raw/ioat: driver for Intel QuickData Technology
cover raw/ioat: driver for Intel QuickData Technology
This patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Bruce Richardson (8):
raw/ioat: add initial support for ioat rawdev driver usertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
MAINTAINERS | 7 +- app/test/Makefile | 1 + app/test/meson.build | 4 + app/test/test_ioat_rawdev.c | 269 +++++++++++++++++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 227 ++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 + drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 29 ++ drivers/raw/ioat/ioat_rawdev.c | 310 ++++++++++++++++++++ drivers/raw/ioat/meson.build | 9 + drivers/raw/ioat/rte_ioat_rawdev.h | 228 ++++++++++++++ drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 3 +- mk/rte.app.mk | 1 + usertools/dpdk-devbind.py | 10 + 21 files changed, 1422 insertions(+), 2 deletions(-) create mode 100644 app/test/test_ioat_rawdev.c create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_ioat_spec.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
Signature · 1 line
2.21.0
PATCH 1/8 raw/ioat: add initial support for ioat rawdev driver
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
--- MAINTAINERS | 7 +- app/test/Makefile | 1 + app/test/meson.build | 1 + app/test/test_ioat_rawdev.c | 22 +++++ config/common_armv8a_linux | 1 + config/common_base | 5 ++ config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 25 ++++++ doc/guides/rel_notes/release_19_08.rst | 11 +++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 ++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 3 +- mk/rte.app.mk | 1 + 19 files changed, 236 insertions(+), 2 deletions(-) create mode 100644 app/test/test_ioat_rawdev.c create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +6 -1
@@ -1042,6 +1042,12 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst +F: app/test/test_ioat_rawdev.c + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/
@@ -1052,7 +1058,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processing -----------------
app/test/Makefile +1 -0
@@ -212,6 +212,7 @@ endif ifeq ($(CONFIG_RTE_LIBRTE_RAWDEV),y) SRCS-y += test_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += test_ioat_rawdev.c endif SRCS-$(CONFIG_RTE_LIBRTE_KVARGS) += test_kvargs.c
app/test/meson.build +1 -0
@@ -52,6 +52,7 @@ test_sources = files('commands.c', 'test_hash_perf.c', 'test_hash_readwrite_lf.c', 'test_interrupts.c', + 'test_ioat_rawdev.c', 'test_ipsec.c', 'test_kni.c', 'test_kvargs.c',
app/test/test_ioat_rawdev.c new file +22 -0
@@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#include "test.h" + +#ifndef RTE_LIBRTE_PMD_IOAT_RAWDEV + +static int +test_ioat_rawdev (void ) {return TEST_SKIPPED ; } + +#else + +static int +test_ioat_rawdev (void ) +{ +return 0 ; +} + +#endif /* RTE_LIBRTE_PMD_IOAT_RAWDEV */ + +REGISTER_TEST_COMMAND (ioat_rawdev_autotest, test_ioat_rawdev);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=n
config/common_base +5 -0
@@ -741,6 +741,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n # CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y +# +# Compile PMD for Intel IOAT raw device +# +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +1 -0
@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +25 -0
@@ -0,0 +1,25 @@ +.. SPDX-License-Identifier: BSD-3-Clause + Copyright(c) 2019 Intel Corporation. + +.. include:: <isonum.txt> + +IOAT Rawdev Driver for Intel\ |reg| QuickData Technology +====================================================================== + +The ``ioat`` rawdev driver provides a poll-mode driver (PMD) for Intel\ |reg| +QuickData Technology, part of Intel\ |reg| I/O Acceleration Technology +`(Intel I/OAT) <https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_. +This PMD allows data copies, for example, cloning packet data, to be +accelerated by hardware rather than having to be done by software, freeing +up CPU cycles for other tasks. + +Compilation +------------ + +For builds done with ``make``, the driver compilation is enabled by the +``CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV`` build configuration option. This is +enabled by default in builds for x86 platforms, and disabled in other +configurations. + +For builds using ``meson`` and ``ninja``, the driver will be built when the +target platform is x86-based.
