bus.c                | 255 +++++++++++++++++++++++
 include/linux/mod_devicetable.h          |   8 +
 include/linux/virtual_bus.h              |  62 ++++++
 scripts/mod/devicetable-offsets.c        |   3 +
 scripts/mod/file2alias.c                 |   7 +
 8 files changed, 435 insertions(+)
 create mode 100644 Documentation/driver-api/virtual_bus.rst
 create mode 100644 drivers/bus/virtual_bus.c
 create mode 100644 include/linux/virtual_bus.h

diff --git a/Documentation/driver-api/virtual_bus.rst b/Documentation/driver-api/virtual_bus.rst
new file mode 100644
index 000000000000..33cef5433dcd
--- /dev/null
+++ b/Documentation/driver-api/virtual_bus.rst
@@ -0,0 +1,88 @@
+===============================
+Virtual Bus Devices and Drivers
+===============================
+
+See <linux/virtual_bus.h> for the models for virtbus_device and virtbus_driver.
+This bus is meant to be a minimalist software based bus to attach generic
+devices and drivers to so that a chunk of data can be passed between them.
+
+Memory Allocation Lifespan and Model
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Originating KO - The kernel object that is defining either the
+virtbus_device or the virtbus_driver and registering it with the virtual_bus.
+
+The originating KO is expected to allocate memory for the virtbus_device or
+virtbus_driver before registering it with the virtual_bus.
+
+For a virtbus_device, this memory is expected to remain viable until the
+device's mandatory .release() callback is accessed.  The memory cleanup will
+ideally be performed within this callback.
+
+For a virtbus_driver, this memory is expected to remain viable until the
+driver's .remove() or .shutdown() callbacks are accessed.  The memory cleanup
+will ideally be performed with these callbacks.
+
+Device Enumeration
+~~~~~~~~~~~~~~~~~~
+
+Enumeration is handled automatically by the bus infrastructure.
+
+Device naming and driver binding
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The virtbus_device.dev.name is the canonical name for the device. It is
+built from two other parts:
+
+        - virtbus_device.match_name (used for matching).
+        - virtbus_device.id (generated automatically from ida_simple calls)
+
+Virtbus device's sub-device names are always in "<name>.<instance>" format.
+
+For a virtbus_device to be matched with a virtbus_driver, the device's match
+name must be populated, as this is what will be evaluated to perform the match.
+
+Common Usage and Structure Design
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The two (or more) Originating KO's (one for the driver and one or more for
+the devices) will have to have a common header file, so that a struct can
+be defined that they both use.  This allows for the following design.
+
+In the common header file outside of the virtual_bus infrastructure, define
+struct virtbus_object:
+
+                 virtbus_object
+                        |
+                -------------------------
+                |                       |
+        virtbus_device          private_struct/data
+
+The virtbus_device cannot be a pointer so that a container_of can be
+performed by the driver.
+
+The device originating KO will allocate a virtbus_object and then populate
+virtbus_device->match_name and register the virtbus_device with virtbus.
+
+When the virtbus_driver registers with the virtual_bus and a match is made,
+the .probe() callback of the driver will be called with the virtbus_device
+struct as a parameter.  This allows the virtbus_driver to perform a
+contain_of call to get access to the virtbus_object, and to the pointer to
+the private_struct/data.
+
+This is the main goal of virtual_bus, to get a common pointer available
+to both of the originating KOs involved in a match.
+
+Mandatory Elements
+~~~~~~~~~~~~~~~~~
+
+virtbus_device:
+        .relase() callback must not be NULL and is expected to perform
+                memory cleanup.
+        .match_name must be populated to be able to match with a driver
+
+virtbus_driver:
+        .probe() callback must not be NULL
+        .remove() callback must not be NULL
+        .shutdown() callback must not be NULL
+        .id_table must not be NULL, used to perform matching
diff --git a/drivers/bus/Kconfig b/drivers/bus/Kconfig
index 6d4e4497b59b..00553c78510c 100644
--- a/drivers/bus/Kconfig
+++ b/drivers/bus/Kconfig
@@ -203,4 +203,14 @@ config DA8XX_MSTPRI
 source "drivers/bus/fsl-mc/Kconfig"
 source "drivers/bus/mhi/Kconfig"
 
+config VIRTUAL_BUS
+       tristate "Software based Virtual Bus"
+       help
+         Provides a software bus for virtbus_devices to be added to it
+         and virtbus_drivers to be registered on it.  It matches driver
+         and device based on id and calls the driver's probe routine.
+         One example is the irdma driver needing to connect with various
+         PCI LAN drivers to request resources (queues) to be able to perform
+         its function.
+
 endmenu
diff --git a/drivers/bus/Makefile b/drivers/bus/Makefile
index 05f32cd694a4..d30828a4768c 100644
--- a/drivers/bus/Makefile
+++ b/drivers/bus/Makefile
@@ -37,3 +37,5 @@ obj-$(CONFIG_DA8XX_MSTPRI)	+= da8xx-mstpri.o
 
 # MHI
 obj-$(CONFIG_MHI_BUS)		+= mhi/
+
+obj-$(CONFIG_VIRTUAL_BUS)	+= virtual_bus.o
diff --git a/drivers/bus/virtual_bus.c b/drivers/bus/virtual_bus.c
new file mode 100644
index 000000000000..fe8000a53a0a
--- /dev/null
+++ b/drivers/bus/virtual_bus.c
@@ -0,0 +1,255 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * virtual_bus.c - lightweight software based bus for virtual devices
+ *
+ * Copyright (c) 2019-2020 Intel Corporation
+ *
+ * Please see Documentation/driver-api/virtual_bus.rst for
+ * more information
+ */
+
+#include <linux/string.h>
+#include <linux/virtual_bus.h>
+#include <linux/of_irq.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pm_runtime.h>
+#include <linux/pm_domain.h>
+#include <linux/acpi.h>
+#include <linux/device.h>
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("Virtual Bus");
+MODULE_AUTHOR("David Ertman <david.m.ertman@intel.com>");
+MODULE_AUTHOR("Kiran Patil <kiran.patil@intel.com>");
+
+static DEFINE_IDA(virtbus_dev_ida);
+#define VIRTBUS_INVALID_ID	0xFFFFFFFF
+
+static const
+struct virtbus_dev_id *virtbus_match_id(const struct virtbus_dev_id *id,
+					struct virtbus_device *vdev)
+{
+	while (id->name[0]) {
+		if (!strcmp(vdev->match_name, id->name))
+			return id;
+		id++;
+	}
+	return NULL;
+}
+
+static int virtbus_match(struct device *dev, struct device_driver *drv)
+{
+	struct virtbus_driver *vdrv = to_virtbus_drv(drv);
+	struct virtbus_device *vdev = to_virtbus_dev(dev);
+
+	return virtbus_match_id(vdrv->id_table, vdev) != NULL;
+}
+
+static int virtbus_probe(struct device *dev)
+{
+	return dev->driver->probe(dev);
+}
+
+static int virtbus_remove(struct device *dev)
+{
+	return dev->driver->remove(dev);
+}
+
+static void virtbus_shutdown(struct device *dev)
+{
+	dev->driver->shutdown(dev);
+}
+
+static int virtbus_suspend(struct device *dev, pm_message_t state)
+{
+	if (!dev->driver->suspend)
+		return 0;
+
+	return dev->driver->suspend(dev, state);
+}
+
+static int virtbus_resume(struct device *dev)
+{
+	if (!dev->driver->resume)
+		return 0;
+
+	return dev->driver->resume(dev);
+}
+
+struct bus_type virtual_bus_type = {
+	.name = "virtbus",
+	.match = virtbus_match,
+	.probe = virtbus_probe,
+	.remove = virtbus_remove,
+	.shutdown = virtbus_shutdown,
+	.suspend = virtbus_suspend,
+	.resume = virtbus_resume,
+};
+
+/**
+ * virtbus_release_device - Destroy a virtbus device
+ * @_dev: device to release
+ */
+static void virtbus_release_device(struct device *_dev)
+{
+	struct virtbus_device *vdev = to_virtbus_dev(_dev);
+	u32 ida = vdev->id;
+
+	vdev->release(vdev);
+	if (ida != VIRTBUS_INVALID_ID)
+		ida_simple_remove(&virtbus_dev_ida, ida);
+}
+
+/**
+ * virtbus_register_device - add a virtual bus device
+ * @vdev: virtual bus device to add
+ */
+int virtbus_register_device(struct virtbus_device *vdev)
+{
+	int ret;
+
+	if (WARN_ON(!vdev->release))
+		return -EINVAL;
+
+	/* All error paths out of this function after the device_initialize
+	 * must perform a put_device() so that the .release() callback is
+	 * called for an error condition.
+	 */
+	device_initialize(&vdev->dev);
+
+	vdev->dev.bus = &virtual_bus_type;
+	vdev->dev.release = virtbus_release_device;
+
+	/* All device IDs are automatically allocated */
+	ret = ida_simple_get(&virtbus_dev_ida, 0, 0, GFP_KERNEL);
+
+	if (ret < 0) {
+		vdev->id = VIRTBUS_INVALID_ID;
+		dev_err(&vdev->dev, "get IDA idx for virtbus device failed!\n");
+		goto device_add_err;
+	}
+
+	vdev->id = ret;
+
+	ret = dev_set_name(&vdev->dev, "%s.%d", vdev->match_name, vdev->id);
+	if (ret) {
+		dev_err(&vdev->dev, "dev_set_name failed for device\n");
+		goto device_add_err;
+	}
+
+	dev_dbg(&vdev->dev, "Registering virtbus device '%s'\n",
+		dev_name(&vdev->dev));
+
+	ret = device_add(&vdev->dev);
+	if (ret)
+		goto device_add_err;
+
+	return 0;
+
+device_add_err:
+	dev_err(&vdev->dev, "Add device to virtbus failed!: %d\n", ret);
+	put_device(&vdev->dev);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(virtbus_register_device);
+
+static int virtbus_probe_driver(struct device *_dev)
+{
+	struct virtbus_driver *vdrv = to_virtbus_drv(_dev->driver);
+	struct virtbus_device *vdev = to_virtbus_dev(_dev);
+	int ret;
+
+	ret = dev_pm_domain_attach(_dev, true);
+	if (ret) {
+		dev_warn(_dev, "Failed to attach to PM Domain : %d\n", ret);
+		return ret;
+	}
+
+	ret = vdrv->probe(vdev);
+	if (ret) {
+		dev_err(&vdev->dev, "Probe returned error\n");
+		dev_pm_domain_detach(_dev, true);
+	}
+
+	return ret;
+}
+
+static int virtbus_remove_driver(struct device *_dev)
+{
+	struct virtbus_driver *vdrv = to_virtbus_drv(_dev->driver);
+	struct virtbus_device *vdev = to_virtbus_dev(_dev);
+	int ret = 0;
+
+	ret = vdrv->remove(vdev);
+	dev_pm_domain_detach(_dev, true);
+
+	return ret;
+}
+
+static void virtbus_shutdown_driver(struct device *_dev)
+{
+	struct virtbus_driver *vdrv = to_virtbus_drv(_dev->driver);
+	struct virtbus_device *vdev = to_virtbus_dev(_dev);
+
+	vdrv->shutdown(vdev);
+}
+
+static int virtbus_suspend_driver(struct device *_dev, pm_message_t state)
+{
+	struct virtbus_driver *vdrv = to_virtbus_drv(_dev->driver);
+	struct virtbus_device *vdev = to_virtbus_dev(_dev);
+
+	if (!vdrv->suspend)
+		return 0;
+
+	return vdrv->suspend(vdev, state);
+}
+
+static int virtbus_resume_driver(struct device *_dev)
+{
+	struct virtbus_driver *vdrv = to_virtbus_drv(_dev->driver);
+	struct virtbus_device *vdev = to_virtbus_dev(_dev);
+
+	if (!vdrv->resume)
+		return 0;
+
+	return vdrv->resume(vdev);
+}
+
+/**
+ * __virtbus_register_driver - register a driver for virtual bus devices
+ * @vdrv: virtbus_driver structure
+ * @owner: owning module/driver
+ */
+int __virtbus_register_driver(struct virtbus_driver *vdrv, struct module *owner)
+{
+	if (!vdrv->probe || !vdrv->remove || !vdrv->shutdown || !vdrv->id_table)
+		return -EINVAL;
+
+	vdrv->driver.owner = owner;
+	vdrv->driver.bus = &virtual_bus_type;
+	vdrv->driver.probe = virtbus_probe_driver;
+	vdrv->driver.remove = virtbus_remove_driver;
+	vdrv->driver.shutdown = virtbus_shutdown_driver;
+	vdrv->driver.suspend = virtbus_suspend_driver;
+	vdrv->driver.resume = virtbus_resume_driver;
+
+	return driver_register(&vdrv->driver);
+}
+EXPORT_SYMBOL_GPL(__virtbus_register_driver);
+
+static int __init virtual_bus_init(void)
+{
+	return bus_register(&virtual_bus_type);
+}
+
+static void __exit virtual_bus_exit(void)
+{
+	bus_unregister(&virtual_bus_type);
+	ida_destroy(&virtbus_dev_ida);
+}
+
+module_init(virtual_bus_init);
+module_exit(virtual_bus_exit);
diff --git a/include/linux/mod_devicetable.h b/include/linux/mod_devicetable.h
index 4c2ddd0941a7..60bcfe75fb94 100644
--- a/include/linux/mod_devicetable.h
+++ b/include/linux/mod_devicetable.h
@@ -832,4 +832,12 @@ struct mhi_device_id {
 	kernel_ulong_t driver_data;
 };
 
+#define VIRTBUS_NAME_SIZE 20
+#define VIRTBUS_MODULE_PREFIX "virtbus:"
+
+struct virtbus_dev_id {
+	char name[VIRTBUS_NAME_SIZE];
+	kernel_ulong_t driver_data;
+};
+
 #endif /* LINUX_MOD_DEVICETABLE_H */
diff --git a/include/linux/virtual_bus.h b/include/linux/virtual_bus.h
new file mode 100644
index 000000000000..4872fd5a9218
--- /dev/null
+++ b/include/linux/virtual_bus.h
@@ -0,0 +1,62 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * virtual_bus.h - lightweight software bus
+ *
+ * Copyright (c) 2019-2020 Intel Corporation
+ *
+ * Please see Documentation/driver-api/virtual_bus.rst for more information
+ */
+
+#ifndef _VIRTUAL_BUS_H_
+#define _VIRTUAL_BUS_H_
+
+#include <linux/device.h>
+
+struct virtbus_device {
+	struct device dev;
+	const char *match_name;
+	void (*release)(struct virtbus_device *);
+	u32 id;
+};
+
+struct virtbus_driver {
+	int (*probe)(struct virtbus_device *);
+	int (*remove)(struct virtbus_device *);
+	void (*shutdown)(struct virtbus_device *);
+	int (*suspend)(struct virtbus_device *, pm_message_t);
+	int (*resume)(struct virtbus_device *);
+	struct device_driver driver;
+	const struct virtbus_dev_id *id_table;
+};
+
+static inline
+struct virtbus_device *to_virtbus_dev(struct device *dev)
+{
+	return container_of(dev, struct virtbus_device, dev);
+}
+
+static inline
+struct virtbus_driver *to_virtbus_drv(struct device_driver *drv)
+{
+	return container_of(drv, struct virtbus_driver, driver);
+}
+
+int virtbus_register_device(struct virtbus_device *vdev);
+
+int
+__virtbus_register_driver(struct virtbus_driver *vdrv, struct module *owner);
+
+#define virtbus_register_driver(vdrv) \
+	__virtbus_register_driver(vdrv, THIS_MODULE)
+
+static inline void virtbus_unregister_device(struct virtbus_device *vdev)
+{
+	device_unregister(&vdev->dev);
+}
+
+static inline void virtbus_unregister_driver(struct virtbus_driver *vdrv)
+{
+	driver_unregister(&vdrv->driver);
+}
+
+#endif /* _VIRTUAL_BUS_H_ */
diff --git a/scripts/mod/devicetable-offsets.c b/scripts/mod/devicetable-offsets.c
index 010be8ba2116..0c8e0e3a7c84 100644
--- a/scripts/mod/devicetable-offsets.c
+++ b/scripts/mod/devicetable-offsets.c
@@ -241,5 +241,8 @@ int main(void)
 	DEVID(mhi_device_id);
 	DEVID_FIELD(mhi_device_id, chan);
 
+	DEVID(virtbus_dev_id);
+	DEVID_FIELD(virtbus_dev_id, name);
+
 	return 0;
 }
diff --git a/scripts/mod/file2alias.c b/scripts/mod/file2alias.c
index 02d5d79da284..7d78fa3fba34 100644
--- a/scripts/mod/file2alias.c
+++ b/scripts/mod/file2alias.c
@@ -1358,7 +1358,13 @@ static int do_mhi_entry(const char *filename, void *symval, char *alias)
 {
 	DEF_FIELD_ADDR(symval, mhi_device_id, chan);
 	sprintf(alias, MHI_DEVICE_MODALIAS_FMT, *chan);
+	return 1;
+}
 
+static int do_virtbus_entry(const char *filename, void *symval, char *alias)
+{
+	DEF_FIELD_ADDR(symval, virtbus_dev_id, name);
+	sprintf(alias, VIRTBUS_MODULE_PREFIX "%s", *name);
 	return 1;
 }
 
@@ -1436,6 +1442,7 @@ static const struct devtable devtable[] = {
 	{"tee", SIZE_tee_client_device_id, do_tee_entry},
 	{"wmi", SIZE_wmi_device_id, do_wmi_entry},
 	{"mhi", SIZE_mhi_device_id, do_mhi_entry},
+	{"virtbus", SIZE_virtbus_dev_id, do_virtbus_entry},
 };
 
 /* Create MODULE_ALIAS() statements.
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 2/9] ice: Create and register virtual bus for RDMA
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 3/9] ice: Complete RDMA peer registration Jeff Kirsher
                   ` (6 subsequent siblings)
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

The RDMA block does not have its own PCI function, instead it must utilize
the ice driver to gain access to the PCI device. Create a virtual bus
device so the irdma driver can register a virtual bus driver to bind to it
and receive device data. The device data contains all of the relevant
information that the irdma peer will need to access this PF's IIDC API
callbacks.