doc/guides/rel_notes/release_19_08.rst +11 -0
@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================
@@ -54,6 +56,15 @@ New Features Also, make sure to start the actual text at the margin. ========================================================= +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of + Intel\ |reg| I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_, + allows data copies to be done by hardware instead + of via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items -------------
drivers/raw/Makefile +1 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright(c) 2019 Intel Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +93 -0
93 lines not on this page · raw message
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright 2019 Intel Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior notice +*/ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() */ +#define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ +#define IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +2 -1
@@ -1,7 +1,8 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +1 -0
@@ -303,6 +303,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev _LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke endif # CONFIG_RTE_LIBRTE_IFPGA_BUS +_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev endif # CONFIG_RTE_LIBRTE_RAWDEV endif # !CONFIG_RTE_BUILD_SHARED_LIBS
Signature · 1 line
2.21.0
PATCH 2/8 usertools/dpdk-devbind.py: add support for IOAT devices
In reply to Bruce RichardsonIn order to allow binding/unbinding of devices for use by the ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++++++++++ usertools/dpdk-devbind.py | 10 ++++++++++ 2 files changed, 21 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -23,3 +23,14 @@ configurations. For builds using ``meson`` and ``ninja``, the driver will be built when the target platform is x86-based. + +Device Setup +------------- + +The Intel\ |reg| QuickData Technology HW devices will need to be bound to a +user-space IO driver for use. The script ``dpdk-devbind.py`` script +included with DPDK can be used to view the state of the devices and to bind +them to a suitable DPDK-supported kernel driver. When querying the +status of the devices, they will appear under the category of "dma +devices", i.e. the command ``dpdk-devbind.py --status-dev dma`` can be used +to see the state of those devices alone.
usertools/dpdk-devbind.py +10 -0
@@ -36,11 +36,17 @@ octeontx2_npa = {'Class': '08', 'Vendor': '177d', 'Device': 'a0fb,a0fc', 'SVendor': None, 'SDevice': None} +intel_ioat_bdw = {'Class': '08', 'Vendor': '8086', 'Device': '6f20,6f21,6f22,6f23,6f24,6f25,6f26,6f27,6f2e,6f2f', + 'SVendor': None, 'SDevice': None} +intel_ioat_skx = {'Class': '08', 'Vendor': '8086', 'Device': '2021', + 'SVendor': None, 'SDevice': None} + network_devices = [network_class, cavium_pkx, avp_vnic, ifpga_class] crypto_devices = [encryption_class, intel_processor_class] eventdev_devices = [cavium_sso, cavium_tim, octeontx2_sso] mempool_devices = [cavium_fpa, octeontx2_npa] compress_devices = [cavium_zip] +dma_devices = [intel_ioat_bdw, intel_ioat_skx] # global dict ethernet devices present. Dictionary indexed by PCI address. # Each device within this is itself a dictionary of device properties
@@ -595,6 +601,8 @@ def show_status(): if status_dev == "compress" or status_dev == "all": show_device_status(compress_devices , "Compress") + if status_dev == "dma" or status_dev == "all": + show_device_status(dma_devices, "DMA") def parse_args(): '''Parses the command-line arguments given by the user and takes the
@@ -670,6 +678,7 @@ def do_arg_actions(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(dma_devices) show_status()
@@ -690,6 +699,7 @@ def main(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(dma_devices) do_arg_actions() if __name__ == "__main__":
Signature · 1 line
2.21.0
PATCH 3/8 raw/ioat: add register definition file
In reply to Bruce RichardsonAdd in the list of registers for the device. File is taken from the SPDK project:
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
--- drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++++++++++++++ 3 files changed, 304 insertions(+), 1 deletion(-) create mode 100644 drivers/raw/ioat/rte_ioat_spec.h
drivers/raw/ioat/Makefile +1 -0
@@ -24,5 +24,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h +SYMLINK-y-include += rte_ioat_spec.h include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/meson.build +2 -1
@@ -5,4 +5,5 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c') deps += ['rawdev', 'bus_pci'] -install_headers('rte_ioat_rawdev.h') +install_headers('rte_ioat_rawdev.h', + 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_spec.h new file +301 -0
301 lines not on this page · raw message
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH 3/8] raw/ioat: add register definition file
drivers/raw/ioat/rte_ioat_spec.h line 19
Quoted 5 lines · On Thu, 30 May 2019 22:25:20 +0100 Bruce Richardson…
drivers/raw/ioat/rte_ioat_spec.h new file +5
-0+ +#ifdef __cplusplus +extern "C" { +#endif +
This a driver private file, why the C++ guard here?
Re: [dpdk-dev] [PATCH 3/8] raw/ioat: add register definition file
Quoted 10 lines · Stephen Hemminger
drivers/raw/ioat/rte_ioat_spec.h line 19
Quoted 5 lines · On Thu, 30 May 2019 22:25:20 +0100 Bruce Richardson…
drivers/raw/ioat/rte_ioat_spec.h new file +5
-0+ +#ifdef __cplusplus +extern "C" { +#endif +This a driver private file, why the C++ guard here?
It's not just a private file, it's included in the regular driver .h file, as the inline functions there need access to the register definitions.
As for whether the functions need to be inline, most of them are inline for performance reasons, and the one that I would consider suitable for moving to the driver is inline also so that the user does not need to link their application against the PMD in order to use the device. Having the header at compile-time, and then the driver loaded at runtime is sufficient.
/Bruce
PATCH 4/8 raw/ioat: create device on probe and destroy on release
In reply to Bruce RichardsonAdd the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 ++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++++++++++- drivers/raw/ioat/rte_ioat_rawdev.h | 20 +++++++ 3 files changed, 121 insertions(+), 3 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -34,3 +34,14 @@ them to a suitable DPDK-supported kernel driver. When querying the status of the devices, they will appear under the category of "dma devices", i.e. the command ``dpdk-devbind.py --status-dev dma`` can be used to see the state of those devices alone. + +Device Probing and Initialization +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Once bound to a suitable kernel device driver, the HW devices will be found +as part of the PCI scan done at application initialization time. No vdev +parameters need to be passed to create or initialize the device. + +Once probed successfully, the device will appear as a ``rawdev``, that is a +"raw device type" inside DPDK, and can be accessed using APIs from the +``rte_rawdev`` library.
drivers/raw/ioat/ioat_rawdev.c +90 -3
111 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +20 -0
@@ -14,6 +14,9 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <rte_memory.h> +#include <rte_ioat_spec.h> +/** Name of the device driver */ #define IOAT_PMD_RAWDEV_NAME rawdev_ioat/** String reported as the device driver name by rte_rawdev_info_get() */
@@ -21,4 +24,21 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* @internal +* Structure representing a device instance +*/ +struct rte_ioat_rawdev { +struct rte_rawdev *rawdev; +volatile struct rte_ioat_registers *regs; +phys_addr_t status_addr; +phys_addr_t ring_addr; + +unsigned short ring_size; +struct rte_ioat_desc *desc_ring; + +/* to report completions, the device will write status back here */ + volatile uint64_t status __rte_cache_aligned; +}; +#endif
Signature · 1 line
2.21.0
PATCH 5/8 raw/ioat: add device info function
In reply to Bruce RichardsonAdd in the "info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.
--- app/test/meson.build | 3 +++ app/test/test_ioat_rawdev.c | 23 ++++++++++++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 5 files changed, 82 insertions(+)
app/test/meson.build +3 -0
@@ -305,6 +305,9 @@ endif if dpdk_conf.has('RTE_LIBRTE_KNI') test_deps += 'kni' endif +if dpdk_conf.has('RTE_LIBRTE_PMD_IOAT_RAWDEV') + test_deps += 'pmd_ioat' +endif cflags = machine_args if cc.has_argument('-Wno-format-truncation')
app/test/test_ioat_rawdev.c +23 -0
@@ -11,9 +11,32 @@ test_ioat_rawdev(void) { return TEST_SKIPPED; }#else +#include <string.h> +#include <unistd.h> + +#include <rte_mbuf.h> +#include <rte_rawdev.h> +#include <rte_ioat_rawdev.h> +static int test_ioat_rawdev (void ) { +const int count =rte_rawdev_count (); +int i, found =0 ; + +printf ("Checking %d rawdevs\n" , count); +for (i =0 ; i < count && !found; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; + found = (rte_rawdev_info_get (i, &info) ==0 && +strcmp (info.driver_name, +IOAT_PMD_RAWDEV_NAME_STR ) ==0 ); + } + +if (!found) { +printf ("No IOAT rawdev found, skipping tests\n" ); +return TEST_SKIPPED ; + } +return 0 ; }
doc/guides/rawdevs/ioat_rawdev.rst +34 -0
@@ -45,3 +45,37 @@ parameters need to be passed to create or initialize the device. Once probed successfully, the device will appear as a ``rawdev``, that is a "raw device type" inside DPDK, and can be accessed using APIs from the ``rte_rawdev`` library. + +Using IOAT Rawdev Devices +-------------------------- + +To use the devices from an application, the rawdev API can be used, along +with definitions taken from the device-specific header file +``rte_ioat_rawdev.h``. This header is needed to get the definition of +structure parameters used by some of the rawdev APIs for IOAT rawdev +devices, as well as providing key functions for using the device for memory +copies. + +Getting Device Information +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Basic information about each rawdev device can be got using the +``rte_rawdev_info_get()`` API. For most applications, this API will be +needed to verify that the rawdev in question is of the expected type. For +example, the following code in ``test_ioat_rawdev.c`` is used to identify +the IOAT rawdev device for use for the tests: + +.. code-block:: C + + for (i = 0; i < count && !found; i++) { + struct rte_rawdev_info info = { .dev_private = NULL }; + found = (rte_rawdev_info_get(i, &info) == 0 && + strcmp(info.driver_name, + IOAT_PMD_RAWDEV_NAME_STR) == 0); + } + +When calling the ``rte_rawdev_info_get()`` API for an IOAT rawdev device, +the ``dev_private`` field in the ``rte_rawdev_info`` struct should either +be NULL, or else be set to point to a structure of type +``rte_ioat_rawdev_config``, in which case the size of the configured device +input ring will be returned in that structure.