Note the header file iidc.h is located under include/linux/net/intel
as this is a unified header file to be used by all consumers of the
IIDC interface.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 MAINTAINERS                                   |   1 +
 drivers/net/ethernet/intel/Kconfig            |   1 +
 drivers/net/ethernet/intel/ice/Makefile       |   1 +
 drivers/net/ethernet/intel/ice/ice.h          |  12 +
 .../net/ethernet/intel/ice/ice_adminq_cmd.h   |   1 +
 drivers/net/ethernet/intel/ice/ice_common.c   |  18 +-
 drivers/net/ethernet/intel/ice/ice_dcb_lib.c  |  31 ++
 drivers/net/ethernet/intel/ice/ice_dcb_lib.h  |   3 +
 .../net/ethernet/intel/ice/ice_hw_autogen.h   |   1 +
 drivers/net/ethernet/intel/ice/ice_idc.c      | 417 ++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_idc_int.h  |  67 +++
 drivers/net/ethernet/intel/ice/ice_lib.c      |  11 +
 drivers/net/ethernet/intel/ice/ice_lib.h      |   2 +
 drivers/net/ethernet/intel/ice/ice_main.c     |  57 ++-
 drivers/net/ethernet/intel/ice/ice_type.h     |   1 +
 include/linux/net/intel/iidc.h                | 337 ++++++++++++++
 16 files changed, 958 insertions(+), 3 deletions(-)
 create mode 100644 drivers/net/ethernet/intel/ice/ice_idc.c
 create mode 100644 drivers/net/ethernet/intel/ice/ice_idc_int.h
 create mode 100644 include/linux/net/intel/iidc.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 0ac9cec0bce6..95a309bc7a37 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -8555,6 +8555,7 @@ F:	Documentation/networking/device_drivers/intel/ixgbevf.rst
 F:	drivers/net/ethernet/intel/
 F:	drivers/net/ethernet/intel/*/
 F:	include/linux/avf/virtchnl.h
+F:	include/linux/net/intel/iidc.h
 
 INTEL FRAMEBUFFER DRIVER (excluding 810 and 815)
 M:	Maik Broemme <mbroemme@libmpq.org>
diff --git a/drivers/net/ethernet/intel/Kconfig b/drivers/net/ethernet/intel/Kconfig
index ad34e4335df2..1a5d51b0f294 100644
--- a/drivers/net/ethernet/intel/Kconfig
+++ b/drivers/net/ethernet/intel/Kconfig
@@ -272,6 +272,7 @@ config I40EVF
 	tristate "Intel(R) Ethernet Adaptive Virtual Function support"
 	select IAVF
 	depends on PCI_MSI
+	select VIRTUAL_BUS
 	---help---
 	  This driver supports virtual functions for Intel XL710,
 	  X710, X722, XXV710, and all devices advertising support for
diff --git a/drivers/net/ethernet/intel/ice/Makefile b/drivers/net/ethernet/intel/ice/Makefile
index 29c6c6743450..73909045da1c 100644
--- a/drivers/net/ethernet/intel/ice/Makefile
+++ b/drivers/net/ethernet/intel/ice/Makefile
@@ -20,6 +20,7 @@ ice-y := ice_main.o	\
 	 ice_flex_pipe.o \
 	 ice_flow.o	\
 	 ice_devlink.o	\
+	 ice_idc.o	\
 	 ice_ethtool.o
 ice-$(CONFIG_PCI_IOV) += ice_virtchnl_pf.o ice_sriov.o
 ice-$(CONFIG_DCB) += ice_dcb.o ice_dcb_nl.o ice_dcb_lib.o
diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 5c11448bfbb3..73366009ef03 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -44,6 +44,7 @@
 #include "ice_switch.h"
 #include "ice_common.h"
 #include "ice_sched.h"
+#include "ice_idc_int.h"
 #include "ice_virtchnl_pf.h"
 #include "ice_sriov.h"
 #include "ice_xsk.h"
@@ -72,6 +73,8 @@ extern const char ice_drv_ver[];
 #define ICE_MAX_LG_RSS_QS	256
 #define ICE_RES_VALID_BIT	0x8000
 #define ICE_RES_MISC_VEC_ID	(ICE_RES_VALID_BIT - 1)
+#define ICE_RDMA_NUM_VECS	4
+#define ICE_RES_RDMA_VEC_ID	(ICE_RES_MISC_VEC_ID - 1)
 #define ICE_INVAL_Q_INDEX	0xffff
 #define ICE_INVAL_VFID		256
 
@@ -330,11 +333,13 @@ struct ice_q_vector {
 
 enum ice_pf_flags {
 	ICE_FLAG_FLTR_SYNC,
+	ICE_FLAG_IWARP_ENA,
 	ICE_FLAG_RSS_ENA,
 	ICE_FLAG_SRIOV_ENA,
 	ICE_FLAG_SRIOV_CAPABLE,
 	ICE_FLAG_DCB_CAPABLE,
 	ICE_FLAG_DCB_ENA,
+	ICE_FLAG_PEER_ENA,
 	ICE_FLAG_ADV_FEATURES,
 	ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA,
 	ICE_FLAG_NO_MEDIA,
@@ -384,6 +389,8 @@ struct ice_pf {
 	struct mutex sw_mutex;		/* lock for protecting VSI alloc flow */
 	struct mutex tc_mutex;		/* lock to protect TC changes */
 	u32 msg_enable;
+	u32 num_rdma_msix;	/* Total MSIX vectors for RDMA driver */
+	u32 rdma_base_vector;
 	u32 hw_csum_rx_error;
 	u32 oicr_idx;		/* Other interrupt cause MSIX vector index */
 	u32 num_avail_sw_msix;	/* remaining MSIX SW vectors left unclaimed */
@@ -410,6 +417,7 @@ struct ice_pf {
 	unsigned long tx_timeout_last_recovery;
 	u32 tx_timeout_recovery_level;
 	char int_name[ICE_INT_NAME_STR_LEN];
+	struct ice_peer_dev_int **peers;
 	u32 sw_int_count;
 };
 
@@ -523,6 +531,10 @@ int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size);
 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size);
 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset);
 void ice_print_link_msg(struct ice_vsi *vsi, bool isup);
+int ice_init_peer_devices(struct ice_pf *pf);
+int
+ice_for_each_peer(struct ice_pf *pf, void *data,
+		  int (*fn)(struct ice_peer_dev_int *, void *));
 int ice_open(struct net_device *netdev);
 int ice_stop(struct net_device *netdev);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index 2381b4014ed6..51baab0621a2 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -108,6 +108,7 @@ struct ice_aqc_list_caps_elem {
 #define ICE_AQC_CAPS_TXQS				0x0042
 #define ICE_AQC_CAPS_MSIX				0x0043
 #define ICE_AQC_CAPS_MAX_MTU				0x0047
+#define ICE_AQC_CAPS_IWARP				0x0051
 
 	u8 major_ver;
 	u8 minor_ver;
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index 2c0d8fd3d5cd..2dca49aed5bb 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -825,7 +825,8 @@ enum ice_status ice_check_reset(struct ice_hw *hw)
 				 GLNVM_ULD_POR_DONE_1_M |\
 				 GLNVM_ULD_PCIER_DONE_2_M)
 
-	uld_mask = ICE_RESET_DONE_MASK;
+	uld_mask = ICE_RESET_DONE_MASK | (hw->func_caps.common_cap.iwarp ?
+					  GLNVM_ULD_PE_DONE_M : 0);
 
 	/* Device is Active; check Global Reset processes are done */
 	for (cnt = 0; cnt < ICE_PF_RESET_WAIT_COUNT; cnt++) {
@@ -1678,6 +1679,11 @@ ice_parse_caps(struct ice_hw *hw, void *buf, u32 cap_count,
 				  "%s: msix_vector_first_id = %d\n", prefix,
 				  caps->msix_vector_first_id);
 			break;
+		case ICE_AQC_CAPS_IWARP:
+			caps->iwarp = (number == 1);
+			ice_debug(hw, ICE_DBG_INIT,
+				  "%s: iwarp = %d\n", prefix, caps->iwarp);
+			break;
 		case ICE_AQC_CAPS_MAX_MTU:
 			caps->max_mtu = number;
 			ice_debug(hw, ICE_DBG_INIT, "%s: max_mtu = %d\n",
@@ -1701,6 +1707,16 @@ ice_parse_caps(struct ice_hw *hw, void *buf, u32 cap_count,
 		ice_debug(hw, ICE_DBG_INIT,
 			  "%s: maxtc = %d (based on #ports)\n", prefix,
 			  caps->maxtc);
+		if (caps->iwarp) {
+			ice_debug(hw, ICE_DBG_INIT, "%s: forcing RDMA off\n",
+				  prefix);
+			caps->iwarp = 0;
+		}
+
+		/* print message only when processing device capabilities */
+		if (dev_p)
+			dev_info(ice_hw_to_dev(hw),
+				 "RDMA functionality is not available with the current device configuration.\n");
 	}
 }
 
diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
index 7bea09363b42..24c0a60fe172 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
@@ -763,6 +763,37 @@ ice_tx_prepare_vlan_flags_dcb(struct ice_ring *tx_ring,
 	return 0;
 }
 
+/**
+ * ice_setup_dcb_qos_info - Setup DCB QoS information
+ * @pf: ptr to ice_pf
+ * @qos_info: QoS param instance
+ */
+void ice_setup_dcb_qos_info(struct ice_pf *pf, struct iidc_qos_params *qos_info)
+{
+	struct ice_dcbx_cfg *dcbx_cfg;
+	u32 up2tc;
+	int i;
+
+	dcbx_cfg = &pf->hw.port_info->local_dcbx_cfg;
+	up2tc = rd32(&pf->hw, PRTDCB_TUP2TC);
+	qos_info->num_apps = dcbx_cfg->numapps;
+
+	qos_info->num_tc = ice_dcb_get_num_tc(dcbx_cfg);
+
+	for (i = 0; i < IIDC_MAX_USER_PRIORITY; i++)
+		qos_info->up2tc[i] = (up2tc >> (i * 3)) & 0x7;
+
+	for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++)
+		qos_info->tc_info[i].rel_bw =
+			dcbx_cfg->etscfg.tcbwtable[i];
+
+	for (i = 0; i < qos_info->num_apps; i++) {
+		qos_info->apps[i].priority = dcbx_cfg->app[i].priority;
+		qos_info->apps[i].prot_id = dcbx_cfg->app[i].prot_id;
+		qos_info->apps[i].selector = dcbx_cfg->app[i].selector;
+	}
+}
+
 /**
  * ice_dcb_process_lldp_set_mib_change - Process MIB change
  * @pf: ptr to ice_pf
diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.h b/drivers/net/ethernet/intel/ice/ice_dcb_lib.h
index 37680e815b02..11457b6ba145 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_dcb_lib.h
@@ -29,6 +29,8 @@ int
 ice_tx_prepare_vlan_flags_dcb(struct ice_ring *tx_ring,
 			      struct ice_tx_buf *first);
 void
+ice_setup_dcb_qos_info(struct ice_pf *pf, struct iidc_qos_params *qos_info);
+void
 ice_dcb_process_lldp_set_mib_change(struct ice_pf *pf,
 				    struct ice_rq_event_info *event);
 void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc);
@@ -82,6 +84,7 @@ ice_tx_prepare_vlan_flags_dcb(struct ice_ring __always_unused *tx_ring,
 #define ice_update_dcb_stats(pf) do {} while (0)
 #define ice_pf_dcb_recfg(pf) do {} while (0)
 #define ice_vsi_cfg_dcb_rings(vsi) do {} while (0)
+#define ice_setup_dcb_qos_info(pf, qos_info) do {} while (0)
 #define ice_dcb_process_lldp_set_mib_change(pf, event) do {} while (0)
 #define ice_set_cgd_num(tlan_ctx, ring) do {} while (0)
 #define ice_vsi_cfg_netdev_tc(vsi, ena_tc) do {} while (0)
diff --git a/drivers/net/ethernet/intel/ice/ice_hw_autogen.h b/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
index 1d37a9f02c1c..3f40736a8295 100644
--- a/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
+++ b/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
@@ -58,6 +58,7 @@
 #define PRTDCB_GENS				0x00083020
 #define PRTDCB_GENS_DCBX_STATUS_S		0
 #define PRTDCB_GENS_DCBX_STATUS_M		ICE_M(0x7, 0)
+#define PRTDCB_TUP2TC				0x001D26C0
 #define GL_PREEXT_L2_PMASK0(_i)			(0x0020F0FC + ((_i) * 4))
 #define GL_PREEXT_L2_PMASK1(_i)			(0x0020F108 + ((_i) * 4))
 #define GLFLXP_RXDID_FLX_WRD_0(_i)		(0x0045c800 + ((_i) * 4))
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
new file mode 100644
index 000000000000..68d6b524d6d4
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -0,0 +1,417 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2019, Intel Corporation. */
+
+/* Inter-Driver Communication */
+#include <linux/virtual_bus.h>
+#include "ice.h"
+#include "ice_lib.h"
+#include "ice_dcb_lib.h"
+
+static struct peer_dev_id ice_peers[] = ASSIGN_PEER_INFO;
+
+/**
+ * ice_peer_state_change - manage state machine for peer
+ * @peer_dev: pointer to peer's configuration
+ * @new_state: the state requested to transition into
+ * @locked: boolean to determine if call made with mutex held
+ *
+ * Any function that calls this is responsible for verifying that
+ * the peer_dev_int struct is valid and capable of handling a
+ * state change
+ *
+ * This function handles all state transitions for peer devices.
+ * The state machine is as follows:
+ *
+ *     +<-----------------------+<-----------------------------+
+ *				|<-------+<----------+	       +
+ *				\/	 +	     +	       +
+ *    INIT  --------------> PROBED --> OPENING	  CLOSED --> REMOVED
+ *					 +           +
+ *				       OPENED --> CLOSING
+ *					 +	     +
+ *				       PREP_RST	     +
+ *					 +	     +
+ *				      PREPPED	     +
+ *					 +---------->+
+ */
+static void
+ice_peer_state_change(struct ice_peer_dev_int *peer_dev, long new_state,
+		      bool locked)
+{
+	struct device *dev = &peer_dev->peer_dev.vdev->dev;
+
+	if (!locked)
+		mutex_lock(&peer_dev->peer_dev_state_mutex);
+
+	switch (new_state) {
+	case ICE_PEER_DEV_STATE_INIT:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_REMOVED,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_INIT, peer_dev->state);
+			dev_dbg(dev, "state transition from _REMOVED to _INIT\n");
+		} else {
+			set_bit(ICE_PEER_DEV_STATE_INIT, peer_dev->state);
+			if (dev)
+				dev_dbg(dev, "state set to _INIT\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_PROBED:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_INIT,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev->state);
+			dev_dbg(dev, "state transition from _INIT to _PROBED\n");
+		} else if (test_and_clear_bit(ICE_PEER_DEV_STATE_REMOVED,
+					      peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev->state);
+			dev_dbg(dev, "state transition from _REMOVED to _PROBED\n");
+		} else if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENING,
+					      peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev->state);
+			dev_dbg(dev, "state transition from _OPENING to _PROBED\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_OPENING:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_PROBED,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_OPENING, peer_dev->state);
+			dev_dbg(dev, "state transition from _PROBED to _OPENING\n");
+		} else if (test_and_clear_bit(ICE_PEER_DEV_STATE_CLOSED,
+					      peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_OPENING, peer_dev->state);
+			dev_dbg(dev, "state transition from _CLOSED to _OPENING\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_OPENED:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENING,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_OPENED, peer_dev->state);
+			dev_dbg(dev, "state transition from _OPENING to _OPENED\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_PREP_RST:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENED,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_PREP_RST, peer_dev->state);
+			dev_dbg(dev, "state transition from _OPENED to _PREP_RST\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_PREPPED:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_PREP_RST,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_PREPPED, peer_dev->state);
+			dev_dbg(dev, "state transition _PREP_RST to _PREPPED\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_CLOSING:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENED,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev->state);
+			dev_dbg(dev, "state transition from _OPENED to _CLOSING\n");
+		}
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_PREPPED,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev->state);
+			dev_dbg(dev, "state transition _PREPPED to _CLOSING\n");
+		}
+		/* NOTE - up to peer to handle this situation correctly */
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_PREP_RST,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev->state);
+			dev_warn(dev, "WARN: Peer state PREP_RST to _CLOSING\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_CLOSED:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_CLOSING,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_CLOSED, peer_dev->state);
+			dev_dbg(dev, "state transition from _CLOSING to _CLOSED\n");
+		}
+		break;
+	case ICE_PEER_DEV_STATE_REMOVED:
+		if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENED,
+				       peer_dev->state) ||
+		    test_and_clear_bit(ICE_PEER_DEV_STATE_CLOSED,
+				       peer_dev->state)) {
+			set_bit(ICE_PEER_DEV_STATE_REMOVED, peer_dev->state);
+			dev_dbg(dev, "state from _OPENED/_CLOSED to _REMOVED\n");
+			/* Clear registration for events when peer removed */
+			bitmap_zero(peer_dev->events, ICE_PEER_DEV_STATE_NBITS);
+		}
+		break;
+	default:
+		break;
+	}
+
+	if (!locked)
+		mutex_unlock(&peer_dev->peer_dev_state_mutex);
+}
+
+/**
+ * ice_for_each_peer - iterate across and call function for each peer dev
+ * @pf: pointer to private board struct
+ * @data: data to pass to function on each call
+ * @fn: pointer to function to call for each peer
+ */
+int
+ice_for_each_peer(struct ice_pf *pf, void *data,
+		  int (*fn)(struct ice_peer_dev_int *, void *))
+{
+	unsigned int i;
+
+	if (!pf->peers)
+		return 0;
+
+	for (i = 0; i < ARRAY_SIZE(ice_peers); i++) {
+		struct ice_peer_dev_int *peer_dev_int;
+
+		peer_dev_int = pf->peers[i];
+		if (peer_dev_int) {
+			int ret = fn(peer_dev_int, data);
+
+			if (ret)
+				return ret;
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * ice_unreg_peer_device - unregister specified device
+ * @peer_dev_int: ptr to peer device internal
+ * @data: ptr to opaque data
+ *
+ * This function invokes device unregistration, removes ID associated with
+ * the specified device.
+ */
+int
+ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int,
+		      void __always_unused *data)
+{
+	struct ice_peer_drv_int *peer_drv_int;
+
+	if (!peer_dev_int)
+		return 0;
+
+	virtbus_unregister_device(peer_dev_int->peer_dev.vdev);
+
+	peer_drv_int = peer_dev_int->peer_drv_int;
+
+	if (peer_dev_int->ice_peer_wq) {
+		if (peer_dev_int->peer_prep_task.func)
+			cancel_work_sync(&peer_dev_int->peer_prep_task);
+		destroy_workqueue(peer_dev_int->ice_peer_wq);
+	}
+
+	kfree(peer_drv_int);
+
+	kfree(peer_dev_int);
+
+	return 0;
+}
+
+/**
+ * ice_unroll_peer - destroy peers and peer_wq in case of error
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data
+ *
+ * This function releases resources in the event of a failure in creating
+ * peer devices or their individual work_queues. Meant to be called from
+ * a ice_for_each_peer invocation
+ */
+int
+ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int,
+		void __always_unused *data)
+{
+	if (peer_dev_int->ice_peer_wq)
+		destroy_workqueue(peer_dev_int->ice_peer_wq);
+	kfree(peer_dev_int);
+
+	return 0;
+}
+
+/**
+ * ice_reserve_peer_qvector - Reserve vector resources for peer drivers
+ * @pf: board private structure to initialize
+ */
+static int ice_reserve_peer_qvector(struct ice_pf *pf)
+{
+	if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+		int index;
+
+		index = ice_get_res(pf, pf->irq_tracker, pf->num_rdma_msix,
+				    ICE_RES_RDMA_VEC_ID);
+		if (index < 0)
+			return index;
+		pf->num_avail_sw_msix -= pf->num_rdma_msix;
+		pf->rdma_base_vector = index;
+	}
+	return 0;
+}
+
+/**
+ * ice_peer_vdev_release - function to map to virtbus_devices release callback
+ * @vdev: pointer to virtbus_device to free
+ */
+static void ice_peer_vdev_release(struct virtbus_device *vdev)
+{
+	struct iidc_virtbus_object *vbo;
+
+	vbo = container_of(vdev, struct iidc_virtbus_object, vdev);
+	kfree(vbo);
+}
+
+/**
+ * ice_init_peer_devices - initializes peer devices
+ * @pf: ptr to ice_pf
+ *
+ * This function initializes peer devices on the virtual bus.
+ */
+int ice_init_peer_devices(struct ice_pf *pf)
+{
+	struct ice_vsi *vsi = pf->vsi[0];
+	struct pci_dev *pdev = pf->pdev;
+	struct device *dev = &pdev->dev;
+	int status = 0;
+	unsigned int i, n;
+
+	/* Reserve vector resources */
+	status = ice_reserve_peer_qvector(pf);
+	if (status < 0) {
+		dev_err(dev, "failed to reserve vectors for peer drivers\n");
+		return status;
+	}
+	for (i = 0; i < ARRAY_SIZE(ice_peers); i++) {
+		struct ice_peer_dev_int *peer_dev_int;
+		struct ice_peer_drv_int *peer_drv_int;
+		struct iidc_qos_params *qos_info;
+		struct iidc_virtbus_object *vbo;
+		struct msix_entry *entry = NULL;
+		struct iidc_peer_dev *peer_dev;
+		struct virtbus_device *vdev;
+		int j;
+
+		/* structure layout needed for container_of's looks like:
+		 * ice_peer_dev_int (internal only ice peer superstruct)
+		 * |--> iidc_peer_dev
+		 * |--> *ice_peer_drv_int
+		 *
+		 * iidc_virtbus_object (container_of parent for vdev)
+		 * |--> virtbus_device
+		 * |--> *iidc_peer_dev (pointer from internal struct)
+		 *
+		 * ice_peer_drv_int (internal only peer_drv struct)
+		 */
+		peer_dev_int = kzalloc(sizeof(*peer_dev_int), GFP_KERNEL);
+		if (!peer_dev_int)
+			goto unroll_prev_peers;
+
+		vbo = kzalloc(sizeof(*vbo), GFP_KERNEL);
+		if (!vbo) {
+			kfree(peer_dev_int);
+			goto unroll_prev_peers;
+		}
+
+		peer_drv_int = kzalloc(sizeof(*peer_drv_int), GFP_KERNEL);
+		if (!peer_drv_int) {
+			kfree(peer_dev_int);
+			kfree(vbo);
+			goto unroll_prev_peers;
+		}
+
+		pf->peers[i] = peer_dev_int;
+		vbo->peer_dev = &peer_dev_int->peer_dev;
+		peer_dev_int->peer_drv_int = peer_drv_int;
+		peer_dev_int->peer_dev.vdev = &vbo->vdev;
+
+		/* Initialize driver values */
+		for (j = 0; j < IIDC_EVENT_NBITS; j++)
+			bitmap_zero(peer_drv_int->current_events[j].type,
+				    IIDC_EVENT_NBITS);
+
+		mutex_init(&peer_dev_int->peer_dev_state_mutex);
+
+		peer_dev = &peer_dev_int->peer_dev;
+		peer_dev->peer_ops = NULL;
+		peer_dev->hw_addr = (u8 __iomem *)pf->hw.hw_addr;
+		peer_dev->peer_dev_id = ice_peers[i].id;
+		peer_dev->pf_vsi_num = vsi->vsi_num;
+		peer_dev->netdev = vsi->netdev;
+
+		peer_dev_int->ice_peer_wq =
+			alloc_ordered_workqueue("ice_peer_wq_%d", WQ_UNBOUND,
+						i);
+		if (!peer_dev_int->ice_peer_wq) {
+			kfree(peer_dev_int);
+			kfree(peer_drv_int);
+			kfree(vbo);
+			goto unroll_prev_peers;
+		}
+
+		peer_dev->pdev = pdev;
+		qos_info = &peer_dev->initial_qos_info;
+
+		/* setup qos_info fields with defaults */
+		qos_info->num_apps = 0;
+		qos_info->num_tc = 1;
+
+		for (j = 0; j < IIDC_MAX_USER_PRIORITY; j++)
+			qos_info->up2tc[j] = 0;
+
+		qos_info->tc_info[0].rel_bw = 100;
+		for (j = 1; j < IEEE_8021QAZ_MAX_TCS; j++)
+			qos_info->tc_info[j].rel_bw = 0;
+
+		/* for DCB, override the qos_info defaults. */
+		ice_setup_dcb_qos_info(pf, qos_info);
+
+		/* make sure peer specific resources such as msix_count and
+		 * msix_entries are initialized
+		 */
+		switch (ice_peers[i].id) {
+		case IIDC_PEER_RDMA_ID:
+			if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+				peer_dev->msix_count = pf->num_rdma_msix;
+				entry = &pf->msix_entries[pf->rdma_base_vector];
+			}
+			break;
+		default:
+			break;
+		}
+
+		peer_dev->msix_entries = entry;
+		ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_INIT,
+				      false);
+
+		vdev = &vbo->vdev;
+		vdev->match_name = ice_peers[i].name;
+		vdev->release = ice_peer_vdev_release;
+		vdev->dev.parent = &pdev->dev;
+
+		status = virtbus_register_device(vdev);
+		if (status) {
+			kfree(peer_dev_int);
+			kfree(peer_drv_int);
+			goto unroll_prev_peers;
+		}
+	}
+
+	return status;
+
+unroll_prev_peers:
+	for (n = 0; n < i; n++) {
+		struct ice_peer_dev_int *prev_peer_dev_int;
+		struct ice_peer_drv_int *prev_peer_drv_int;
+		struct virtbus_device *vdev;
+
+		prev_peer_dev_int = pf->peers[n];
+		prev_peer_drv_int = prev_peer_dev_int->peer_drv_int;
+		vdev = prev_peer_dev_int->peer_dev.vdev;
+
+		virtbus_unregister_device(vdev);
+
+		kfree(prev_peer_dev_int);
+		kfree(prev_peer_drv_int);
+	}
+	return -ENOMEM;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
new file mode 100644
index 000000000000..daac19c45490
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -0,0 +1,67 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2019, Intel Corporation. */
+
+#ifndef _ICE_IDC_INT_H_
+#define _ICE_IDC_INT_H_
+
+#include <linux/net/intel/iidc.h>
+#include "ice.h"
+
+enum ice_peer_dev_state {
+	ICE_PEER_DEV_STATE_INIT,
+	ICE_PEER_DEV_STATE_PROBED,
+	ICE_PEER_DEV_STATE_OPENING,
+	ICE_PEER_DEV_STATE_OPENED,
+	ICE_PEER_DEV_STATE_PREP_RST,
+	ICE_PEER_DEV_STATE_PREPPED,
+	ICE_PEER_DEV_STATE_CLOSING,
+	ICE_PEER_DEV_STATE_CLOSED,
+	ICE_PEER_DEV_STATE_REMOVED,
+	ICE_PEER_DEV_STATE_API_RDY,
+	ICE_PEER_DEV_STATE_NBITS,               /* must be last */
+};
+
+enum ice_peer_drv_state {
+	ICE_PEER_DRV_STATE_MBX_RDY,
+	ICE_PEER_DRV_STATE_NBITS,               /* must be last */
+};
+
+struct ice_peer_drv_int {
+	struct iidc_peer_drv *peer_drv;
+
+	/* States associated with peer driver */
+	DECLARE_BITMAP(state, ICE_PEER_DRV_STATE_NBITS);
+
+	/* if this peer_dev is the originator of an event, these are the
+	 * most recent events of each type
+	 */
+	struct iidc_event current_events[IIDC_EVENT_NBITS];
+};
+
+struct ice_peer_dev_int {
+	struct ice_peer_drv_int *peer_drv_int; /* driver private structure */
+	struct iidc_peer_dev peer_dev;
+
+	/* if this peer_dev is the originator of an event, these are the
+	 * most recent events of each type
+	 */
+	struct iidc_event current_events[IIDC_EVENT_NBITS];
+	/* Events a peer has registered to be notified about */
+	DECLARE_BITMAP(events, IIDC_EVENT_NBITS);
+
+	/* States associated with peer device */
+	DECLARE_BITMAP(state, ICE_PEER_DEV_STATE_NBITS);
+	struct mutex peer_dev_state_mutex; /* peer_dev state mutex */
+
+	/* per peer workqueue */
+	struct workqueue_struct *ice_peer_wq;
+
+	struct work_struct peer_prep_task;
+	struct work_struct peer_close_task;
+
+	enum iidc_close_reason rst_type;
+};
+
+int ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
+#endif /* !_ICE_IDC_INT_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 2f256bf45efc..205ac5900551 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -504,6 +504,17 @@ bool ice_is_safe_mode(struct ice_pf *pf)
 	return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
 }
 