drivers/raw/ioat/ioat_rawdev.c +11 -0
@@ -24,10 +24,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +} +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_info_get = ioat_dev_info_get, };struct rte_rawdev *rawdev =NULL ;
drivers/raw/ioat/rte_ioat_rawdev.h +11 -0
@@ -24,6 +24,17 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* Configuration structure for an ioat rawdev instance +* +* This structure is to be passed as the ".dev_private" parameter when +* calling the rte_rawdev_get_info() and rte_rawdev_configure() APIs on +* an ioat rawdev instance. +*/ +struct rte_ioat_rawdev_config { +unsigned short ring_size; +}; +/** * @internal * Structure representing a device instance
Signature · 1 line
2.21.0
PATCH 6/8 raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonAllow initializing a driver instance.
--- app/test/test_ioat_rawdev.c | 35 +++++++++++++- doc/guides/rawdevs/ioat_rawdev.rst | 32 +++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 75 ++++++++++++++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 14 ++++++ 4 files changed, 155 insertions(+), 1 deletion(-)
app/test/test_ioat_rawdev.c +34 -1
@@ -18,6 +18,39 @@ test_ioat_rawdev(void) { return TEST_SKIPPED; }#include <rte_rawdev.h> #include <rte_ioat_rawdev.h> +static int +run_ioat_tests (int dev_id) +{ +#define IOAT_TEST_RINGSIZE 512 +struct rte_ioat_rawdev_config p = { .ring_size =-1 }; +struct rte_rawdev_info info = { .dev_private = &p }; + +rte_rawdev_info_get (dev_id, &info); +if (p.ring_size !=0 ) { +printf ("Error, initial ring size is non-zero (%d)\n" , + (int )p.ring_size); +return -1 ; + } + + p.ring_size =IOAT_TEST_RINGSIZE ; +if (rte_rawdev_configure (dev_id, &info) !=0 ) { +printf ("Error with rte_rawdev_configure()\n" ); +return -1 ; + } +rte_rawdev_info_get (dev_id, &info); +if (p.ring_size !=IOAT_TEST_RINGSIZE ) { +printf ("Error, ring size is not %d (%d)\n" , +IOAT_TEST_RINGSIZE , (int )p.ring_size); +return -1 ; + } + +if (rte_rawdev_start (dev_id) !=0 ) { +printf ("Error with rte_rawdev_start()\n" ); +return -1 ; + } +return 0 ; +} +static int test_ioat_rawdev (void ) {
@@ -37,7 +70,7 @@ test_ioat_rawdev(void) return TEST_SKIPPED; } -return 0 ; +return run_ioat_tests (i); }#endif /* RTE_LIBRTE_PMD_IOAT_RAWDEV */
doc/guides/rawdevs/ioat_rawdev.rst +32 -0
@@ -79,3 +79,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API.
drivers/raw/ioat/ioat_rawdev.c +75 -0
87 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +14 -0
@@ -47,9 +47,23 @@ struct rte_ioat_rawdev {unsigned short ring_size;struct rte_ioat_desc *desc_ring; +__m128i *hdls;/* completion handles for returning to user */ /* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; }; +/** +* @internal +* Structure representing a descriptor for a copy operation +*/ +struct rte_ioat_desc { + uint32_t xfer_size __rte_cache_aligned; +uint32_t desc_control; +phys_addr_t src_addr;/* 64 bits */ +phys_addr_t dest_addr;/* 64 bits */ +phys_addr_t next_desc_addr;/* 64 bits */ +uint64_t op_type_specific[4 ]; +}; +#endif
Signature · 1 line
2.21.0
PATCH 7/8 raw/ioat: add statistics functions
In reply to Bruce RichardsonAdd stats functions to track what is happening in the driver, and put unit tests to check those.
--- app/test/test_ioat_rawdev.c | 38 ++++++++++++++++++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 14 ++++++++++ drivers/raw/ioat/ioat_rawdev.c | 44 ++++++++++++++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 6 ++++ 4 files changed, 102 insertions(+)
app/test/test_ioat_rawdev.c +38 -0
@@ -24,6 +24,11 @@ run_ioat_tests(int dev_id)#define IOAT_TEST_RINGSIZE 512struct rte_ioat_rawdev_config p = { .ring_size =-1 };struct rte_rawdev_info info = { .dev_private = &p }; +struct rte_rawdev_xstats_name *snames =NULL ; +uint64_t *stats =NULL ; +unsigned int *ids =NULL ; +unsigned int nb_xstats; +unsigned int i;rte_rawdev_info_get (dev_id, &info);if (p.ring_size !=0 ) {
@@ -48,6 +53,39 @@ run_ioat_tests(int dev_id)printf ("Error with rte_rawdev_start()\n" );return -1 ; } + +/* allocate memory for xstats names and values */ + nb_xstats =rte_rawdev_xstats_names_get (dev_id,NULL ,0 ); + + snames =malloc (sizeof (*snames) * nb_xstats); +if (snames ==NULL ) { +printf ("Error allocating xstat names memory\n" ); +return -1 ; + } +rte_rawdev_xstats_names_get (dev_id, snames, nb_xstats); + + ids =malloc (sizeof (*ids) * nb_xstats); +if (ids ==NULL ) { +printf ("Error allocating xstat ids memory\n" ); +return -1 ; + } +for (i =0 ; i < nb_xstats; i++) + ids[i] = i; + + stats =malloc (sizeof (*stats) * nb_xstats); +if (stats ==NULL ) { +printf ("Error allocating xstat memory\n" ); +return -1 ; + } + +rte_rawdev_xstats_get (dev_id, ids, stats, nb_xstats); +for (i =0 ; i < nb_xstats; i++) +printf ("%s: %" PRIu64" " , snames[i].name, stats[i]); +printf ("\n" ); + +free (snames); +free (stats); +free (ids);return 0 ; }
doc/guides/rawdevs/ioat_rawdev.rst +14 -0
@@ -111,3 +111,17 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. + +Querying Device Statistics +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The statistics from the IOAT rawdev device can be got via the xstats +functions in the ``rte_rawdev`` library, i.e. +``rte_rawdev_xstats_names_get()``, ``rte_rawdev_xstats_get()`` and +``rte_rawdev_xstats_by_name_get``. The statistics returned for each device +instance are: + +* ``failed_enqueues`` +* ``successful_enqueues`` +* ``copies_started`` +* ``copies_completed``
drivers/raw/ioat/ioat_rawdev.c +44 -0
62 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +6 -0
@@ -49,6 +49,12 @@ struct rte_ioat_rawdev {struct rte_ioat_desc *desc_ring;__m128i *hdls;/* completion handles for returning to user */ +/* some statistics for tracking, if added/changed update xstats fns*/ + uint64_t enqueue_failed __rte_cache_aligned; +uint64_t enqueued; +uint64_t started; +uint64_t completed; +/* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; };
Signature · 1 line
2.21.0
PATCH 8/8 raw/ioat: add local API to perform copies
In reply to Bruce RichardsonAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
--- app/test/test_ioat_rawdev.c | 159 ++++++++++++++++++++++++++++- doc/guides/rawdevs/ioat_rawdev.rst | 100 ++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 155 +++++++++++++++++++++++++++- 3 files changed, 410 insertions(+), 4 deletions(-)
app/test/test_ioat_rawdev.c +156 -3
@@ -18,6 +18,131 @@ test_ioat_rawdev(void) { return TEST_SKIPPED; }#include <rte_rawdev.h> #include <rte_ioat_rawdev.h> +static struct rte_mempool *pool; + +static int +test_enqueue_copies (int dev_id) +{ +const unsigned int length =1024 ; +unsigned int i; + +do { +struct rte_mbuf *src, *dst; +char *src_data, *dst_data; +struct rte_mbuf *completed[2 ] = {0 }; + +/* test doing a single copy */ + src =rte_pktmbuf_alloc (pool); + dst =rte_pktmbuf_alloc (pool); + src->data_len = src->pkt_len = length; + dst->data_len = dst->pkt_len = length; + src_data = rte_pktmbuf_mtod(src, char *); + dst_data = rte_pktmbuf_mtod(dst, char *); + +for (i =0 ; i < length; i++) + src_data[i] =rand () &0xFF ; + +if (rte_ioat_enqueue_copy (dev_id, + src->buf_iova + src->data_off, + dst->buf_iova + dst->data_off, + length, + (uintptr_t )src, + (uintptr_t )dst, +0 /* no fence */ ) !=1 ) { +printf ("Error with rte_ioat_enqueue_copy\n" ); +return -1 ; + } +rte_ioat_do_copies (dev_id); +usleep (10 ); + +if (rte_ioat_completed_copies (dev_id,1 , (void *)&completed[0 ], + (void *)&completed[1 ]) !