+/**
+ * ice_is_peer_ena
+ * @pf: pointer to the PF struct
+ *
+ * returns true if peer devices/drivers are supported, false otherwise
+ */
+bool ice_is_peer_ena(struct ice_pf *pf)
+{
+	return test_bit(ICE_FLAG_PEER_ENA, pf->flags);
+}
+
 /**
  * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
  * @vsi: the VSI being cleaned up
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.h b/drivers/net/ethernet/intel/ice/ice_lib.h
index 04ca00799364..db07cc065b10 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_lib.h
@@ -104,6 +104,8 @@ ice_vsi_cfg_mac_fltr(struct ice_vsi *vsi, const u8 *macaddr, bool set);
 
 bool ice_is_safe_mode(struct ice_pf *pf);
 
+bool ice_is_peer_ena(struct ice_pf *pf);
+
 bool ice_is_dflt_vsi_in_use(struct ice_sw *sw);
 
 bool ice_is_vsi_dflt_vsi(struct ice_sw *sw, struct ice_vsi *vsi);
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 5b190c257124..033e463bcdf1 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -5,6 +5,7 @@
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
+#include <linux/virtual_bus.h>
 #include "ice.h"
 #include "ice_base.h"
 #include "ice_lib.h"
@@ -2690,6 +2691,12 @@ static void ice_set_pf_caps(struct ice_pf *pf)
 {
 	struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
 
+	clear_bit(ICE_FLAG_IWARP_ENA, pf->flags);
+	clear_bit(ICE_FLAG_PEER_ENA, pf->flags);
+	if (func_caps->common_cap.iwarp) {
+		set_bit(ICE_FLAG_IWARP_ENA, pf->flags);
+		set_bit(ICE_FLAG_PEER_ENA, pf->flags);
+	}
 	clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
 	if (func_caps->common_cap.dcb)
 		set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
@@ -2769,6 +2776,16 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 	v_budget += needed;
 	v_left -= needed;
 
+	/* reserve vectors for RDMA peer driver */
+	if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+		needed = ICE_RDMA_NUM_VECS;
+		if (v_left < needed)
+			goto no_hw_vecs_left_err;
+		pf->num_rdma_msix = needed;
+		v_budget += needed;
+		v_left -= needed;
+	}
+
 	pf->msix_entries = devm_kcalloc(dev, v_budget,
 					sizeof(*pf->msix_entries), GFP_KERNEL);
 
@@ -2793,16 +2810,19 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 	if (v_actual < v_budget) {
 		dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
 			 v_budget, v_actual);
-/* 2 vectors for LAN (traffic + OICR) */
+/* 2 vectors for LAN and RDMA (traffic + OICR) */
 #define ICE_MIN_LAN_VECS 2
+#define ICE_MIN_RDMA_VECS 2
+#define ICE_MIN_VECS (ICE_MIN_LAN_VECS + ICE_MIN_RDMA_VECS)
 
-		if (v_actual < ICE_MIN_LAN_VECS) {
+		if (v_actual < ICE_MIN_VECS) {
 			/* error if we can't get minimum vectors */
 			pci_disable_msix(pf->pdev);
 			err = -ERANGE;
 			goto msix_err;
 		} else {
 			pf->num_lan_msix = ICE_MIN_LAN_VECS;
+			pf->num_rdma_msix = ICE_MIN_RDMA_VECS;
 		}
 	}
 
@@ -2818,6 +2838,7 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 	err = -ERANGE;
 exit_err:
 	pf->num_lan_msix = 0;
+	pf->num_rdma_msix = 0;
 	return err;
 }
 
@@ -3362,6 +3383,26 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
 
 	/* initialize DDP driven features */
 
+	/* init peers only if supported */
+	if (ice_is_peer_ena(pf)) {
+		pf->peers = devm_kcalloc(dev, IIDC_MAX_NUM_PEERS,
+					 sizeof(*pf->peers), GFP_KERNEL);
+		if (!pf->peers) {
+			err = -ENOMEM;
+			goto err_init_peer_unroll;
+		}
+
+		err = ice_init_peer_devices(pf);
+		if (err) {
+			dev_err(dev, "Failed to initialize peer devices: 0x%x\n",
+				err);
+			err = -EIO;
+			goto err_init_peer_unroll;
+		}
+	} else {
+		dev_warn(dev, "RDMA is not supported on this device\n");
+	}
+
 	/* Note: DCB init failure is non-fatal to load */
 	if (ice_init_pf_dcb(pf, false)) {
 		clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
@@ -3375,6 +3416,14 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
 
 	return 0;
 
+err_init_peer_unroll:
+	if (ice_is_peer_ena(pf)) {
+		ice_for_each_peer(pf, NULL, ice_unroll_peer);
+		if (pf->peers) {
+			devm_kfree(dev, pf->peers);
+			pf->peers = NULL;
+		}
+	}
 err_alloc_sw_unroll:
 	ice_devlink_destroy_port(pf);
 	set_bit(__ICE_SERVICE_DIS, pf->state);
@@ -3423,6 +3472,10 @@ static void ice_remove(struct pci_dev *pdev)
 
 	ice_devlink_destroy_port(pf);
 	ice_vsi_release_all(pf);
+	if (ice_is_peer_ena(pf)) {
+		ice_for_each_peer(pf, NULL, ice_unreg_peer_device);
+		devm_kfree(&pdev->dev, pf->peers);
+	}
 	ice_free_irq_msix_misc(pf);
 	ice_for_each_vsi(pf, i) {
 		if (!pf->vsi[i])
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index 4ce5f92fca4a..42b2d700bc1f 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -189,6 +189,7 @@ struct ice_hw_common_caps {
 	u8 rss_table_entry_width;	/* RSS Entry width in bits */
 
 	u8 dcb;
+	u8 iwarp;
 };
 
 /* Function specific capabilities */
diff --git a/include/linux/net/intel/iidc.h b/include/linux/net/intel/iidc.h
new file mode 100644
index 000000000000..8056e6d8c4cc
--- /dev/null
+++ b/include/linux/net/intel/iidc.h
@@ -0,0 +1,337 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2019, Intel Corporation. */
+
+#ifndef _IIDC_H_
+#define _IIDC_H_
+
+#include <linux/dcbnl.h>
+#include <linux/device.h>
+#include <linux/if_ether.h>
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/virtual_bus.h>
+
+enum iidc_event_type {
+	IIDC_EVENT_LINK_CHANGE,
+	IIDC_EVENT_MTU_CHANGE,
+	IIDC_EVENT_TC_CHANGE,
+	IIDC_EVENT_API_CHANGE,
+	IIDC_EVENT_MBX_CHANGE,
+	IIDC_EVENT_NBITS		/* must be last */
+};
+
+enum iidc_res_type {
+	IIDC_INVAL_RES,
+	IIDC_VSI,
+	IIDC_VEB,
+	IIDC_EVENT_Q,
+	IIDC_EGRESS_CMPL_Q,
+	IIDC_CMPL_EVENT_Q,
+	IIDC_ASYNC_EVENT_Q,
+	IIDC_DOORBELL_Q,
+	IIDC_RDMA_QSETS_TXSCHED,
+};
+
+enum iidc_peer_reset_type {
+	IIDC_PEER_PFR,
+	IIDC_PEER_CORER,
+	IIDC_PEER_CORER_SW_CORE,
+	IIDC_PEER_CORER_SW_FULL,
+	IIDC_PEER_GLOBR,
+};
+
+/* reason notified to peer driver as part of event handling */
+enum iidc_close_reason {
+	IIDC_REASON_INVAL,
+	IIDC_REASON_HW_UNRESPONSIVE,
+	IIDC_REASON_INTERFACE_DOWN, /* Administrative down */
+	IIDC_REASON_PEER_DRV_UNREG, /* peer driver getting unregistered */
+	IIDC_REASON_PEER_DEV_UNINIT,
+	IIDC_REASON_GLOBR_REQ,
+	IIDC_REASON_CORER_REQ,
+	/* Reason #7 reserved */
+	IIDC_REASON_PFR_REQ = 8,
+	IIDC_REASON_HW_RESET_PENDING,
+	IIDC_REASON_RECOVERY_MODE,
+	IIDC_REASON_PARAM_CHANGE,
+};
+
+enum iidc_rdma_filter {
+	IIDC_RDMA_FILTER_INVAL,
+	IIDC_RDMA_FILTER_IWARP,
+	IIDC_RDMA_FILTER_ROCEV2,
+	IIDC_RDMA_FILTER_BOTH,
+};
+
+/* Struct to hold per DCB APP info */
+struct iidc_dcb_app_info {
+	u8  priority;
+	u8  selector;
+	u16 prot_id;
+};
+
+struct iidc_peer_dev;
+
+#define IIDC_MAX_USER_PRIORITY		8
+#define IIDC_MAX_APPS			8
+
+/* Struct to hold per RDMA Qset info */
+struct iidc_rdma_qset_params {
+	u32 teid;	/* qset TEID */
+	u16 qs_handle; /* RDMA driver provides this */
+	u16 vsi_id; /* VSI index */
+	u8 tc; /* TC branch the QSet should belong to */
+	u8 reserved[3];
+};
+
+struct iidc_res_base {
+	/* Union for future provision e.g. other res_type */
+	union {
+		struct iidc_rdma_qset_params qsets;
+	} res;
+};
+
+struct iidc_res {
+	/* Type of resource. Filled by peer driver */
+	enum iidc_res_type res_type;
+	/* Count requested by peer driver */
+	u16 cnt_req;
+
+	/* Number of resources allocated. Filled in by callee.
+	 * Based on this value, caller to fill up "resources"
+	 */
+	u16 res_allocated;
+
+	/* Unique handle to resources allocated. Zero if call fails.
+	 * Allocated by callee and for now used by caller for internal
+	 * tracking purpose.
+	 */
+	u32 res_handle;
+
+	/* Peer driver has to allocate sufficient memory, to accommodate
+	 * cnt_requested before calling this function.
+	 * Memory has to be zero initialized. It is input/output param.
+	 * As a result of alloc_res API, this structures will be populated.
+	 */
+	struct iidc_res_base res[1];
+};
+
+struct iidc_qos_info {
+	u64 tc_ctx;
+	u8 rel_bw;
+	u8 prio_type;
+	u8 egress_virt_up;
+	u8 ingress_virt_up;
+};
+
+/* Struct to hold QoS info */
+struct iidc_qos_params {
+	struct iidc_qos_info tc_info[IEEE_8021QAZ_MAX_TCS];
+	u8 up2tc[IIDC_MAX_USER_PRIORITY];
+	u8 vsi_relative_bw;
+	u8 vsi_priority_type;
+	u32 num_apps;
+	struct iidc_dcb_app_info apps[IIDC_MAX_APPS];
+	u8 num_tc;
+};
+
+union iidc_event_info {
+	/* IIDC_EVENT_LINK_CHANGE */
+	struct {
+		struct net_device *lwr_nd;
+		u16 vsi_num; /* HW index of VSI corresponding to lwr ndev */
+		u8 new_link_state;
+		u8 lport;
+	} link_info;
+	/* IIDC_EVENT_MTU_CHANGE */
+	u16 mtu;
+	/* IIDC_EVENT_TC_CHANGE */
+	struct iidc_qos_params port_qos;
+	/* IIDC_EVENT_API_CHANGE */
+	u8 api_rdy;
+	/* IIDC_EVENT_MBX_CHANGE */
+	u8 mbx_rdy;
+};
+
+/* iidc_event elements are to be passed back and forth between the device
+ * owner and the peer drivers. They are to be used to both register/unregister
+ * for event reporting and to report an event (events can be either device
+ * owner generated or peer generated).
+ *
+ * For (un)registering for events, the structure needs to be populated with:
+ *   reporter - pointer to the iidc_peer_dev struct of the peer (un)registering
+ *   type - bitmap with bits set for event types to (un)register for
+ *
+ * For reporting events, the structure needs to be populated with:
+ *   reporter - pointer to peer that generated the event (NULL for ice)
+ *   type - bitmap with single bit set for this event type
+ *   info - union containing data relevant to this event type
+ */
+struct iidc_event {
+	struct iidc_peer_dev *reporter;
+	DECLARE_BITMAP(type, IIDC_EVENT_NBITS);
+	union iidc_event_info info;
+};
+
+/* Following APIs are implemented by device owner and invoked by peer
+ * drivers
+ */
+struct iidc_ops {
+	/* APIs to allocate resources such as VEB, VSI, Doorbell queues,
+	 * completion queues, Tx/Rx queues, etc...
+	 */
+	int (*alloc_res)(struct iidc_peer_dev *peer_dev,
+			 struct iidc_res *res,
+			 int partial_acceptable);
+	int (*free_res)(struct iidc_peer_dev *peer_dev,
+			struct iidc_res *res);
+
+	int (*is_vsi_ready)(struct iidc_peer_dev *peer_dev);
+	int (*peer_register)(struct iidc_peer_dev *peer_dev);
+	int (*peer_unregister)(struct iidc_peer_dev *peer_dev);
+	int (*request_reset)(struct iidc_peer_dev *dev,
+			     enum iidc_peer_reset_type reset_type);
+
+	void (*notify_state_change)(struct iidc_peer_dev *dev,
+				    struct iidc_event *event);
+
+	/* Notification APIs */
+	void (*reg_for_notification)(struct iidc_peer_dev *dev,
+				     struct iidc_event *event);
+	void (*unreg_for_notification)(struct iidc_peer_dev *dev,
+				       struct iidc_event *event);
+	int (*update_vsi_filter)(struct iidc_peer_dev *peer_dev,
+				 enum iidc_rdma_filter filter, bool enable);
+	int (*vc_send)(struct iidc_peer_dev *peer_dev, u32 vf_id, u8 *msg,
+		       u16 len);
+};
+
+/* Following APIs are implemented by peer drivers and invoked by device
+ * owner
+ */
+struct iidc_peer_ops {
+	void (*event_handler)(struct iidc_peer_dev *peer_dev,
+			      struct iidc_event *event);
+
+	/* Why we have 'open' and when it is expected to be called:
+	 * 1. symmetric set of API w.r.t close
+	 * 2. To be invoked form driver initialization path
+	 *     - call peer_driver:open once device owner is fully
+	 *     initialized
+	 * 3. To be invoked upon RESET complete
+	 */
+	int (*open)(struct iidc_peer_dev *peer_dev);
+
+	/* Peer's close function is to be called when the peer needs to be
+	 * quiesced. This can be for a variety of reasons (enumerated in the
+	 * iidc_close_reason enum struct). A call to close will only be
+	 * followed by a call to either remove or open. No IDC calls from the
+	 * peer should be accepted until it is re-opened.
+	 *
+	 * The *reason* parameter is the reason for the call to close. This
+	 * can be for any reason enumerated in the iidc_close_reason struct.
+	 * It's primary reason is for the peer's bookkeeping and in case the
+	 * peer want to perform any different tasks dictated by the reason.
+	 */
+	void (*close)(struct iidc_peer_dev *peer_dev,
+		      enum iidc_close_reason reason);
+
+	int (*vc_receive)(struct iidc_peer_dev *peer_dev, u32 vf_id, u8 *msg,
+			  u16 len);
+	/* tell RDMA peer to prepare for TC change in a blocking call
+	 * that will directly precede the change event
+	 */
+	void (*prep_tc_change)(struct iidc_peer_dev *peer_dev);
+};
+
+#define IIDC_PEER_RDMA_NAME	"intel,ice,rdma"
+#define IIDC_PEER_RDMA_ID	0x00000010
+#define IIDC_MAX_NUM_PEERS	4
+
+/* The const struct that instantiates peer_dev_id needs to be initialized
+ * in the .c with the macro ASSIGN_PEER_INFO.
+ * For example:
+ * static const struct peer_dev_id peer_dev_ids[] = ASSIGN_PEER_INFO;
+ */
+struct peer_dev_id {
+	char *name;
+	int id;
+};
+
+#define ASSIGN_PEER_INFO						\
+{									\
+	{ .name = IIDC_PEER_RDMA_NAME, .id = IIDC_PEER_RDMA_ID },	\
+}
+
+#define iidc_peer_priv(x) ((x)->peer_priv)
+
+/* Structure representing peer specific information, each peer using the IIDC
+ * interface will have an instance of this struct dedicated to it.
+ */
+struct iidc_peer_dev {
+	struct pci_dev *pdev; /* PCI device of corresponding to main function */
+	struct virtbus_device *vdev; /* virtual device for this peer */
+	/* KVA / Linear address corresponding to BAR0 of underlying
+	 * pci_device.
+	 */
+	u8 __iomem *hw_addr;
+	int peer_dev_id;
+
+	/* Opaque pointer for peer specific data tracking.  This memory will
+	 * be alloc'd and freed by the peer driver and used for private data
+	 * accessible only to the specific peer.  It is stored here so that
+	 * when this struct is passed to the peer via an IDC call, the data
+	 * can be accessed by the peer at that time.
+	 * The peers should only retrieve the pointer by the macro:
+	 *    iidc_peer_priv(struct iidc_peer_dev *)
+	 */
+	void *peer_priv;
+
+	u8 ftype;	/* PF(false) or VF (true) */
+
+	/* Data VSI created by driver */
+	u16 pf_vsi_num;
+
+	struct iidc_qos_params initial_qos_info;
+	struct net_device *netdev;
+
+	/* Based on peer driver type, this shall point to corresponding MSIx
+	 * entries in pf->msix_entries (which were allocated as part of driver
+	 * initialization) e.g. for RDMA driver, msix_entries reserved will be
+	 * num_online_cpus + 1.
+	 */
+	u16 msix_count; /* How many vectors are reserved for this device */
+	struct msix_entry *msix_entries;
+
+	/* Following struct contains function pointers to be initialized
+	 * by device owner and called by peer driver
+	 */
+	const struct iidc_ops *ops;
+
+	/* Following struct contains function pointers to be initialized
+	 * by peer driver and called by device owner
+	 */
+	const struct iidc_peer_ops *peer_ops;
+
+	/* Pointer to peer_drv struct to be populated by peer driver */
+	struct iidc_peer_drv *peer_drv;
+};
+
+struct iidc_virtbus_object {
+	struct virtbus_device vdev;
+	struct iidc_peer_dev *peer_dev;
+};
+
+/* structure representing peer driver
+ * Peer driver to initialize those function ptrs and it will be invoked
+ * by device owner as part of driver_registration via bus infrastructure
+ */
+struct iidc_peer_drv {
+	u16 driver_id;
+#define IIDC_PEER_DEVICE_OWNER		0
+#define IIDC_PEER_RDMA_DRIVER		4
+
+	const char *name;
+
+};
+#endif /* _IIDC_H_*/
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 3/9] ice: Complete RDMA peer registration
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 2/9] ice: Create and register virtual bus for RDMA Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 4/9] ice: Support resource allocation requests Jeff Kirsher
                   ` (5 subsequent siblings)
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