=1 ) { +printf ("Error with rte_ioat_completed_copies\n" ); +return -1 ; + } +if (completed[0 ] != src || completed[1 ] != dst) { +printf ("Error with completions: got (%p, %p), not (%p,%p)\n" , + completed[0 ], completed[1 ], src, dst); +return -1 ; + } + +for (i =0 ; i < length; i++) +if (dst_data[i] != src_data[i]) { +printf ("Data mismatch at char %u\n" , i); +return -1 ; + } +rte_pktmbuf_free (src); +rte_pktmbuf_free (dst); + }while (0 ); + +/* test doing multiple copies */ +do { +struct rte_mbuf *srcs[32 ], *dsts[32 ]; +struct rte_mbuf *completed_src[64 ]; +struct rte_mbuf *completed_dst[64 ]; +unsigned int j; + +for (i =0 ; i <RTE_DIM (srcs); i++) { +char *src_data; + + srcs[i] =rte_pktmbuf_alloc (pool); + dsts[i] =rte_pktmbuf_alloc (pool); + srcs[i]->data_len = srcs[i]->pkt_len = length; + dsts[i]->data_len = dsts[i]->pkt_len = length; + src_data = rte_pktmbuf_mtod(srcs[i], char *); + +for (j =0 ; j < length; j++) + src_data[j] =rand () &0xFF ; + +if (rte_ioat_enqueue_copy (dev_id, + srcs[i]->buf_iova + srcs[i]->data_off, + dsts[i]->buf_iova + dsts[i]->data_off, + length, + (uintptr_t )srcs[i], + (uintptr_t )dsts[i], +0 /* nofence */ ) !=1 ) { +printf ("Error with rte_ioat_enqueue_copy for buffer %u\n" , + i); +return -1 ; + } + } +rte_ioat_do_copies (dev_id); +usleep (100 ); + +if (rte_ioat_completed_copies (dev_id,64 , (void *)completed_src, + (void *)completed_dst) !=RTE_DIM (srcs)) { +printf ("Error with rte_ioat_completed_copies\n" ); +return -1 ; + } +for (i =0 ; i <RTE_DIM (srcs); i++) { +char *src_data, *dst_data; + +if (completed_src[i] != srcs[i]) { +printf ("Error with source pointer %u\n" , i); +return -1 ; + } +if (completed_dst[i] != dsts[i]) { +printf ("Error with dest pointer %u\n" , i); +return -1 ; + } + + src_data = rte_pktmbuf_mtod(srcs[i], char *); + dst_data = rte_pktmbuf_mtod(dsts[i], char *); +for (j =0 ; j < length; j++) +if (src_data[j] != dst_data[j]) { +printf ("Error with copy of packet %u, byte %u\n" , + i, j); +return -1 ; + } +rte_pktmbuf_free (srcs[i]); +rte_pktmbuf_free (dsts[i]); + } + + }while (0 ); + +return 0 ; +} +static int run_ioat_tests (int dev_id) {
@@ -54,6 +179,17 @@ run_ioat_tests(int dev_id)return -1 ; } + pool =rte_pktmbuf_pool_create ("TEST_IOAT_POOL" , +256 ,/* n == num elements */ +32 ,/* cache size */ +0 ,/* priv size */ +2048 ,/* data room size */ + info.socket_id); +if (pool ==NULL ) { +printf ("Error with mempool creation\n" ); +return -1 ; + } +/* allocate memory for xstats names and values */ nb_xstats =rte_rawdev_xstats_names_get (dev_id,NULL ,0 );
@@ -78,15 +214,32 @@ run_ioat_tests(int dev_id)return -1 ; } -rte_rawdev_xstats_get (dev_id, ids, stats, nb_xstats); -for (i =0 ; i < nb_xstats; i++) -printf ("%s: %" PRIu64" " , snames[i].name, stats[i]); +/* run the test cases */ +for (i =0 ; i <100 ; i++) { +unsigned int j; + +if (test_enqueue_copies (dev_id) !=0 ) + goto err; + +rte_rawdev_xstats_get (dev_id, ids, stats, nb_xstats); +for (j =0 ; j < nb_xstats; j++) +printf ("%s: %" PRIu64" " , snames[j].name, stats[j]); +printf ("\r" ); + }printf ("\n" ); +rte_mempool_free (pool);free (snames);free (stats);free (ids);return 0 ; + +err: +rte_mempool_free (pool); +free (snames); +free (stats); +free (ids); +return -1 ; } static int
doc/guides/rawdevs/ioat_rawdev.rst +100 -0
106 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +154 -1
171 lines not on this page · raw message
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH 8/8] raw/ioat: add local API to perform copies
30/05/2019 23:25, Bruce Richardson:
Quoted 3 lines
Add local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
Would be nice to have a small design overview in the guide.
We may need also to answer these questions:
What are the performance gains? In which use-case? What are the hardware requirements?
Re: [dpdk-dev] [PATCH 8/8] raw/ioat: add local API to perform copies
Quoted 11 lines · Thomas Monjalon
30/05/2019 23:25, Bruce Richardson:
Quoted 3 lines
Add local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
Would be nice to have a small design overview in the guide.