Ensure that the peer supports the minimal set of operations required
for operation and, if so, open the connection to the peer.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_idc.c      | 288 ++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_idc_int.h  |  36 +++
 drivers/net/ethernet/intel/ice/ice_lib.c      |  33 ++
 drivers/net/ethernet/intel/ice/ice_lib.h      |   2 +
 drivers/net/ethernet/intel/ice/ice_main.c     |  18 +-
 drivers/net/ethernet/intel/ice/ice_switch.c   |  23 ++
 drivers/net/ethernet/intel/ice/ice_switch.h   |   2 +
 .../net/ethernet/intel/ice/ice_virtchnl_pf.c  |  25 --
 8 files changed, 401 insertions(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 68d6b524d6d4..499c1b77dfc9 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -146,6 +146,78 @@ ice_peer_state_change(struct ice_peer_dev_int *peer_dev, long new_state,
 		mutex_unlock(&peer_dev->peer_dev_state_mutex);
 }
 
+/**
+ * ice_peer_close - close a peer device
+ * @peer_dev_int: device to close
+ * @data: pointer to opaque data
+ *
+ * This function will also set the state bit for the peer to CLOSED. This
+ * function is meant to be called from a ice_for_each_peer().
+ */
+int ice_peer_close(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+	enum iidc_close_reason reason = *(enum iidc_close_reason *)(data);
+	struct iidc_peer_dev *peer_dev;
+	struct ice_pf *pf;
+	int i;
+
+	peer_dev = &peer_dev_int->peer_dev;
+	/* return 0 so ice_for_each_peer will continue closing other peers */
+	if (!ice_validate_peer_dev(peer_dev))
+		return 0;
+	pf = pci_get_drvdata(peer_dev->pdev);
+
+	if (test_bit(__ICE_DOWN, pf->state) ||
+	    test_bit(__ICE_SUSPENDED, pf->state) ||
+	    test_bit(__ICE_NEEDS_RESTART, pf->state))
+		return 0;
+
+	mutex_lock(&peer_dev_int->peer_dev_state_mutex);
+
+	/* no peer driver, already closed, closing or opening nothing to do */
+	if (test_bit(ICE_PEER_DEV_STATE_CLOSED, peer_dev_int->state) ||
+	    test_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev_int->state) ||
+	    test_bit(ICE_PEER_DEV_STATE_OPENING, peer_dev_int->state) ||
+	    test_bit(ICE_PEER_DEV_STATE_REMOVED, peer_dev_int->state))
+		goto peer_close_out;
+
+	/* Set the peer state to CLOSING */
+	ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSING, true);
+
+	for (i = 0; i < IIDC_EVENT_NBITS; i++)
+		bitmap_zero(peer_dev_int->current_events[i].type,
+			    IIDC_EVENT_NBITS);
+
+	if (peer_dev->peer_ops && peer_dev->peer_ops->close)
+		peer_dev->peer_ops->close(peer_dev, reason);
+
+	/* Set the peer state to CLOSED */
+	ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSED, true);
+
+peer_close_out:
+	mutex_unlock(&peer_dev_int->peer_dev_state_mutex);
+
+	return 0;
+}
+
+/**
+ * ice_peer_update_vsi - update the pf_vsi info in peer_dev struct
+ * @peer_dev_int: pointer to peer dev internal struct
+ * @data: opaque pointer - VSI to be updated
+ */
+int ice_peer_update_vsi(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+	struct ice_vsi *vsi = (struct ice_vsi *)data;
+	struct iidc_peer_dev *peer_dev;
+
+	peer_dev = &peer_dev_int->peer_dev;
+	if (!peer_dev)
+		return 0;
+
+	peer_dev->pf_vsi_num = vsi->vsi_num;
+	return 0;
+}
+
 /**
  * ice_for_each_peer - iterate across and call function for each peer dev
  * @pf: pointer to private board struct
@@ -176,6 +248,89 @@ ice_for_each_peer(struct ice_pf *pf, void *data,
 	return 0;
 }
 
+/**
+ * ice_finish_init_peer_device - complete peer device initialization
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data
+ *
+ * This function completes remaining initialization of peer_devices
+ */
+int
+ice_finish_init_peer_device(struct ice_peer_dev_int *peer_dev_int,
+			    void __always_unused *data)
+{
+	struct iidc_peer_dev *peer_dev;
+	struct iidc_peer_drv *peer_drv;
+	struct device *dev;
+	struct ice_pf *pf;
+	int ret = 0;
+
+	peer_dev = &peer_dev_int->peer_dev;
+	/* peer_dev will not always be populated at the time of this check */
+	if (!ice_validate_peer_dev(peer_dev))
+		return ret;
+
+	peer_drv = peer_dev->peer_drv;
+	pf = pci_get_drvdata(peer_dev->pdev);
+	dev = ice_pf_to_dev(pf);
+	/* There will be several assessments of the peer_dev's state in this
+	 * chunk of logic.  We need to hold the peer_dev_int's state mutex
+	 * for the entire part so that the flow progresses without another
+	 * context changing things mid-flow
+	 */
+	mutex_lock(&peer_dev_int->peer_dev_state_mutex);
+
+	if (!peer_dev->peer_ops) {
+		dev_err(dev, "peer_ops not defined on peer dev\n");
+		goto init_unlock;
+	}
+
+	if (!peer_dev->peer_ops->open) {
+		dev_err(dev, "peer_ops:open not defined on peer dev\n");
+		goto init_unlock;
+	}
+
+	if (!peer_dev->peer_ops->close) {
+		dev_err(dev, "peer_ops:close not defined on peer dev\n");
+		goto init_unlock;
+	}
+
+	/* Peer driver expected to set driver_id during registration */
+	if (!peer_drv->driver_id) {
+		dev_err(dev, "Peer driver did not set driver_id\n");
+		goto init_unlock;
+	}
+
+	if ((test_bit(ICE_PEER_DEV_STATE_CLOSED, peer_dev_int->state) ||
+	     test_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev_int->state)) &&
+	    ice_pf_state_is_nominal(pf)) {
+		/* If the RTNL is locked, we defer opening the peer
+		 * until the next time this function is called by the
+		 * service task.
+		 */
+		if (rtnl_is_locked())
+			goto init_unlock;
+		ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_OPENING,
+				      true);
+		ret = peer_dev->peer_ops->open(peer_dev);
+		if (ret) {
+			dev_err(dev, "Peer %d failed to open\n",
+				peer_dev->peer_dev_id);
+			ice_peer_state_change(peer_dev_int,
+					      ICE_PEER_DEV_STATE_PROBED, true);
+			goto init_unlock;
+		}
+
+		ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_OPENED,
+				      true);
+	}
+
+init_unlock:
+	mutex_unlock(&peer_dev_int->peer_dev_state_mutex);
+
+	return ret;
+}
+
 /**
  * ice_unreg_peer_device - unregister specified device
  * @peer_dev_int: ptr to peer device internal
@@ -200,6 +355,9 @@ ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int,
 	if (peer_dev_int->ice_peer_wq) {
 		if (peer_dev_int->peer_prep_task.func)
 			cancel_work_sync(&peer_dev_int->peer_prep_task);
+
+		if (peer_dev_int->peer_close_task.func)
+			cancel_work_sync(&peer_dev_int->peer_close_task);
 		destroy_workqueue(peer_dev_int->ice_peer_wq);
 	}
 
@@ -230,6 +388,134 @@ ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int,
 	return 0;
 }
 
+/**
+ * ice_peer_unregister - request to unregister peer
+ * @peer_dev: peer device
+ *
+ * This function triggers close/remove on peer_dev allowing peer
+ * to unregister.
+ */
+static int ice_peer_unregister(struct iidc_peer_dev *peer_dev)
+{
+	enum iidc_close_reason reason = IIDC_REASON_PEER_DEV_UNINIT;
+	struct ice_peer_dev_int *peer_dev_int;
+	struct ice_pf *pf;
+	int ret;
+
+	if (!ice_validate_peer_dev(peer_dev))
+		return -EINVAL;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	if (ice_is_reset_in_progress(pf->state))
+		return -EBUSY;
+
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+
+	ret = ice_peer_close(peer_dev_int, &reason);
+	if (ret)
+		return ret;
+
+	peer_dev->peer_ops = NULL;
+
+	ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_REMOVED, false);
+
+	return 0;
+}
+
+/**
+ * ice_peer_register - Called by peer to open communication with LAN
+ * @peer_dev: ptr to peer device
+ *
+ * registering peer is expected to populate the ice_peerdrv->name field
+ * before calling this function.
+ */
+static int ice_peer_register(struct iidc_peer_dev *peer_dev)
+{
+	struct ice_peer_drv_int *peer_drv_int;
+	struct ice_peer_dev_int *peer_dev_int;
+	struct iidc_peer_drv *peer_drv;
+
+	if (!peer_dev) {
+		pr_err("Failed to reg peer dev: peer_dev ptr NULL\n");
+		return -EINVAL;
+	}
+
+	if (!peer_dev->pdev) {
+		pr_err("Failed to reg peer dev: peer dev pdev NULL\n");
+		return -EINVAL;
+	}
+
+	if (!peer_dev->peer_ops || !peer_dev->ops) {
+		pr_err("Failed to reg peer dev: peer dev peer_ops/ops NULL\n");
+		return -EINVAL;
+	}
+
+	peer_drv = peer_dev->peer_drv;
+	if (!peer_drv) {
+		pr_err("Failed to reg peer dev: peer drv NULL\n");
+		return -EINVAL;
+	}
+
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+	peer_drv_int = peer_dev_int->peer_drv_int;
+	if (!peer_drv_int) {
+		pr_err("Failed to match peer_drv_int to peer_dev\n");
+		return -EINVAL;
+	}
+
+	peer_drv_int->peer_drv = peer_drv;
+
+	ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_PROBED, false);
+
+	return 0;
+}
+
+/**
+ * ice_peer_update_vsi_filter - update main VSI filters for RDMA
+ * @peer_dev: pointer to RDMA peer device
+ * @filter: selection of filters to enable or disable
+ * @enable: bool whether to enable or disable filters
+ */
+static int
+ice_peer_update_vsi_filter(struct iidc_peer_dev *peer_dev,
+			   enum iidc_rdma_filter __always_unused filter,
+			   bool enable)
+{
+	struct ice_vsi *vsi;
+	struct ice_pf *pf;
+	int ret;
+
+	if (!ice_validate_peer_dev(peer_dev))
+		return -EINVAL;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+
+	vsi = ice_get_main_vsi(pf);
+	if (!vsi)
+		return -EINVAL;
+
+	ret = ice_cfg_iwarp_fltr(&pf->hw, vsi->idx, enable);
+
+	if (ret) {
+		dev_err(ice_pf_to_dev(pf), "Failed to  %sable iWARP filtering\n",
+			enable ? "en" : "dis");
+	} else {
+		if (enable)
+			vsi->info.q_opt_flags |= ICE_AQ_VSI_Q_OPT_PE_FLTR_EN;
+		else
+			vsi->info.q_opt_flags &= ~ICE_AQ_VSI_Q_OPT_PE_FLTR_EN;
+	}
+
+	return ret;
+}
+
+/* Initialize the ice_ops struct, which is used in 'ice_init_peer_devices' */
+static const struct iidc_ops ops = {
+	.peer_register			= ice_peer_register,
+	.peer_unregister		= ice_peer_unregister,
+	.update_vsi_filter		= ice_peer_update_vsi_filter,
+};
+
 /**
  * ice_reserve_peer_qvector - Reserve vector resources for peer drivers
  * @pf: board private structure to initialize
@@ -364,6 +650,8 @@ int ice_init_peer_devices(struct ice_pf *pf)
 
 		/* for DCB, override the qos_info defaults. */
 		ice_setup_dcb_qos_info(pf, qos_info);
+		/* Initialize ice_ops */
+		peer_dev->ops = &ops;
 
 		/* make sure peer specific resources such as msix_count and
 		 * msix_entries are initialized
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
index daac19c45490..d22e6f5bb50e 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc_int.h
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -62,6 +62,42 @@ struct ice_peer_dev_int {
 	enum iidc_close_reason rst_type;
 };
 
+int ice_peer_update_vsi(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_peer_close(struct ice_peer_dev_int *peer_dev_int, void *data);
+int
+ice_finish_init_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
+
+static inline struct
+ice_peer_dev_int *peer_to_ice_dev_int(struct iidc_peer_dev *peer_dev)
+{
+	return  container_of(peer_dev, struct ice_peer_dev_int, peer_dev);
+}
+
+static inline bool ice_validate_peer_dev(struct iidc_peer_dev *peer_dev)
+{
+	struct ice_peer_dev_int *peer_dev_int;
+	struct ice_pf *pf;
+
+	if (!peer_dev || !peer_dev->pdev)
+		return false;
+
+	if (!peer_dev->peer_ops)
+		return false;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	if (!pf)
+		return false;
+
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+	if (!peer_dev_int)
+		return false;
+
+	if (test_bit(ICE_PEER_DEV_STATE_REMOVED, peer_dev_int->state) ||
+	    test_bit(ICE_PEER_DEV_STATE_INIT, peer_dev_int->state))
+		return false;
+
+	return true;
+}
 #endif /* !_ICE_IDC_INT_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 205ac5900551..5043d5ed1b2a 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1379,6 +1379,30 @@ ice_add_mac_to_list(struct ice_vsi *vsi, struct list_head *add_list,
 	return 0;
 }
 
+/**
+ * ice_pf_state_is_nominal - checks the PF for nominal state
+ * @pf: pointer to PF to check
+ *
+ * Check the PF's state for a collection of bits that would indicate
+ * the PF is in a state that would inhibit normal operation for
+ * driver functionality.
+ *
+ * Returns true if PF is in a nominal state, false otherwise
+ */
+bool ice_pf_state_is_nominal(struct ice_pf *pf)
+{
+	DECLARE_BITMAP(check_bits, __ICE_STATE_NBITS) = { 0 };
+
+	if (!pf)
+		return false;
+
+	bitmap_set(check_bits, 0, __ICE_STATE_NOMINAL_CHECK_BITS);
+	if (bitmap_intersects(pf->state, check_bits, __ICE_STATE_NBITS))
+		return false;
+
+	return true;
+}
+
 /**
  * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
  * @vsi: the VSI to be updated
@@ -2390,6 +2414,15 @@ void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
  */
 void ice_vsi_close(struct ice_vsi *vsi)
 {
+	enum iidc_close_reason reason = IIDC_REASON_INTERFACE_DOWN;
+
+	if (!ice_is_safe_mode(vsi->back) && vsi->type == ICE_VSI_PF) {
+		int ret = ice_for_each_peer(vsi->back, &reason, ice_peer_close);
+
+		if (ret)
+			dev_dbg(ice_pf_to_dev(vsi->back), "Peer device did not implement close function\n");
+	}
+
 	if (!test_and_set_bit(__ICE_DOWN, vsi->state))
 		ice_down(vsi);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.h b/drivers/net/ethernet/intel/ice/ice_lib.h
index db07cc065b10..f77ddd6883c3 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_lib.h
@@ -14,6 +14,8 @@ ice_add_mac_to_list(struct ice_vsi *vsi, struct list_head *add_list,
 
 void ice_free_fltr_list(struct device *dev, struct list_head *h);
 
+bool ice_pf_state_is_nominal(struct ice_pf *pf);
+
 void ice_update_eth_stats(struct ice_vsi *vsi);
 
 int ice_vsi_cfg_rxqs(struct ice_vsi *vsi);
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 033e463bcdf1..ac0c6d5b01e4 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -1492,6 +1492,9 @@ static void ice_service_task(struct work_struct *work)
 		return;
 	}
 
+	/* Invoke remaining initialization of peer devices */
+	ice_for_each_peer(pf, NULL, ice_finish_init_peer_device);
+
 	ice_process_vflr_event(pf);
 	ice_clean_mailboxq_subtask(pf);
 
@@ -3451,6 +3454,7 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
 static void ice_remove(struct pci_dev *pdev)
 {
 	struct ice_pf *pf = pci_get_drvdata(pdev);
+	enum iidc_close_reason reason;
 	int i;
 
 	if (!pf)
@@ -3467,8 +3471,12 @@ static void ice_remove(struct pci_dev *pdev)
 		ice_free_vfs(pf);
 	}
 
-	set_bit(__ICE_DOWN, pf->state);
 	ice_service_task_stop(pf);
+	if (ice_is_peer_ena(pf)) {
+		reason = IIDC_REASON_INTERFACE_DOWN;
+		ice_for_each_peer(pf, &reason, ice_peer_close);
+	}
+	set_bit(__ICE_DOWN, pf->state);
 
 	ice_devlink_destroy_port(pf);
 	ice_vsi_release_all(pf);
@@ -4785,7 +4793,15 @@ static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
 		dev_err(dev, "PF VSI rebuild failed: %d\n", err);
 		goto err_vsi_rebuild;
 	}
+	if (ice_is_peer_ena(pf)) {
+		struct ice_vsi *vsi = ice_get_main_vsi(pf);
 
+		if (!vsi) {
+			dev_err(dev, "No PF_VSI to update peer\n");
+			goto err_vsi_rebuild;
+		}
+		ice_for_each_peer(pf, vsi, ice_peer_update_vsi);
+	}
 	if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
 		err = ice_vsi_rebuild_by_type(pf, ICE_VSI_VF);
 		if (err) {
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.c b/drivers/net/ethernet/intel/ice/ice_switch.c
index 51825a203e35..cf8e1553599a 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.c
+++ b/drivers/net/ethernet/intel/ice/ice_switch.c
@@ -430,6 +430,29 @@ ice_update_vsi(struct ice_hw *hw, u16 vsi_handle, struct ice_vsi_ctx *vsi_ctx,
 	return ice_aq_update_vsi(hw, vsi_ctx, cd);
 }
 
+/**
+ * ice_cfg_iwarp_fltr - enable/disable iWARP filtering on VSI
+ * @hw: pointer to HW struct
+ * @vsi_handle: VSI SW index
+ * @enable: boolean for enable/disable
+ */
+enum ice_status
+ice_cfg_iwarp_fltr(struct ice_hw *hw, u16 vsi_handle, bool enable)
+{
+	struct ice_vsi_ctx *ctx;
+
+	ctx = ice_get_vsi_ctx(hw, vsi_handle);
+	if (!ctx)
+		return ICE_ERR_DOES_NOT_EXIST;
+
+	if (enable)
+		ctx->info.q_opt_flags |= ICE_AQ_VSI_Q_OPT_PE_FLTR_EN;
+	else
+		ctx->info.q_opt_flags &= ~ICE_AQ_VSI_Q_OPT_PE_FLTR_EN;
+
+	return ice_update_vsi(hw, vsi_handle, ctx, NULL);
+}
+
 /**
  * ice_aq_alloc_free_vsi_list
  * @hw: pointer to the HW struct
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.h b/drivers/net/ethernet/intel/ice/ice_switch.h
index fa14b9545dab..96010d3d96fd 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.h
+++ b/drivers/net/ethernet/intel/ice/ice_switch.h
@@ -220,6 +220,8 @@ void ice_remove_vsi_fltr(struct ice_hw *hw, u16 vsi_handle);
 enum ice_status
 ice_add_vlan(struct ice_hw *hw, struct list_head *m_list);
 enum ice_status ice_remove_vlan(struct ice_hw *hw, struct list_head *v_list);
+enum ice_status
+ice_cfg_iwarp_fltr(struct ice_hw *hw, u16 vsi_handle, bool enable);
 
 /* Promisc/defport setup for VSIs */
 enum ice_status
diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
index 15191a325918..07f3d4b456c7 100644
--- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
@@ -1375,31 +1375,6 @@ static int ice_alloc_vfs(struct ice_pf *pf, u16 num_alloc_vfs)
 	return ret;
 }
 
-/**
- * ice_pf_state_is_nominal - checks the PF for nominal state
- * @pf: pointer to PF to check
- *
- * Check the PF's state for a collection of bits that would indicate
- * the PF is in a state that would inhibit normal operation for
- * driver functionality.
- *
- * Returns true if PF is in a nominal state.
- * Returns false otherwise
- */
-static bool ice_pf_state_is_nominal(struct ice_pf *pf)
-{
-	DECLARE_BITMAP(check_bits, __ICE_STATE_NBITS) = { 0 };
-
-	if (!pf)
-		return false;
-
-	bitmap_set(check_bits, 0, __ICE_STATE_NOMINAL_CHECK_BITS);
-	if (bitmap_intersects(pf->state, check_bits, __ICE_STATE_NBITS))
-		return false;
-
-	return true;
-}
-
 /**
  * ice_pci_sriov_ena - Enable or change number of VFs
  * @pf: pointer to the PF structure
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 4/9] ice: Support resource allocation requests
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 2/9] ice: Create and register virtual bus for RDMA Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 3/9] ice: Complete RDMA peer registration Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 5/9] ice: Enable event notifications Jeff Kirsher
                   ` (4 subsequent siblings)
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

Enable the peer device to request queue sets from the PF.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h          |   1 +
 .../net/ethernet/intel/ice/ice_adminq_cmd.h   |  32 +++
 drivers/net/ethernet/intel/ice/ice_common.c   | 188 ++++++++++++++
 drivers/net/ethernet/intel/ice/ice_common.h   |   9 +
 drivers/net/ethernet/intel/ice/ice_idc.c      | 244 ++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_sched.c    |  69 ++++-
 drivers/net/ethernet/intel/ice/ice_switch.c   |   4 +
 drivers/net/ethernet/intel/ice/ice_switch.h   |   2 +
 drivers/net/ethernet/intel/ice/ice_type.h     |   3 +
 9 files changed, 547 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 73366009ef03..6ad1894eca3f 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -296,6 +296,7 @@ struct ice_vsi {
 	u16 req_rxq;			 /* User requested Rx queues */
 	u16 num_rx_desc;
 	u16 num_tx_desc;
+	u16 qset_handle[ICE_MAX_TRAFFIC_CLASS];
 	struct ice_tc_cfg tc_cfg;
 	struct bpf_prog *xdp_prog;
 	struct ice_ring **xdp_rings;	 /* XDP ring array */
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index 51baab0621a2..a1066c4bf40d 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -1536,6 +1536,36 @@ struct ice_aqc_dis_txq {
 	struct ice_aqc_dis_txq_item qgrps[1];
 };
 
+/* Add Tx RDMA Queue Set (indirect 0x0C33) */
+struct ice_aqc_add_rdma_qset {
+	u8 num_qset_grps;
+	u8 reserved[7];
+	__le32 addr_high;
+	__le32 addr_low;
+};
+
+/* This is the descriptor of each qset entry for the Add Tx RDMA Queue Set
+ * command (0x0C33). Only used within struct ice_aqc_add_rdma_qset.
+ */
+struct ice_aqc_add_tx_rdma_qset_entry {
+	__le16 tx_qset_id;
+	u8 rsvd[2];
+	__le32 qset_teid;
+	struct ice_aqc_txsched_elem info;
+};
+
+/* The format of the command buffer for Add Tx RDMA Queue Set(0x0C33)
+ * is an array of the following structs. Please note that the length of
+ * each struct ice_aqc_add_rdma_qset is variable due to the variable
+ * number of queues in each group!
+ */
+struct ice_aqc_add_rdma_qset_data {
+	__le32 parent_teid;
+	__le16 num_qsets;
+	u8 rsvd[2];
+	struct ice_aqc_add_tx_rdma_qset_entry rdma_qsets[1];
+};
+
 /* Configure Firmware Logging Command (indirect 0xFF09)
  * Logging Information Read Response (indirect 0xFF10)
  * Note: The 0xFF10 command has no input parameters.
@@ -1732,6 +1762,7 @@ struct ice_aq_desc {
 		struct ice_aqc_get_set_rss_key get_set_rss_key;
 		struct ice_aqc_add_txqs add_txqs;
 		struct ice_aqc_dis_txqs dis_txqs;
+		struct ice_aqc_add_rdma_qset add_rdma_qset;
 		struct ice_aqc_add_get_update_free_vsi vsi_cmd;
 		struct ice_aqc_add_update_free_vsi_resp add_update_free_vsi_res;
 		struct ice_aqc_fw_logging fw_logging;
@@ -1867,6 +1898,7 @@ enum ice_adminq_opc {
 	/* Tx queue handling commands/events */
 	ice_aqc_opc_add_txqs				= 0x0C30,
 	ice_aqc_opc_dis_txqs				= 0x0C31,
+	ice_aqc_opc_add_rdma_qset			= 0x0C33,
 
 	/* package commands */
 	ice_aqc_opc_download_pkg			= 0x0C40,
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index 2dca49aed5bb..c760fae4aed4 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -2917,6 +2917,59 @@ ice_aq_dis_lan_txq(struct ice_hw *hw, u8 num_qgrps,
 	return status;
 }
 
+/**
+ * ice_aq_add_rdma_qsets
+ * @hw: pointer to the hardware structure
+ * @num_qset_grps: Number of RDMA Qset groups
+ * @qset_list: list of qset groups to be added
+ * @buf_size: size of buffer for indirect command
+ * @cd: pointer to command details structure or NULL
+ *
+ * Add Tx RDMA Qsets (0x0C33)
+ */
+static enum ice_status
+ice_aq_add_rdma_qsets(struct ice_hw *hw, u8 num_qset_grps,
+		      struct ice_aqc_add_rdma_qset_data *qset_list,
+		      u16 buf_size, struct ice_sq_cd *cd)
+{
+	struct ice_aqc_add_rdma_qset_data *list;
+	u16 i, sum_header_size, sum_q_size = 0;
+	struct ice_aqc_add_rdma_qset *cmd;
+	struct ice_aq_desc desc;
+
+	cmd = &desc.params.add_rdma_qset;
+
+	ice_fill_dflt_direct_cmd_desc(&desc, ice_aqc_opc_add_rdma_qset);
+
+	if (!qset_list)
+		return ICE_ERR_PARAM;
+
+	if (num_qset_grps > ICE_LAN_TXQ_MAX_QGRPS)
+		return ICE_ERR_PARAM;
+
+	sum_header_size = num_qset_grps *
+		(sizeof(*qset_list) - sizeof(*qset_list->rdma_qsets));
+
+	list = qset_list;
+	for (i = 0; i < num_qset_grps; i++) {
+		struct ice_aqc_add_tx_rdma_qset_entry *qset = list->rdma_qsets;
+		u16 num_qsets = le16_to_cpu(list->num_qsets);
+
+		sum_q_size += num_qsets * sizeof(*qset);
+		list = (struct ice_aqc_add_rdma_qset_data *)
+			(qset + num_qsets);
+	}
+
+	if (buf_size != (sum_header_size + sum_q_size))
+		return ICE_ERR_PARAM;
+
+	desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
+
+	cmd->num_qset_grps = num_qset_grps;
+
+	return ice_aq_send_cmd(hw, &desc, qset_list, buf_size, cd);
+}
+
 /* End of FW Admin Queue command wrappers */
 
 /**
@@ -3388,6 +3441,141 @@ ice_cfg_vsi_lan(struct ice_port_info *pi, u16 vsi_handle, u8 tc_bitmap,
 			      ICE_SCHED_NODE_OWNER_LAN);
 }
 
+/**
+ * ice_cfg_vsi_rdma - configure the VSI RDMA queues
+ * @pi: port information structure
+ * @vsi_handle: software VSI handle
+ * @tc_bitmap: TC bitmap
+ * @max_rdmaqs: max RDMA queues array per TC
+ *
+ * This function adds/updates the VSI RDMA queues per TC.
+ */
+enum ice_status
+ice_cfg_vsi_rdma(struct ice_port_info *pi, u16 vsi_handle, u8 tc_bitmap,
+		 u16 *max_rdmaqs)
+{
+	return ice_cfg_vsi_qs(pi, vsi_handle, tc_bitmap, max_rdmaqs,
+			      ICE_SCHED_NODE_OWNER_RDMA);
+}
+
+/**
+ * ice_ena_vsi_rdma_qset
+ * @pi: port information structure
+ * @vsi_handle: software VSI handle
+ * @tc: TC number
+ * @rdma_qset: pointer to RDMA qset
+ * @num_qsets: number of RDMA qsets
+ * @qset_teid: pointer to qset node teids
+ *
+ * This function adds RDMA qset
+ */
+enum ice_status
+ice_ena_vsi_rdma_qset(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
+		      u16 *rdma_qset, u16 num_qsets, u32 *qset_teid)
+{
+	struct ice_aqc_txsched_elem_data node = { 0 };
+	struct ice_aqc_add_rdma_qset_data *buf;
+	struct ice_sched_node *parent;
+	enum ice_status status;
+	struct ice_hw *hw;
+	u16 i, buf_size;
+
+	if (!pi || pi->port_state != ICE_SCHED_PORT_STATE_READY)
+		return ICE_ERR_CFG;
+	hw = pi->hw;
+
+	if (!ice_is_vsi_valid(hw, vsi_handle))
+		return ICE_ERR_PARAM;
+
+	buf_size = struct_size(buf, rdma_qsets, num_qsets - 1);
+	buf = kzalloc(buf_size, GFP_KERNEL);
+	if (!buf)
+		return ICE_ERR_NO_MEMORY;
+	mutex_lock(&pi->sched_lock);
+
+	parent = ice_sched_get_free_qparent(pi, vsi_handle, tc,
+					    ICE_SCHED_NODE_OWNER_RDMA);
+	if (!parent) {
+		status = ICE_ERR_PARAM;
+		goto rdma_error_exit;
+	}
+	buf->parent_teid = parent->info.node_teid;
+	node.parent_teid = parent->info.node_teid;
+
+	buf->num_qsets = cpu_to_le16(num_qsets);
+	for (i = 0; i < num_qsets; i++) {
+		buf->rdma_qsets[i].tx_qset_id = cpu_to_le16(rdma_qset[i]);
+		buf->rdma_qsets[i].info.valid_sections =
+						ICE_AQC_ELEM_VALID_GENERIC;
+	}
+	status = ice_aq_add_rdma_qsets(hw, 1, buf, buf_size, NULL);
+	if (status) {
+		ice_debug(hw, ICE_DBG_RDMA, "add RDMA qset failed\n");
+		goto rdma_error_exit;
+	}
+	node.data.elem_type = ICE_AQC_ELEM_TYPE_LEAF;
+	for (i = 0; i < num_qsets; i++) {
+		node.node_teid = buf->rdma_qsets[i].qset_teid;
+		status = ice_sched_add_node(pi, hw->num_tx_sched_layers - 1,
+					    &node);
+		if (status)
+			break;
+		qset_teid[i] = le32_to_cpu(node.node_teid);
+	}
+rdma_error_exit:
+	mutex_unlock(&pi->sched_lock);
+	kfree(buf);
+	return status;
+}
+
+/**
+ * ice_dis_vsi_rdma_qset - free RDMA resources
+ * @pi: port_info struct
+ * @count: number of RDMA qsets to free
+ * @qset_teid: TEID of qset node
+ * @q_id: list of queue IDs being disabled
+ */
+enum ice_status
+ice_dis_vsi_rdma_qset(struct ice_port_info *pi, u16 count, u32 *qset_teid,
+		      u16 *q_id)
+{
+	struct ice_aqc_dis_txq_item qg_list;
+	enum ice_status status = 0;
+	u16 qg_size;
+	int i;
+
+	if (!pi || pi->port_state != ICE_SCHED_PORT_STATE_READY)
+		return ICE_ERR_CFG;
+
+	qg_size = sizeof(qg_list);
+
+	mutex_lock(&pi->sched_lock);
+
+	for (i = 0; i < count; i++) {
+		struct ice_sched_node *node;
+
+		node = ice_sched_find_node_by_teid(pi->root, qset_teid[i]);
+		if (!node)
+			continue;
+
+		qg_list.parent_teid = node->info.parent_teid;
+		qg_list.num_qs = 1;
+		qg_list.q_id[0] =
+			cpu_to_le16(q_id[i] |
+				    ICE_AQC_Q_DIS_BUF_ELEM_TYPE_RDMA_QSET);
+
+		status = ice_aq_dis_lan_txq(pi->hw, 1, &qg_list, qg_size,
+					    ICE_NO_RESET, 0, NULL);
+		if (status)
+			break;
+
+		ice_free_sched_node(pi, node);
+	}
+
+	mutex_unlock(&pi->sched_lock);
+	return status;
+}
+
 /**
  * ice_replay_pre_init - replay pre initialization
  * @hw: pointer to the HW struct
diff --git a/drivers/net/ethernet/intel/ice/ice_common.h b/drivers/net/ethernet/intel/ice/ice_common.h
index 8104f3d64d96..db63fd6b5608 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.h
+++ b/drivers/net/ethernet/intel/ice/ice_common.h
@@ -125,6 +125,15 @@ ice_aq_sff_eeprom(struct ice_hw *hw, u16 lport, u8 bus_addr,
 		  bool write, struct ice_sq_cd *cd);
 
 enum ice_status
+ice_cfg_vsi_rdma(struct ice_port_info *pi, u16 vsi_handle, u8 tc_bitmap,
+		 u16 *max_rdmaqs);
+enum ice_status
+ice_ena_vsi_rdma_qset(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
+		      u16 *rdma_qset, u16 num_qsets, u32 *qset_teid);
+enum ice_status
+ice_dis_vsi_rdma_qset(struct ice_port_info *pi, u16 count, u32 *qset_teid,
+		      u16 *q_id);
+enum ice_status
 ice_dis_vsi_txq(struct ice_port_info *pi, u16 vsi_handle, u8 tc, u8 num_queues,
 		u16 *q_handle, u16 *q_ids, u32 *q_teids,
 		enum ice_disq_rst_src rst_src, u16 vmvf_num,
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 499c1b77dfc9..05fa5c61e2d3 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -388,6 +388,248 @@ ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int,
 	return 0;
 }
 