We may need also to answer these questions:
What are the performance gains? In which use-case? What are the hardware requirements?
Thanks for the comments. Will try and add more details in subsequent revisions.
/Bruce
Re: [dpdk-dev] [PATCH 8/8] raw/ioat: add local API to perform copies
In reply to Bruce RichardsonQuoted 8 lines
-----Original Message-----
From: dev <[email protected]> On Behalf Of Bruce Richardson
Sent: Friday, May 31, 2019 2:55 AM
To: [email protected]
Cc: Bruce Richardson <[email protected]>
Subject: [dpdk-dev] [PATCH 8/8] raw/ioat: add local API to perform copies
Quoted 7 lines
Add local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
Signed-off-by Bruce Richardson <[email protected]>--- app/test/test_ioat_rawdev.c | 159 ++++++++++++++++++++++++++++-
How about moving this test to drivers/raw/ioat/ directory by implementing dev_selftest ops? IMO, it is better to not pollute app/test with PMD specific test cases.
Re: [dpdk-dev] [PATCH 8/8] raw/ioat: add local API to perform copies
Quoted 12 lines · Jerin Jacob Kollanukkaran
Quoted 12 lines
-----Original Message-----
From: dev <[email protected]> On Behalf Of Bruce Richardson
Sent: Friday, May 31, 2019 2:55 AM
To: [email protected]
Cc: Bruce Richardson <[email protected]>
Subject: [dpdk-dev] [PATCH 8/8] raw/ioat: add local API to perform copiesAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
Quoted 8 lines
Quoted 3 lines
Signed-off-by Bruce Richardson <[email protected]>--- app/test/test_ioat_rawdev.c | 159 ++++++++++++++++++++++++++++-How about moving this test to
drivers/raw/ioat/directory by implementingdev_selftestops? IMO, it is better to not pollute app/test with PMD specific test cases.
Let me look into it.
/Bruce
v2 cover raw/ioat: driver for Intel QuickData Technology
In reply to Bruce RichardsonThis patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
V2:
* moved tests to rawdev selftest function
* some checkpatch and other small cleanups
* added extra documentation details on supported hardware
* aligned the changes to dpdk-devbind with the changes in the NTB set for consistency
Bruce Richardson (8):
raw/ioat: add initial support for ioat rawdev driver usertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
MAINTAINERS | 6 +- app/test/test_rawdev.c | 19 ++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 265 +++++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 + drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 31 ++ drivers/raw/ioat/ioat_rawdev.c | 313 ++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 235 +++++++++++++++ drivers/raw/ioat/meson.build | 10 + drivers/raw/ioat/rte_ioat_rawdev.h | 242 +++++++++++++++ drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + usertools/dpdk-devbind.py | 10 + 20 files changed, 1460 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_ioat_spec.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
Signature · 1 line
2.21.0
PATCH v2 1/8 raw/ioat: add initial support for ioat rawdev driver
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
---
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
--- MAINTAINERS | 6 +- app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 ++ config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 ++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 +++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 ++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 258 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +5 -1
@@ -1047,6 +1047,11 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/
@@ -1057,7 +1062,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processing -----------------
app/test/test_rawdev.c +8 -0
@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }REGISTER_TEST_COMMAND (rawdev_autotest, test_rawdev_selftest_skeleton); + +static int +test_rawdev_selftest_ioat (void ) +{ +return TEST_SKIPPED ; +} + +REGISTER_TEST_COMMAND (ioat_rawdev_autotest, test_rawdev_selftest_ioat);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=n
config/common_base +5 -0
@@ -746,6 +746,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n # CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y +# +# Compile PMD for Intel IOAT raw device +# +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +1 -0
@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +63 -0
63 lines not on this page · raw message
doc/guides/rel_notes/release_19_08.rst +11 -0
@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================
@@ -88,6 +90,15 @@ New Features * Added multi-queue support to allow one af_xdp vdev with multiple netdev queues +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of + Intel\ |reg| I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_, + allows data copies to be done by hardware instead + of via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items -------------
drivers/raw/Makefile +1 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright(c) 2019 Intel Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +93 -0
93 lines not on this page · raw message
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright 2019 Intel Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior notice +*/ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() */ +#define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ +#define IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +3 -1
@@ -1,7 +1,9 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat', + 'skeleton_rawdev'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +1 -0
@@ -301,6 +301,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev _LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke endif # CONFIG_RTE_LIBRTE_IFPGA_BUS +_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev endif # CONFIG_RTE_LIBRTE_RAWDEV endif # !CONFIG_RTE_BUILD_SHARED_LIBS
Signature · 1 line
2.21.0
PATCH v2 2/8 usertools/dpdk-devbind.py: add support for IOAT devices
In reply to Bruce RichardsonIn order to allow binding/unbinding of devices for use by the ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.