+/**
+ * ice_find_vsi - Find the VSI from VSI ID
+ * @pf: The PF pointer to search in
+ * @vsi_num: The VSI ID to search for
+ */
+static struct ice_vsi *ice_find_vsi(struct ice_pf *pf, u16 vsi_num)
+{
+	int i;
+
+	ice_for_each_vsi(pf, i)
+		if (pf->vsi[i] && pf->vsi[i]->vsi_num == vsi_num)
+			return  pf->vsi[i];
+	return NULL;
+}
+
+/**
+ * ice_peer_alloc_rdma_qsets - Allocate Leaf Nodes for RDMA Qset
+ * @peer_dev: peer that is requesting the Leaf Nodes
+ * @res: Resources to be allocated
+ * @partial_acceptable: If partial allocation is acceptable to the peer
+ *
+ * This function allocates Leaf Nodes for given RDMA Qset resources
+ * for the peer device.
+ */
+static int
+ice_peer_alloc_rdma_qsets(struct iidc_peer_dev *peer_dev, struct iidc_res *res,
+			  int __always_unused partial_acceptable)
+{
+	u16 max_rdmaqs[ICE_MAX_TRAFFIC_CLASS];
+	enum ice_status status;
+	struct ice_vsi *vsi;
+	struct device *dev;
+	struct ice_pf *pf;
+	int i, ret = 0;
+	u32 *qset_teid;
+	u16 *qs_handle;
+
+	if (!ice_validate_peer_dev(peer_dev) || !res)
+		return -EINVAL;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	dev = ice_pf_to_dev(pf);
+
+	if (res->cnt_req > ICE_MAX_TXQ_PER_TXQG)
+		return -EINVAL;
+
+	qset_teid = kcalloc(res->cnt_req, sizeof(*qset_teid), GFP_KERNEL);
+	if (!qset_teid)
+		return -ENOMEM;
+
+	qs_handle = kcalloc(res->cnt_req, sizeof(*qs_handle), GFP_KERNEL);
+	if (!qs_handle) {
+		kfree(qset_teid);
+		return -ENOMEM;
+	}
+
+	ice_for_each_traffic_class(i)
+		max_rdmaqs[i] = 0;
+
+	for (i = 0; i < res->cnt_req; i++) {
+		struct iidc_rdma_qset_params *qset;
+
+		qset = &res->res[i].res.qsets;
+		if (qset->vsi_id != peer_dev->pf_vsi_num) {
+			dev_err(dev, "RDMA QSet invalid VSI requested\n");
+			ret = -EINVAL;
+			goto out;
+		}
+		max_rdmaqs[qset->tc]++;
+		qs_handle[i] = qset->qs_handle;
+	}
+
+	vsi = ice_find_vsi(pf, peer_dev->pf_vsi_num);
+	if (!vsi) {
+		dev_err(dev, "RDMA QSet invalid VSI\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	status = ice_cfg_vsi_rdma(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
+				  max_rdmaqs);
+	if (status) {
+		dev_err(dev, "Failed VSI RDMA qset config\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	for (i = 0; i < res->cnt_req; i++) {
+		struct iidc_rdma_qset_params *qset;
+
+		qset = &res->res[i].res.qsets;
+		status = ice_ena_vsi_rdma_qset(vsi->port_info, vsi->idx,
+					       qset->tc, &qs_handle[i], 1,
+					       &qset_teid[i]);
+		if (status) {
+			dev_err(dev, "Failed VSI RDMA qset enable\n");
+			ret = -EINVAL;
+			goto out;
+		}
+		vsi->qset_handle[qset->tc] = qset->qs_handle;
+		qset->teid = qset_teid[i];
+	}
+
+out:
+	kfree(qset_teid);
+	kfree(qs_handle);
+	return ret;
+}
+
+/**
+ * ice_peer_free_rdma_qsets - Free leaf nodes for RDMA Qset
+ * @peer_dev: peer that requested qsets to be freed
+ * @res: Resource to be freed
+ */
+static int
+ice_peer_free_rdma_qsets(struct iidc_peer_dev *peer_dev, struct iidc_res *res)
+{
+	enum ice_status status;
+	int count, i, ret = 0;
+	struct ice_vsi *vsi;
+	struct device *dev;
+	struct ice_pf *pf;
+	u16 vsi_id;
+	u32 *teid;
+	u16 *q_id;
+
+	if (!ice_validate_peer_dev(peer_dev) || !res)
+		return -EINVAL;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	dev = ice_pf_to_dev(pf);
+
+	count = res->res_allocated;
+	if (count > ICE_MAX_TXQ_PER_TXQG)
+		return -EINVAL;
+
+	teid = kcalloc(count, sizeof(*teid), GFP_KERNEL);
+	if (!teid)
+		return -ENOMEM;
+
+	q_id = kcalloc(count, sizeof(*q_id), GFP_KERNEL);
+	if (!q_id) {
+		kfree(teid);
+		return -ENOMEM;
+	}
+
+	vsi_id = res->res[0].res.qsets.vsi_id;
+	vsi = ice_find_vsi(pf, vsi_id);
+	if (!vsi) {
+		dev_err(dev, "RDMA Invalid VSI\n");
+		ret = -EINVAL;
+		goto rdma_free_out;
+	}
+
+	for (i = 0; i < count; i++) {
+		struct iidc_rdma_qset_params *qset;
+
+		qset = &res->res[i].res.qsets;
+		if (qset->vsi_id != vsi_id) {
+			dev_err(dev, "RDMA Invalid VSI ID\n");
+			ret = -EINVAL;
+			goto rdma_free_out;
+		}
+		q_id[i] = qset->qs_handle;
+		teid[i] = qset->teid;
+
+		vsi->qset_handle[qset->tc] = 0;
+	}
+
+	status = ice_dis_vsi_rdma_qset(vsi->port_info, count, teid, q_id);
+	if (status)
+		ret = -EINVAL;
+
+rdma_free_out:
+	kfree(teid);
+	kfree(q_id);
+
+	return ret;
+}
+
+/**
+ * ice_peer_alloc_res - Allocate requested resources for peer device
+ * @peer_dev: peer that is requesting resources
+ * @res: Resources to be allocated
+ * @partial_acceptable: If partial allocation is acceptable to the peer
+ *
+ * This function allocates requested resources for the peer device.
+ */
+static int
+ice_peer_alloc_res(struct iidc_peer_dev *peer_dev, struct iidc_res *res,
+		   int partial_acceptable)
+{
+	struct ice_pf *pf;
+	int ret;
+
+	if (!ice_validate_peer_dev(peer_dev) || !res)
+		return -EINVAL;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	if (!ice_pf_state_is_nominal(pf))
+		return -EBUSY;
+
+	switch (res->res_type) {
+	case IIDC_RDMA_QSETS_TXSCHED:
+		ret = ice_peer_alloc_rdma_qsets(peer_dev, res,
+						partial_acceptable);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}
+
+/**
+ * ice_peer_free_res - Free given resources
+ * @peer_dev: peer that is requesting freeing of resources
+ * @res: Resources to be freed
+ *
+ * Free/Release resources allocated to given peer device.
+ */
+static int
+ice_peer_free_res(struct iidc_peer_dev *peer_dev, struct iidc_res *res)
+{
+	int ret;
+
+	if (!ice_validate_peer_dev(peer_dev) || !res)
+		return -EINVAL;
+
+	switch (res->res_type) {
+	case IIDC_RDMA_QSETS_TXSCHED:
+		ret = ice_peer_free_rdma_qsets(peer_dev, res);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}
+
 /**
  * ice_peer_unregister - request to unregister peer
  * @peer_dev: peer device
@@ -511,6 +753,8 @@ ice_peer_update_vsi_filter(struct iidc_peer_dev *peer_dev,
 
 /* Initialize the ice_ops struct, which is used in 'ice_init_peer_devices' */
 static const struct iidc_ops ops = {
+	.alloc_res			= ice_peer_alloc_res,
+	.free_res			= ice_peer_free_res,
 	.peer_register			= ice_peer_register,
 	.peer_unregister		= ice_peer_unregister,
 	.update_vsi_filter		= ice_peer_update_vsi_filter,
diff --git a/drivers/net/ethernet/intel/ice/ice_sched.c b/drivers/net/ethernet/intel/ice/ice_sched.c
index eae707ddf8e8..2f618d051b56 100644
--- a/drivers/net/ethernet/intel/ice/ice_sched.c
+++ b/drivers/net/ethernet/intel/ice/ice_sched.c
@@ -577,6 +577,50 @@ ice_alloc_lan_q_ctx(struct ice_hw *hw, u16 vsi_handle, u8 tc, u16 new_numqs)
 	return 0;
 }
 
+/**
+ * ice_alloc_rdma_q_ctx - allocate RDMA queue contexts for the given VSI and TC
+ * @hw: pointer to the HW struct
+ * @vsi_handle: VSI handle
+ * @tc: TC number
+ * @new_numqs: number of queues
+ */
+static enum ice_status
+ice_alloc_rdma_q_ctx(struct ice_hw *hw, u16 vsi_handle, u8 tc, u16 new_numqs)
+{
+	struct ice_vsi_ctx *vsi_ctx;
+	struct ice_q_ctx *q_ctx;
+
+	vsi_ctx = ice_get_vsi_ctx(hw, vsi_handle);
+	if (!vsi_ctx)
+		return ICE_ERR_PARAM;
+	/* allocate RDMA queue contexts */
+	if (!vsi_ctx->rdma_q_ctx[tc]) {
+		vsi_ctx->rdma_q_ctx[tc] = devm_kcalloc(ice_hw_to_dev(hw),
+						       new_numqs,
+						       sizeof(*q_ctx),
+						       GFP_KERNEL);
+		if (!vsi_ctx->rdma_q_ctx[tc])
+			return ICE_ERR_NO_MEMORY;
+		vsi_ctx->num_rdma_q_entries[tc] = new_numqs;
+		return 0;
+	}
+	/* num queues are increased, update the queue contexts */
+	if (new_numqs > vsi_ctx->num_rdma_q_entries[tc]) {
+		u16 prev_num = vsi_ctx->num_rdma_q_entries[tc];
+
+		q_ctx = devm_kcalloc(ice_hw_to_dev(hw), new_numqs,
+				     sizeof(*q_ctx), GFP_KERNEL);
+		if (!q_ctx)
+			return ICE_ERR_NO_MEMORY;
+		memcpy(q_ctx, vsi_ctx->rdma_q_ctx[tc],
+		       prev_num * sizeof(*q_ctx));
+		devm_kfree(ice_hw_to_dev(hw), vsi_ctx->rdma_q_ctx[tc]);
+		vsi_ctx->rdma_q_ctx[tc] = q_ctx;
+		vsi_ctx->num_rdma_q_entries[tc] = new_numqs;
+	}
+	return 0;
+}
+
 /**
  * ice_aq_rl_profile - performs a rate limiting task
  * @hw: pointer to the HW struct
@@ -1599,13 +1643,22 @@ ice_sched_update_vsi_child_nodes(struct ice_port_info *pi, u16 vsi_handle,
 	if (!vsi_ctx)
 		return ICE_ERR_PARAM;
 
-	prev_numqs = vsi_ctx->sched.max_lanq[tc];
+	if (owner == ICE_SCHED_NODE_OWNER_LAN)
+		prev_numqs = vsi_ctx->sched.max_lanq[tc];
+	else
+		prev_numqs = vsi_ctx->sched.max_rdmaq[tc];
 	/* num queues are not changed or less than the previous number */
 	if (new_numqs <= prev_numqs)
 		return status;
-	status = ice_alloc_lan_q_ctx(hw, vsi_handle, tc, new_numqs);
-	if (status)
-		return status;
+	if (owner == ICE_SCHED_NODE_OWNER_LAN) {
+		status = ice_alloc_lan_q_ctx(hw, vsi_handle, tc, new_numqs);
+		if (status)
+			return status;
+	} else {
+		status = ice_alloc_rdma_q_ctx(hw, vsi_handle, tc, new_numqs);
+		if (status)
+			return status;
+	}
 
 	if (new_numqs)
 		ice_sched_calc_vsi_child_nodes(hw, new_numqs, new_num_nodes);
@@ -1620,7 +1673,10 @@ ice_sched_update_vsi_child_nodes(struct ice_port_info *pi, u16 vsi_handle,
 					       new_num_nodes, owner);
 	if (status)
 		return status;
-	vsi_ctx->sched.max_lanq[tc] = new_numqs;
+	if (owner == ICE_SCHED_NODE_OWNER_LAN)
+		vsi_ctx->sched.max_lanq[tc] = new_numqs;
+	else
+		vsi_ctx->sched.max_rdmaq[tc] = new_numqs;
 
 	return 0;
 }
@@ -1686,6 +1742,7 @@ ice_sched_cfg_vsi(struct ice_port_info *pi, u16 vsi_handle, u8 tc, u16 maxqs,
 		 * recreate the child nodes all the time in these cases.
 		 */
 		vsi_ctx->sched.max_lanq[tc] = 0;
+		vsi_ctx->sched.max_rdmaq[tc] = 0;
 	}
 
 	/* update the VSI child nodes */
@@ -1817,6 +1874,8 @@ ice_sched_rm_vsi_cfg(struct ice_port_info *pi, u16 vsi_handle, u8 owner)
 		}
 		if (owner == ICE_SCHED_NODE_OWNER_LAN)
 			vsi_ctx->sched.max_lanq[i] = 0;
+		else
+			vsi_ctx->sched.max_rdmaq[i] = 0;
 	}
 	status = 0;
 
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.c b/drivers/net/ethernet/intel/ice/ice_switch.c
index cf8e1553599a..eeb1b0e6f716 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.c
+++ b/drivers/net/ethernet/intel/ice/ice_switch.c
@@ -310,6 +310,10 @@ static void ice_clear_vsi_q_ctx(struct ice_hw *hw, u16 vsi_handle)
 			devm_kfree(ice_hw_to_dev(hw), vsi->lan_q_ctx[i]);
 			vsi->lan_q_ctx[i] = NULL;
 		}
+		if (vsi->rdma_q_ctx[i]) {
+			devm_kfree(ice_hw_to_dev(hw), vsi->rdma_q_ctx[i]);
+			vsi->rdma_q_ctx[i] = NULL;
+		}
 	}
 }
 
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.h b/drivers/net/ethernet/intel/ice/ice_switch.h
index 96010d3d96fd..acd2f150c30b 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.h
+++ b/drivers/net/ethernet/intel/ice/ice_switch.h
@@ -26,6 +26,8 @@ struct ice_vsi_ctx {
 	u8 vf_num;
 	u16 num_lan_q_entries[ICE_MAX_TRAFFIC_CLASS];
 	struct ice_q_ctx *lan_q_ctx[ICE_MAX_TRAFFIC_CLASS];
+	u16 num_rdma_q_entries[ICE_MAX_TRAFFIC_CLASS];
+	struct ice_q_ctx *rdma_q_ctx[ICE_MAX_TRAFFIC_CLASS];
 };
 
 enum ice_sw_fwd_act_type {
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index 42b2d700bc1f..3ada92536540 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -45,6 +45,7 @@ static inline u32 ice_round_to_num(u32 N, u32 R)
 #define ICE_DBG_FLOW		BIT_ULL(9)
 #define ICE_DBG_SW		BIT_ULL(13)
 #define ICE_DBG_SCHED		BIT_ULL(14)
+#define ICE_DBG_RDMA		BIT_ULL(15)
 #define ICE_DBG_PKG		BIT_ULL(16)
 #define ICE_DBG_RES		BIT_ULL(17)
 #define ICE_DBG_AQ_MSG		BIT_ULL(24)
@@ -282,6 +283,7 @@ struct ice_sched_node {
 	u8 tc_num;
 	u8 owner;
 #define ICE_SCHED_NODE_OWNER_LAN	0
+#define ICE_SCHED_NODE_OWNER_RDMA	2
 };
 
 /* Access Macros for Tx Sched Elements data */
@@ -353,6 +355,7 @@ struct ice_sched_vsi_info {
 	struct ice_sched_node *ag_node[ICE_MAX_TRAFFIC_CLASS];
 	struct list_head list_entry;
 	u16 max_lanq[ICE_MAX_TRAFFIC_CLASS];
+	u16 max_rdmaq[ICE_MAX_TRAFFIC_CLASS];
 };
 
 /* driver defines the policy */
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 5/9] ice: Enable event notifications
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
                   ` (2 preceding siblings ...)
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 4/9] ice: Support resource allocation requests Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 6/9] ice: Allow reset operations Jeff Kirsher
                   ` (3 subsequent siblings)
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

Enable registration of notifications. Peer devices can register to be
notified of certain events as well as notify the driver of its state
changes.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_dcb_lib.c |  37 ++++
 drivers/net/ethernet/intel/ice/ice_idc.c     | 221 +++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_idc_int.h |   1 +
 drivers/net/ethernet/intel/ice/ice_main.c    |  27 ++-
 4 files changed, 280 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
index 24c0a60fe172..c4f8be0c0b24 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
@@ -168,6 +168,30 @@ void ice_vsi_cfg_dcb_rings(struct ice_vsi *vsi)
 	}
 }
 
+/**
+ * ice_peer_prep_tc_change - Pre-notify RDMA Peer in blocking call of TC change
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data
+ */
+static int
+ice_peer_prep_tc_change(struct ice_peer_dev_int *peer_dev_int,
+			void __always_unused *data)
+{
+	struct iidc_peer_dev *peer_dev;
+
+	peer_dev = &peer_dev_int->peer_dev;
+	if (!ice_validate_peer_dev(peer_dev))
+		return 0;
+
+	if (!test_bit(ICE_PEER_DEV_STATE_OPENED, peer_dev_int->state))
+		return 0;
+
+	if (peer_dev->peer_ops && peer_dev->peer_ops->prep_tc_change)
+		peer_dev->peer_ops->prep_tc_change(peer_dev);
+
+	return 0;
+}
+
 /**
  * ice_dcb_bwchk - check if ETS bandwidth input parameters are correct
  * @pf: pointer to the PF struct
@@ -248,6 +272,9 @@ int ice_pf_dcb_cfg(struct ice_pf *pf, struct ice_dcbx_cfg *new_cfg, bool locked)
 		return -ENOMEM;
 
 	dev_info(dev, "Commit DCB Configuration to the hardware\n");
+	/* Notify capable peers about impending change to TCs */
+	ice_for_each_peer(pf, NULL, ice_peer_prep_tc_change);
+
 	pf_vsi = ice_get_main_vsi(pf);
 	if (!pf_vsi) {
 		dev_dbg(dev, "PF VSI doesn't exist\n");
@@ -580,6 +607,7 @@ static int ice_dcb_noncontig_cfg(struct ice_pf *pf)
 void ice_pf_dcb_recfg(struct ice_pf *pf)
 {
 	struct ice_dcbx_cfg *dcbcfg = &pf->hw.port_info->local_dcbx_cfg;
+	struct iidc_event *event;
 	u8 tc_map = 0;
 	int v, ret;
 
@@ -615,6 +643,15 @@ void ice_pf_dcb_recfg(struct ice_pf *pf)
 		if (vsi->type == ICE_VSI_PF)
 			ice_dcbnl_set_all(vsi);
 	}
+	event = kzalloc(sizeof(*event), GFP_KERNEL);
+	if (!event)
+		return;
+
+	set_bit(IIDC_EVENT_TC_CHANGE, event->type);
+	event->reporter = NULL;
+	ice_setup_dcb_qos_info(pf, &event->info.port_qos);
+	ice_for_each_peer(pf, event, ice_peer_check_for_reg);
+	kfree(event);
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 05fa5c61e2d3..0fb1080c19d7 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -218,6 +218,72 @@ int ice_peer_update_vsi(struct ice_peer_dev_int *peer_dev_int, void *data)
 	return 0;
 }
 
+/**
+ * ice_check_peer_drv_for_events - check peer_drv for events to report
+ * @peer_dev: peer device to report to
+ */
+static void ice_check_peer_drv_for_events(struct iidc_peer_dev *peer_dev)
+{
+	const struct iidc_peer_ops *p_ops = peer_dev->peer_ops;
+	struct ice_peer_dev_int *peer_dev_int;
+	struct ice_peer_drv_int *peer_drv_int;
+	int i;
+
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+	if (!peer_dev_int)
+		return;
+	peer_drv_int = peer_dev_int->peer_drv_int;
+
+	for_each_set_bit(i, peer_dev_int->events, IIDC_EVENT_NBITS) {
+		struct iidc_event *curr = &peer_drv_int->current_events[i];
+
+		if (!bitmap_empty(curr->type, IIDC_EVENT_NBITS) &&
+		    p_ops->event_handler)
+			p_ops->event_handler(peer_dev, curr);
+	}
+}
+
+/**
+ * ice_check_peer_for_events - check peer_devs for events new peer reg'd for
+ * @src_peer_int: peer to check for events
+ * @data: ptr to opaque data, to be used for the peer struct that opened
+ *
+ * This function is to be called when a peer device is opened.
+ *
+ * Since a new peer opening would have missed any events that would
+ * have happened before its opening, we need to walk the peers and see
+ * if any of them have events that the new peer cares about
+ *
+ * This function is meant to be called by a device_for_each_child.
+ */
+static int
+ice_check_peer_for_events(struct ice_peer_dev_int *src_peer_int, void *data)
+{
+	struct iidc_peer_dev *new_peer = (struct iidc_peer_dev *)data;
+	const struct iidc_peer_ops *p_ops = new_peer->peer_ops;
+	struct ice_peer_dev_int *new_peer_int;
+	struct iidc_peer_dev *src_peer;
+	int i;
+
+	src_peer = &src_peer_int->peer_dev;
+	if (!ice_validate_peer_dev(new_peer) ||
+	    !ice_validate_peer_dev(src_peer))
+		return 0;
+
+	new_peer_int = peer_to_ice_dev_int(new_peer);
+
+	for_each_set_bit(i, new_peer_int->events, IIDC_EVENT_NBITS) {
+		struct iidc_event *curr = &src_peer_int->current_events[i];
+
+		if (!bitmap_empty(curr->type, IIDC_EVENT_NBITS) &&
+		    new_peer->peer_dev_id != src_peer->peer_dev_id &&
+		    p_ops->event_handler)
+			p_ops->event_handler(new_peer, curr);
+	}
+
+	return 0;
+}
+
 /**
  * ice_for_each_peer - iterate across and call function for each peer dev
  * @pf: pointer to private board struct
@@ -323,6 +389,9 @@ ice_finish_init_peer_device(struct ice_peer_dev_int *peer_dev_int,
 
 		ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_OPENED,
 				      true);
+		ret = ice_for_each_peer(pf, peer_dev,
+					ice_check_peer_for_events);
+		ice_check_peer_drv_for_events(peer_dev);
 	}
 
 init_unlock:
@@ -630,6 +699,155 @@ ice_peer_free_res(struct iidc_peer_dev *peer_dev, struct iidc_res *res)
 	return ret;
 }
 
+/**
+ * ice_peer_reg_for_notif - register a peer to receive specific notifications
+ * @peer_dev: peer that is registering for event notifications
+ * @events: mask of event types peer is registering for
+ */
+static void
+ice_peer_reg_for_notif(struct iidc_peer_dev *peer_dev,
+		       struct iidc_event *events)
+{
+	struct ice_peer_dev_int *peer_dev_int;
+	struct ice_pf *pf;
+
+	if (!ice_validate_peer_dev(peer_dev) || !events)
+		return;
+
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+	pf = pci_get_drvdata(peer_dev->pdev);
+
+	bitmap_or(peer_dev_int->events, peer_dev_int->events, events->type,
+		  IIDC_EVENT_NBITS);
+
+	/* Check to see if any events happened previous to peer registering */
+	ice_for_each_peer(pf, peer_dev, ice_check_peer_for_events);
+	ice_check_peer_drv_for_events(peer_dev);
+}
+
+/**
+ * ice_peer_unreg_for_notif - unreg a peer from receiving certain notifications
+ * @peer_dev: peer that is unregistering from event notifications
+ * @events: mask of event types peer is unregistering for
+ */
+static void
+ice_peer_unreg_for_notif(struct iidc_peer_dev *peer_dev,
+			 struct iidc_event *events)
+{
+	struct ice_peer_dev_int *peer_dev_int;
+
+	if (!ice_validate_peer_dev(peer_dev) || !events)
+		return;
+
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+
+	bitmap_andnot(peer_dev_int->events, peer_dev_int->events, events->type,
+		      IIDC_EVENT_NBITS);
+}
+
+/**
+ * ice_peer_check_for_reg - check to see if any peers are reg'd for event
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data, to be used for ice_event to report
+ *
+ * This function is to be called by device_for_each_child to handle an
+ * event reported by a peer or the ice driver.
+ */
+int ice_peer_check_for_reg(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+	struct iidc_event *event = (struct iidc_event *)data;
+	DECLARE_BITMAP(comp_events, IIDC_EVENT_NBITS);
+	struct iidc_peer_dev *peer_dev;
+	bool check = true;
+
+	peer_dev = &peer_dev_int->peer_dev;
+
+	if (!ice_validate_peer_dev(peer_dev) || !data)
+	/* If invalid dev, in this case return 0 instead of error
+	 * because caller ignores this return value
+	 */
+		return 0;
+
+	if (event->reporter)
+		check = event->reporter->peer_dev_id != peer_dev->peer_dev_id;
+
+	if (bitmap_and(comp_events, event->type, peer_dev_int->events,
+		       IIDC_EVENT_NBITS) &&
+	    (test_bit(ICE_PEER_DEV_STATE_OPENED, peer_dev_int->state) ||
+	     test_bit(ICE_PEER_DEV_STATE_PREP_RST, peer_dev_int->state) ||
+	     test_bit(ICE_PEER_DEV_STATE_PREPPED, peer_dev_int->state)) &&
+	    check &&
+	    peer_dev->peer_ops->event_handler)
+		peer_dev->peer_ops->event_handler(peer_dev, event);
+
+	return 0;
+}
+
+/**
+ * ice_peer_report_state_change - accept report of a peer state change
+ * @peer_dev: peer that is sending notification about state change
+ * @event: ice_event holding info on what the state change is
+ *
+ * We also need to parse the list of peers to see if anyone is registered
+ * for notifications about this state change event, and if so, notify them.
+ */
+static void
+ice_peer_report_state_change(struct iidc_peer_dev *peer_dev,
+			     struct iidc_event *event)
+{
+	struct ice_peer_dev_int *peer_dev_int;
+	struct ice_peer_drv_int *peer_drv_int;
+	int e_type, drv_event = 0;
+	struct ice_pf *pf;
+
+	if (!ice_validate_peer_dev(peer_dev) || !event)
+		return;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	peer_dev_int = peer_to_ice_dev_int(peer_dev);
+	peer_drv_int = peer_dev_int->peer_drv_int;
+
+	e_type = find_first_bit(event->type, IIDC_EVENT_NBITS);
+	if (!e_type)
+		return;
+
+	switch (e_type) {
+	/* Check for peer_drv events */
+	case IIDC_EVENT_MBX_CHANGE:
+		drv_event = 1;
+		if (event->info.mbx_rdy)
+			set_bit(ICE_PEER_DRV_STATE_MBX_RDY,
+				peer_drv_int->state);
+		else
+			clear_bit(ICE_PEER_DRV_STATE_MBX_RDY,
+				  peer_drv_int->state);
+		break;
+
+	/* Check for peer_dev events */
+	case IIDC_EVENT_API_CHANGE:
+		if (event->info.api_rdy)
+			set_bit(ICE_PEER_DEV_STATE_API_RDY,
+				peer_dev_int->state);
+		else
+			clear_bit(ICE_PEER_DEV_STATE_API_RDY,
+				  peer_dev_int->state);
+		break;
+
+	default:
+		return;
+	}
+
+	/* store the event and state to notify any new peers opening */
+	if (drv_event)
+		memcpy(&peer_drv_int->current_events[e_type], event,
+		       sizeof(*event));
+	else
+		memcpy(&peer_dev_int->current_events[e_type], event,
+		       sizeof(*event));
+
+	ice_for_each_peer(pf, event, ice_peer_check_for_reg);
+}
+
 /**
  * ice_peer_unregister - request to unregister peer
  * @peer_dev: peer device
@@ -755,6 +973,9 @@ ice_peer_update_vsi_filter(struct iidc_peer_dev *peer_dev,
 static const struct iidc_ops ops = {
 	.alloc_res			= ice_peer_alloc_res,
 	.free_res			= ice_peer_free_res,
+	.reg_for_notification		= ice_peer_reg_for_notif,
+	.unreg_for_notification		= ice_peer_unreg_for_notif,
+	.notify_state_change		= ice_peer_report_state_change,
 	.peer_register			= ice_peer_register,
 	.peer_unregister		= ice_peer_unregister,
 	.update_vsi_filter		= ice_peer_update_vsi_filter,
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
index d22e6f5bb50e..1d3d5cafc977 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc_int.h
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -66,6 +66,7 @@ int ice_peer_update_vsi(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_peer_close(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_peer_check_for_reg(struct ice_peer_dev_int *peer_dev_int, void *data);
 int
 ice_finish_init_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index ac0c6d5b01e4..d1a528da9128 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -4862,7 +4862,9 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu)
 	struct ice_netdev_priv *np = netdev_priv(netdev);
 	struct ice_vsi *vsi = np->vsi;
 	struct ice_pf *pf = vsi->back;
+	struct iidc_event *event;
 	u8 count = 0;
+	int err = 0;
 
 	if (new_mtu == netdev->mtu) {
 		netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
@@ -4904,27 +4906,40 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu)
 		return -EBUSY;
 	}
 
+	event = kzalloc(sizeof(*event), GFP_KERNEL);
+	if (!event)
+		return -ENOMEM;
+
 	netdev->mtu = new_mtu;
 
 	/* if VSI is up, bring it down and then back up */
 	if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
-		int err;
-
 		err = ice_down(vsi);
 		if (err) {
-			netdev_err(netdev, "change MTU if_up err %d\n", err);
-			return err;
+			netdev_err(netdev, "change MTU if_down err %d\n", err);
+			goto free_event;
 		}
 
 		err = ice_up(vsi);
 		if (err) {
 			netdev_err(netdev, "change MTU if_up err %d\n", err);
-			return err;
+			goto free_event;
 		}
 	}
 
+	if (ice_is_safe_mode(pf))
+		goto out;
+
+	set_bit(IIDC_EVENT_MTU_CHANGE, event->type);
+	event->reporter = NULL;
+	event->info.mtu = new_mtu;
+	ice_for_each_peer(pf, event, ice_peer_check_for_reg);
+
+out:
 	netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
-	return 0;
+free_event:
+	kfree(event);
+	return err;
 }
 
 /**
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 6/9] ice: Allow reset operations
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
                   ` (3 preceding siblings ...)
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 5/9] ice: Enable event notifications Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 7/9] ice: Pass through communications to VF Jeff Kirsher
                   ` (2 subsequent siblings)
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

Enable the PF to notify peers when it's going to reset so that peer devices
can prepare accordingly. Also enable the peer devices to request the PF to
reset.

Implement ice_peer_is_vsi_ready() so the peer device can determine when the
VSI is ready for operations following a reset.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_idc.c     | 140 +++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_idc_int.h |   1 +
 drivers/net/ethernet/intel/ice/ice_lib.c     |   6 +
 drivers/net/ethernet/intel/ice/ice_main.c    |   3 +
 4 files changed, 150 insertions(+)

diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 0fb1080c19d7..748e9134a113 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -218,6 +218,40 @@ int ice_peer_update_vsi(struct ice_peer_dev_int *peer_dev_int, void *data)
 	return 0;
 }
 
+/**
+ * ice_close_peer_for_reset - queue work to close peer for reset
+ * @peer_dev_int: pointer peer dev internal struct
+ * @data: pointer to opaque data used for reset type
+ */
+int ice_close_peer_for_reset(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+	struct iidc_peer_dev *peer_dev;
+	enum ice_reset_req reset;
+
+	peer_dev = &peer_dev_int->peer_dev;
+	if (!ice_validate_peer_dev(peer_dev))
+		return 0;
+
+	reset = *(enum ice_reset_req *)data;
+
+	switch (reset) {
+	case ICE_RESET_GLOBR:
+		peer_dev_int->rst_type = IIDC_REASON_GLOBR_REQ;
+		break;
+	case ICE_RESET_CORER:
+		peer_dev_int->rst_type = IIDC_REASON_CORER_REQ;
+		break;
+	case ICE_RESET_PFR:
+		peer_dev_int->rst_type = IIDC_REASON_PFR_REQ;
+		break;
+	default:
+		/* reset type is invalid */
+		return 1;
+	}
+	queue_work(peer_dev_int->ice_peer_wq, &peer_dev_int->peer_close_task);
+	return 0;
+}
+
 /**
  * ice_check_peer_drv_for_events - check peer_drv for events to report
  * @peer_dev: peer device to report to
@@ -930,6 +964,74 @@ static int ice_peer_register(struct iidc_peer_dev *peer_dev)
 	return 0;
 }
 
+/**
+ * ice_peer_request_reset - accept request from peer to perform a reset
+ * @peer_dev: peer device that is request a reset
+ * @reset_type: type of reset the peer is requesting
+ */
+static int
+ice_peer_request_reset(struct iidc_peer_dev *peer_dev,
+		       enum iidc_peer_reset_type reset_type)
+{
+	enum ice_reset_req reset;
+	struct ice_pf *pf;
+
+	if (!ice_validate_peer_dev(peer_dev))
+		return -EINVAL;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+
+	switch (reset_type) {
+	case IIDC_PEER_PFR:
+		reset = ICE_RESET_PFR;
+		break;
+	case IIDC_PEER_CORER:
+		reset = ICE_RESET_CORER;
+		break;
+	case IIDC_PEER_GLOBR:
+		reset = ICE_RESET_GLOBR;
+		break;
+	default:
+		dev_err(ice_pf_to_dev(pf), "incorrect reset request from peer\n");
+		return -EINVAL;
+	}
+
+	return ice_schedule_reset(pf, reset);
+}
+
+/**
+ * ice_peer_is_vsi_ready - query if VSI in nominal state
+ * @peer_dev: pointer to iidc_peer_dev struct
+ */
+static int ice_peer_is_vsi_ready(struct iidc_peer_dev *peer_dev)
+{
+	DECLARE_BITMAP(check_bits, __ICE_STATE_NBITS) = { 0 };
+	struct ice_netdev_priv *np;
+	struct ice_vsi *vsi;
+
+	/* If the peer_dev or associated values are not valid, then return
+	 * 0 as there is no ready port associated with the values passed in
+	 * as parameters.
+	 */
+
+	if (!ice_validate_peer_dev(peer_dev))
+		return 0;
+
+	if (!peer_dev->netdev)
+		return 0;
+
+	np = netdev_priv(peer_dev->netdev);
+	vsi = np->vsi;
+	if (!vsi)
+		return 0;
+
+	bitmap_set(check_bits, 0, __ICE_STATE_NOMINAL_CHECK_BITS);
+	if (bitmap_intersects(vsi->state, check_bits, __ICE_STATE_NBITS))
+		return 0;
+
+	return 1;
+}
+
 /**
  * ice_peer_update_vsi_filter - update main VSI filters for RDMA
  * @peer_dev: pointer to RDMA peer device
@@ -973,9 +1075,11 @@ ice_peer_update_vsi_filter(struct iidc_peer_dev *peer_dev,
 static const struct iidc_ops ops = {
 	.alloc_res			= ice_peer_alloc_res,
 	.free_res			= ice_peer_free_res,
+	.is_vsi_ready			= ice_peer_is_vsi_ready,
 	.reg_for_notification		= ice_peer_reg_for_notif,
 	.unreg_for_notification		= ice_peer_unreg_for_notif,
 	.notify_state_change		= ice_peer_report_state_change,
+	.request_reset			= ice_peer_request_reset,
 	.peer_register			= ice_peer_register,
 	.peer_unregister		= ice_peer_unregister,
 	.update_vsi_filter		= ice_peer_update_vsi_filter,
@@ -1000,6 +1104,41 @@ static int ice_reserve_peer_qvector(struct ice_pf *pf)
 	return 0;
 }
 
+/**
+ * ice_peer_close_task - call peer's close asynchronously
+ * @work: pointer to work_struct contained by the peer_dev_int struct
+ *
+ * This method (asynchronous) of calling a peer's close function is
+ * meant to be used in the reset path.
+ */
+static void ice_peer_close_task(struct work_struct *work)
+{
+	struct ice_peer_dev_int *peer_dev_int;
+	struct iidc_peer_dev *peer_dev;
+
+	peer_dev_int = container_of(work, struct ice_peer_dev_int,
+				    peer_close_task);
+
+	peer_dev = &peer_dev_int->peer_dev;
+	if (!peer_dev || !peer_dev->peer_ops)
+		return;
+
+	/* If this peer_dev is going to close, we do not want any state changes
+	 * to happen until after we successfully finish or abort the close.
+	 * Grab the peer_dev_state_mutex to protect this flow
+	 */
+	mutex_lock(&peer_dev_int->peer_dev_state_mutex);
+
+	ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSING, true);
+
+	if (peer_dev->peer_ops->close)
+		peer_dev->peer_ops->close(peer_dev, peer_dev_int->rst_type);
+
+	ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSED, true);
+
+	mutex_unlock(&peer_dev_int->peer_dev_state_mutex);
+}
+
 /**
  * ice_peer_vdev_release - function to map to virtbus_devices release callback
  * @vdev: pointer to virtbus_device to free
@@ -1098,6 +1237,7 @@ int ice_init_peer_devices(struct ice_pf *pf)
 			kfree(vbo);
 			goto unroll_prev_peers;
 		}
+		INIT_WORK(&peer_dev_int->peer_close_task, ice_peer_close_task);
 
 		peer_dev->pdev = pdev;
 		qos_info = &peer_dev->initial_qos_info;
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
index 1d3d5cafc977..90e165434aea 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc_int.h
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -63,6 +63,7 @@ struct ice_peer_dev_int {
 };
 
 int ice_peer_update_vsi(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_close_peer_for_reset(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
 int ice_peer_close(struct ice_peer_dev_int *peer_dev_int, void *data);
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 5043d5ed1b2a..34b41b1039f1 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -2416,6 +2416,12 @@ void ice_vsi_close(struct ice_vsi *vsi)
 {
 	enum iidc_close_reason reason = IIDC_REASON_INTERFACE_DOWN;
 
+	if (test_bit(__ICE_CORER_REQ, vsi->back->state))
+		reason = IIDC_REASON_CORER_REQ;
+	if (test_bit(__ICE_GLOBR_REQ, vsi->back->state))
+		reason = IIDC_REASON_GLOBR_REQ;
+	if (test_bit(__ICE_PFR_REQ, vsi->back->state))
+		reason = IIDC_REASON_PFR_REQ;
 	if (!ice_is_safe_mode(vsi->back) && vsi->type == ICE_VSI_PF) {
 		int ret = ice_for_each_peer(vsi->back, &reason, ice_peer_close);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index d1a528da9128..c7eb51bae33d 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -560,6 +560,9 @@ static void ice_reset_subtask(struct ice_pf *pf)
 		/* return if no valid reset type requested */
 		if (reset_type == ICE_RESET_INVAL)
 			return;
+		if (ice_is_peer_ena(pf))
+			ice_for_each_peer(pf, &reset_type,
+					  ice_close_peer_for_reset);
 		ice_prepare_for_reset(pf);
 
 		/* make sure we are ready to rebuild */
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 7/9] ice: Pass through communications to VF
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
                   ` (4 preceding siblings ...)
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 6/9] ice: Allow reset operations Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 8/9] i40e: Move client header location Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 9/9] i40e: Register a virtbus device to provide RDMA Jeff Kirsher
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Dave Ertman <david.m.ertman@intel.com>

Allow for forwarding of RDMA and VF virt channel messages. The driver will
forward messages from the RDMA driver to the VF via the vc_send operation
and invoke the peer's vc_receive() call when receiving a virt channel
message destined for the peer driver.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h          |  1 +
 drivers/net/ethernet/intel/ice/ice_idc.c      | 34 +++++++++++++++++++
 .../net/ethernet/intel/ice/ice_virtchnl_pf.c  | 34 +++++++++++++++++++
 3 files changed, 69 insertions(+)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 6ad1894eca3f..0e45e080a41f 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -392,6 +392,7 @@ struct ice_pf {
 	u32 msg_enable;
 	u32 num_rdma_msix;	/* Total MSIX vectors for RDMA driver */
 	u32 rdma_base_vector;
+	struct iidc_peer_dev *rdma_peer;
 	u32 hw_csum_rx_error;
 	u32 oicr_idx;		/* Other interrupt cause MSIX vector index */
 	u32 num_avail_sw_msix;	/* remaining MSIX SW vectors left unclaimed */
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 748e9134a113..d287728b3cc8 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -1071,6 +1071,38 @@ ice_peer_update_vsi_filter(struct iidc_peer_dev *peer_dev,
 	return ret;
 }
 
+/**
+ * ice_peer_vc_send - send a virt channel message from RDMA peer
+ * @peer_dev: pointer to RDMA peer dev
+ * @vf_id: the absolute VF ID of recipient of message
+ * @msg: pointer to message contents
+ * @len: len of message
+ */
+static int
+ice_peer_vc_send(struct iidc_peer_dev *peer_dev, u32 vf_id, u8 *msg, u16 len)
+{
+	struct ice_pf *pf;
+	int err;
+
+	if (!ice_validate_peer_dev(peer_dev))
+		return -EINVAL;
+	if (!msg || !len)
+		return -ENOMEM;
+
+	pf = pci_get_drvdata(peer_dev->pdev);
+	if (vf_id >= pf->num_alloc_vfs || len > ICE_AQ_MAX_BUF_LEN)
+		return -EINVAL;
+
+	/* VIRTCHNL_OP_IWARP is being used for RoCEv2 msg also */
+	err = ice_aq_send_msg_to_vf(&pf->hw, vf_id, VIRTCHNL_OP_IWARP, 0, msg,
+				    len, NULL);
+	if (err)
+		dev_err(ice_pf_to_dev(pf), "Unable to send RDMA msg to VF, error %d\n",
+			err);
+
+	return err;
+}
+
 /* Initialize the ice_ops struct, which is used in 'ice_init_peer_devices' */
 static const struct iidc_ops ops = {
 	.alloc_res			= ice_peer_alloc_res,
@@ -1083,6 +1115,7 @@ static const struct iidc_ops ops = {
 	.peer_register			= ice_peer_register,
 	.peer_unregister		= ice_peer_unregister,
 	.update_vsi_filter		= ice_peer_update_vsi_filter,
+	.vc_send			= ice_peer_vc_send,
 };
 
 /**
@@ -1264,6 +1297,7 @@ int ice_init_peer_devices(struct ice_pf *pf)
 		switch (ice_peers[i].id) {
 		case IIDC_PEER_RDMA_ID:
 			if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+				pf->rdma_peer = peer_dev;
 				peer_dev->msix_count = pf->num_rdma_msix;
 				entry = &pf->msix_entries[pf->rdma_base_vector];
 			}
diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
index 07f3d4b456c7..95e39fef0a26 100644
--- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
@@ -3170,6 +3170,37 @@ static int ice_vc_dis_vlan_stripping(struct ice_vf *vf)
 				     v_ret, NULL, 0);
 }
 
+/**
+ * ice_vc_rdma_msg - send msg to RDMA PF from VF
+ * @vf: pointer to VF info
+ * @msg: pointer to msg buffer
+ * @len: length of the message
+ *
+ * This function is called indirectly from the AQ clean function.
+ */
+static int ice_vc_rdma_msg(struct ice_vf *vf, u8 *msg, u16 len)
+{
+	struct iidc_peer_dev *rdma_peer;
+	int ret;
+
+	rdma_peer = vf->pf->rdma_peer;
+	if (!rdma_peer) {
+		pr_err("Invalid RDMA peer attempted to send message to peer\n");
+		return -EIO;
+	}
+
+	if (!rdma_peer->peer_ops || !rdma_peer->peer_ops->vc_receive) {
+		pr_err("Incomplete RMDA peer attempting to send msg\n");
+		return -EINVAL;
+	}
+
+	ret = rdma_peer->peer_ops->vc_receive(rdma_peer, vf->vf_id, msg, len);
+	if (ret)
+		pr_err("Failed to send message to RDMA peer, error %d\n", ret);
+
+	return ret;
+}
+
 /**
  * ice_vf_init_vlan_stripping - enable/disable VLAN stripping on initialization
  * @vf: VF to enable/disable VLAN stripping for on initialization
@@ -3304,6 +3335,9 @@ void ice_vc_process_vf_msg(struct ice_pf *pf, struct ice_rq_event_info *event)
 	case VIRTCHNL_OP_DISABLE_VLAN_STRIPPING:
 		err = ice_vc_dis_vlan_stripping(vf);
 		break;
+	case VIRTCHNL_OP_IWARP:
+		err = ice_vc_rdma_msg(vf, msg, msglen);
+		break;
 	case VIRTCHNL_OP_UNKNOWN:
 	default:
 		dev_err(dev, "Unsupported opcode %d from VF %d\n", v_opcode,
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 8/9] i40e: Move client header location
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
                   ` (5 preceding siblings ...)
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 7/9] ice: Pass through communications to VF Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 9/9] i40e: Register a virtbus device to provide RDMA Jeff Kirsher
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Shiraz Saleem <shiraz.saleem@intel.com>

Move i40e_client.h to include/linux/net/intel/*
since its shared between i40iw and i40e.

Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/infiniband/hw/i40iw/Makefile                            | 1 -
 drivers/infiniband/hw/i40iw/i40iw.h                             | 2 +-
 drivers/net/ethernet/intel/i40e/i40e.h                          | 2 +-
 drivers/net/ethernet/intel/i40e/i40e_client.c                   | 2 +-
 .../intel/i40e => include/linux/net/intel}/i40e_client.h        | 0
 5 files changed, 3 insertions(+), 4 deletions(-)
 rename {drivers/net/ethernet/intel/i40e => include/linux/net/intel}/i40e_client.h (100%)

diff --git a/drivers/infiniband/hw/i40iw/Makefile b/drivers/infiniband/hw/i40iw/Makefile
index 8942f8229945..34da9eba8a7c 100644
--- a/drivers/infiniband/hw/i40iw/Makefile
+++ b/drivers/infiniband/hw/i40iw/Makefile
@@ -1,5 +1,4 @@
 # SPDX-License-Identifier: GPL-2.0
-ccflags-y :=  -I $(srctree)/drivers/net/ethernet/intel/i40e
 
 obj-$(CONFIG_INFINIBAND_I40IW) += i40iw.o
 
diff --git a/drivers/infiniband/hw/i40iw/i40iw.h b/drivers/infiniband/hw/i40iw/i40iw.h
index 3c62c9327a9c..1ba7561f9cbb 100644
--- a/drivers/infiniband/hw/i40iw/i40iw.h
+++ b/drivers/infiniband/hw/i40iw/i40iw.h
@@ -45,6 +45,7 @@
 #include <linux/slab.h>
 #include <linux/io.h>
 #include <linux/crc32c.h>
+#include <linux/net/intel/i40e_client.h>
 #include <rdma/ib_smi.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_pack.h>
@@ -57,7 +58,6 @@
 #include "i40iw_d.h"
 #include "i40iw_hmc.h"
 
-#include <i40e_client.h>
 #include "i40iw_type.h"
 #include "i40iw_p.h"
 #include <rdma/i40iw-abi.h>
diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h
index e95b8da45e07..5ff0828a6f50 100644
--- a/drivers/net/ethernet/intel/i40e/i40e.h
+++ b/drivers/net/ethernet/intel/i40e/i40e.h
@@ -38,7 +38,7 @@
 #include <net/xdp_sock.h>
 #include "i40e_type.h"
 #include "i40e_prototype.h"
-#include "i40e_client.h"
+#include <linux/net/intel/i40e_client.h>
 #include <linux/avf/virtchnl.h>
 #include "i40e_virtchnl_pf.h"
 #include "i40e_txrx.h"
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.c b/drivers/net/ethernet/intel/i40e/i40e_client.c
index e81530ca08d0..befd3018183f 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_client.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.c
@@ -3,10 +3,10 @@
 
 #include <linux/list.h>
 #include <linux/errno.h>
+#include <linux/net/intel/i40e_client.h>
 
 #include "i40e.h"
 #include "i40e_prototype.h"
-#include "i40e_client.h"
 
 static const char i40e_client_interface_version_str[] = I40E_CLIENT_VERSION_STR;
 static struct i40e_client *registered_client;
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.h b/include/linux/net/intel/i40e_client.h
similarity index 100%
rename from drivers/net/ethernet/intel/i40e/i40e_client.h
rename to include/linux/net/intel/i40e_client.h
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [Intel-wired-lan] [net-next v3 9/9] i40e: Register a virtbus device to provide RDMA
  2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
                   ` (6 preceding siblings ...)
  2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 8/9] i40e: Move client header location Jeff Kirsher
@ 2020-05-04 16:43 ` Jeff Kirsher
  7 siblings, 0 replies; 9+ messages in thread
From: Jeff Kirsher @ 2020-05-04 16:43 UTC (permalink / raw)
  To: intel-wired-lan

From: Shiraz Saleem <shiraz.saleem@intel.com>

Register client virtbus device on the virtbus for the RDMA
virtbus driver (irdma) to bind to. It allows to realize a
single RDMA driver capable of working with multiple netdev
drivers over multi-generation Intel HW supporting RDMA.
There is also no load ordering dependencies between i40e and
irdma.

Summary of changes:
* Support to add/remove virtbus devices
* Add 2 new client ops.
	* i40e_client_device_register() which is called during RDMA
	  probe() per PF. Validate client drv OPs and schedule service
	  task to call open()
	* i40e_client_device_unregister() called during RDMA remove()
	  per PF. Call client close() and release_qvlist.
* The global register/unregister calls exported for i40iw are retained
  until i40iw is removed from the kernel.

Signed-off-by: Mustafa Ismail <mustafa.ismail@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/Kconfig            |   1 +
 drivers/net/ethernet/intel/i40e/i40e_client.c | 131 +++++++++++++++---
 include/linux/net/intel/i40e_client.h         |  15 ++
 3 files changed, 127 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/intel/Kconfig b/drivers/net/ethernet/intel/Kconfig
index 1a5d51b0f294..7a61e9d5e36e 100644
--- a/drivers/net/ethernet/intel/Kconfig
+++ b/drivers/net/ethernet/intel/Kconfig
@@ -241,6 +241,7 @@ config I40E
 	tristate "Intel(R) Ethernet Controller XL710 Family support"
 	imply PTP_1588_CLOCK
 	depends on PCI
+	select VIRTUAL_BUS
 	---help---
 	  This driver supports Intel(R) Ethernet Controller XL710 Family of
 	  devices.  For more information on how to identify your adapter, go
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.c b/drivers/net/ethernet/intel/i40e/i40e_client.c
index befd3018183f..fdce8af3ec4f 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_client.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 /* Copyright(c) 2013 - 2018 Intel Corporation. */
 
+#include <linux/net/intel/i40e_client.h>
 #include <linux/list.h>
 #include <linux/errno.h>
 #include <linux/net/intel/i40e_client.h>
@@ -30,11 +31,17 @@ static int i40e_client_update_vsi_ctxt(struct i40e_info *ldev,
 				       bool is_vf, u32 vf_id,
 				       u32 flag, u32 valid_flag);
 
+static int i40e_client_device_register(struct i40e_info *ldev);
+
+static void i40e_client_device_unregister(struct i40e_info *ldev);
+
 static struct i40e_ops i40e_lan_ops = {
 	.virtchnl_send = i40e_client_virtchnl_send,
 	.setup_qvlist = i40e_client_setup_qvlist,
 	.request_reset = i40e_client_request_reset,
 	.update_vsi_ctxt = i40e_client_update_vsi_ctxt,
+	.client_device_register = i40e_client_device_register,
+	.client_device_unregister = i40e_client_device_unregister,
 };
 
 /**
@@ -275,6 +282,37 @@ void i40e_client_update_msix_info(struct i40e_pf *pf)
 	cdev->lan_info.msix_entries = &pf->msix_entries[pf->iwarp_base_vector];
 }
 
+static void i40e_virtdev_release(struct virtbus_device *vdev)
+{
+	struct i40e_virtbus_device *i40e_vdev =
+			container_of(vdev, struct i40e_virtbus_device, vdev);
+
+	kfree(i40e_vdev);
+}
+
+static int i40e_init_client_virtdev(struct i40e_info *ldev)
+{
+	struct pci_dev *pdev = ldev->pcidev;
+	struct i40e_virtbus_device *i40e_vdev;
+	int ret;
+
+	i40e_vdev = kzalloc(sizeof(*i40e_vdev), GFP_KERNEL);
+	if (!i40e_vdev)
+		return -ENOMEM;
+
+	i40e_vdev->vdev.match_name = I40E_PEER_RDMA_NAME;
+	i40e_vdev->vdev.dev.parent = &pdev->dev;
+	i40e_vdev->vdev.release = i40e_virtdev_release;
+	i40e_vdev->ldev = ldev;
+	ldev->vdev = &i40e_vdev->vdev;
+
+	ret = virtbus_register_device(&i40e_vdev->vdev);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
 /**
  * i40e_client_add_instance - add a client instance struct to the instance list
  * @pf: pointer to the board struct
@@ -288,9 +326,6 @@ static void i40e_client_add_instance(struct i40e_pf *pf)
 	struct netdev_hw_addr *mac = NULL;
 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
 
-	if (!registered_client || pf->cinst)
-		return;
-
 	cdev = kzalloc(sizeof(*cdev), GFP_KERNEL);
 	if (!cdev)
 		return;
@@ -310,11 +345,8 @@ static void i40e_client_add_instance(struct i40e_pf *pf)
 	cdev->lan_info.fw_build = pf->hw.aq.fw_build;
 	set_bit(__I40E_CLIENT_INSTANCE_NONE, &cdev->state);
 
-	if (i40e_client_get_params(vsi, &cdev->lan_info.params)) {
-		kfree(cdev);
-		cdev = NULL;
-		return;
-	}
+	if (i40e_client_get_params(vsi, &cdev->lan_info.params))
+		goto free_cdev;
 
 	mac = list_first_entry(&cdev->lan_info.netdev->dev_addrs.list,
 			       struct netdev_hw_addr, list);
@@ -326,7 +358,17 @@ static void i40e_client_add_instance(struct i40e_pf *pf)
 	cdev->client = registered_client;
 	pf->cinst = cdev;
 
-	i40e_client_update_msix_info(pf);
+	cdev->lan_info.msix_count = pf->num_iwarp_msix;
+	cdev->lan_info.msix_entries = &pf->msix_entries[pf->iwarp_base_vector];
+
+	if (i40e_init_client_virtdev(&cdev->lan_info))
+		goto free_cdev;
+
+	return;
+
+free_cdev:
+	kfree(cdev);
+	pf->cinst = NULL;
 }
 
 /**
@@ -347,7 +389,7 @@ void i40e_client_del_instance(struct i40e_pf *pf)
  **/
 void i40e_client_subtask(struct i40e_pf *pf)
 {
-	struct i40e_client *client = registered_client;
+	struct i40e_client *client;
 	struct i40e_client_instance *cdev;
 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
 	int ret = 0;
@@ -361,9 +403,11 @@ void i40e_client_subtask(struct i40e_pf *pf)
 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
 		return;
 
-	if (!client || !cdev)
+	if (!cdev || !cdev->client)
 		return;
 
+	client = cdev->client;
+
 	/* Here we handle client opens. If the client is down, and
 	 * the netdev is registered, then open the client.
 	 */
@@ -424,16 +468,8 @@ int i40e_lan_add_device(struct i40e_pf *pf)
 		 pf->hw.pf_id, pf->hw.bus.bus_id,
 		 pf->hw.bus.device, pf->hw.bus.func);
 
-	/* If a client has already been registered, we need to add an instance
-	 * of it to our new LAN device.
-	 */
-	if (registered_client)
-		i40e_client_add_instance(pf);
+	i40e_client_add_instance(pf);
 
-	/* Since in some cases register may have happened before a device gets
-	 * added, we can schedule a subtask to go initiate the clients if
-	 * they can be launched at probe time.
-	 */
 	set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
 	i40e_service_event_schedule(pf);
 
@@ -453,6 +489,8 @@ int i40e_lan_del_device(struct i40e_pf *pf)
 	struct i40e_device *ldev, *tmp;
 	int ret = -ENODEV;
 
+	virtbus_unregister_device(pf->cinst->lan_info.vdev);
+
 	/* First, remove any client instance. */
 	i40e_client_del_instance(pf);
 
@@ -733,6 +771,59 @@ static int i40e_client_update_vsi_ctxt(struct i40e_info *ldev,
 	return err;
 }
 
+static int i40e_client_device_register(struct i40e_info *ldev)
+{
+	struct i40e_client *client;
+	struct i40e_pf *pf;
+
+	if (!ldev) {
+		pr_err("Failed to reg client dev: ldev ptr NULL\n");
+		return -EINVAL;
+	}
+
+	client = ldev->client;
+	pf = ldev->pf;
+	if (!client) {
+		pr_err("Failed to reg client dev: client ptr NULL\n");
+		return -EINVAL;
+	}
+
+	if (!ldev->ops || !client->ops) {
+		pr_err("Failed to reg client dev: client dev peer_ops/ops NULL\n");
+		return -EINVAL;
+	}
+
+	pf->cinst->client = ldev->client;
+	set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
+	i40e_service_event_schedule(pf);
+
+	return 0;
+}
+
+static void i40e_client_device_unregister(struct i40e_info *ldev)
+{
+	struct i40e_pf *pf = ldev->pf;
+	struct i40e_client_instance *cdev = pf->cinst;
+
+	while (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
+		usleep_range(500, 1000);
+
+	if (!cdev || !cdev->client || !cdev->client->ops ||
+	    !cdev->client->ops->close) {
+		dev_err(&pf->pdev->dev, "Cannot close client device\n");
+		return;
+	}
+	cdev->client->ops->close(&cdev->lan_info, cdev->client, false);
+	clear_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state);
+	i40e_client_release_qvlist(&cdev->lan_info);
+	pf->cinst->client = NULL;
+	clear_bit(__I40E_SERVICE_SCHED, pf->state);
+}
+
+/* Retain legacy global registration/unregistration calls till i40iw is
+ * deprecated from the kernel. The irdma unified driver does not use these
+ * exported symbols.
+ */
 /**
  * i40e_register_client - Register a i40e client driver with the L2 driver
  * @client: pointer to the i40e_client struct
diff --git a/include/linux/net/intel/i40e_client.h b/include/linux/net/intel/i40e_client.h
index 72994baf4941..4a83648cf5fd 100644
--- a/include/linux/net/intel/i40e_client.h
+++ b/include/linux/net/intel/i40e_client.h
@@ -4,6 +4,9 @@
 #ifndef _I40E_CLIENT_H_
 #define _I40E_CLIENT_H_
 
+#include <linux/virtual_bus.h>
+
+#define I40E_PEER_RDMA_NAME	"intel,i40e,rdma"
 #define I40E_CLIENT_STR_LENGTH 10
 
 /* Client interface version should be updated anytime there is a change in the
@@ -84,6 +87,7 @@ struct i40e_info {
 	u8 lanmac[6];
 	struct net_device *netdev;
 	struct pci_dev *pcidev;
+	struct virtbus_device *vdev;
 	u8 __iomem *hw_addr;
 	u8 fid;	/* function id, PF id or VF id */
 #define I40E_CLIENT_FTYPE_PF 0
@@ -97,6 +101,7 @@ struct i40e_info {
 	struct i40e_qvlist_info *qvlist_info;
 	struct i40e_params params;
 	struct i40e_ops *ops;
+	struct i40e_client *client;
 
 	u16 msix_count;	 /* number of msix vectors*/
 	/* Array down below will be dynamically allocated based on msix_count */
@@ -107,6 +112,11 @@ struct i40e_info {
 	u32 fw_build;                   /* firmware build number */
 };
 
+struct i40e_virtbus_device {
+	struct virtbus_device vdev;
+	struct i40e_info *ldev;
+};
+
 #define I40E_CLIENT_RESET_LEVEL_PF   1
 #define I40E_CLIENT_RESET_LEVEL_CORE 2
 #define I40E_CLIENT_VSI_FLAG_TCP_ENABLE  BIT(1)
@@ -132,6 +142,11 @@ struct i40e_ops {
 			       struct i40e_client *client,
 			       bool is_vf, u32 vf_id,
 			       u32 flag, u32 valid_flag);
+
+	int (*client_device_register)(struct i40e_info *ldev);
+
+	void (*client_device_unregister)(struct i40e_info *ldev);
+
 };
 
 struct i40e_client_ops {
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2020-05-04 16:43 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-05-04 16:43 [Intel-wired-lan] [net-next v3 1/9] Implementation of Virtual Bus Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 2/9] ice: Create and register virtual bus for RDMA Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 3/9] ice: Complete RDMA peer registration Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 4/9] ice: Support resource allocation requests Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 5/9] ice: Enable event notifications Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 6/9] ice: Allow reset operations Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 7/9] ice: Pass through communications to VF Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 8/9] i40e: Move client header location Jeff Kirsher
2020-05-04 16:43 ` [Intel-wired-lan] [net-next v3 9/9] i40e: Register a virtbus device to provide RDMA Jeff Kirsher

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.