---
V2: rather than adding a DMA section, add to "misc (rawdev)" section where
other device types, e.g. ntb can also do.
NOTE: this set largely overlaps with the equivalent changes made by the patchset adding NTB support [1]. Since it's unclear which set will be added first, this set is based off the latest head.
[1] http://patches.dpdk.org/patch/55127/
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++++++++++ usertools/dpdk-devbind.py | 10 ++++++++++ 2 files changed, 21 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -61,3 +61,14 @@ configurations. For builds using ``meson`` and ``ninja``, the driver will be built when the target platform is x86-based. + +Device Setup +------------- + +The Intel\ |reg| QuickData Technology HW devices will need to be bound to a +user-space IO driver for use. The script ``dpdk-devbind.py`` script +included with DPDK can be used to view the state of the devices and to bind +them to a suitable DPDK-supported kernel driver. When querying the status +of the devices, they will appear under the category of "Misc (rawdev) +devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be +used to see the state of those devices alone.
usertools/dpdk-devbind.py +10 -0
@@ -36,11 +36,17 @@ octeontx2_npa = {'Class': '08', 'Vendor': '177d', 'Device': 'a0fb,a0fc', 'SVendor': None, 'SDevice': None} +intel_ioat_bdw = {'Class': '08', 'Vendor': '8086', 'Device': '6f20,6f21,6f22,6f23,6f24,6f25,6f26,6f27,6f2e,6f2f', + 'SVendor': None, 'SDevice': None} +intel_ioat_skx = {'Class': '08', 'Vendor': '8086', 'Device': '2021', + 'SVendor': None, 'SDevice': None} + network_devices = [network_class, cavium_pkx, avp_vnic, ifpga_class] crypto_devices = [encryption_class, intel_processor_class] eventdev_devices = [cavium_sso, cavium_tim, octeontx2_sso] mempool_devices = [cavium_fpa, octeontx2_npa] compress_devices = [cavium_zip] +misc_devices = [intel_ioat_bdw, intel_ioat_skx] # global dict ethernet devices present. Dictionary indexed by PCI address. # Each device within this is itself a dictionary of device properties
@@ -595,6 +601,8 @@ def show_status(): if status_dev == "compress" or status_dev == "all": show_device_status(compress_devices , "Compress") + if status_dev == "misc" or status_dev == "all": + show_device_status(misc_devices, "Misc (rawdev)") def parse_args(): '''Parses the command-line arguments given by the user and takes the
@@ -670,6 +678,7 @@ def do_arg_actions(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) show_status()
@@ -690,6 +699,7 @@ def main(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) do_arg_actions() if __name__ == "__main__":
Signature · 1 line
2.21.0
PATCH v2 3/8 raw/ioat: add register definition file
In reply to Bruce RichardsonAdd in the list of registers for the device. File is taken from the SPDK project:
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
--- drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++++++++++++++ 3 files changed, 304 insertions(+), 1 deletion(-) create mode 100644 drivers/raw/ioat/rte_ioat_spec.h
drivers/raw/ioat/Makefile +1 -0
@@ -24,5 +24,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h +SYMLINK-y-include += rte_ioat_spec.h include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/meson.build +2 -1
@@ -5,4 +5,5 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c') deps += ['rawdev', 'bus_pci'] -install_headers('rte_ioat_rawdev.h') +install_headers('rte_ioat_rawdev.h', + 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_spec.h new file +301 -0
301 lines not on this page · raw message
Signature · 1 line
2.21.0
PATCH v2 4/8 raw/ioat: create device on probe and destroy on release
In reply to Bruce RichardsonAdd the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 ++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++++++++++- drivers/raw/ioat/rte_ioat_rawdev.h | 20 +++++++ 3 files changed, 121 insertions(+), 3 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -72,3 +72,14 @@ them to a suitable DPDK-supported kernel driver. When querying the status of the devices, they will appear under the category of "Misc (rawdev) devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be used to see the state of those devices alone. + +Device Probing and Initialization +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Once bound to a suitable kernel device driver, the HW devices will be found +as part of the PCI scan done at application initialization time. No vdev +parameters need to be passed to create or initialize the device. + +Once probed successfully, the device will appear as a ``rawdev``, that is a +"raw device type" inside DPDK, and can be accessed using APIs from the +``rte_rawdev`` library.
drivers/raw/ioat/ioat_rawdev.c +90 -3
111 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +20 -0
@@ -14,6 +14,9 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <rte_memory.h> +#include <rte_ioat_spec.h> +/** Name of the device driver */ #define IOAT_PMD_RAWDEV_NAME rawdev_ioat/** String reported as the device driver name by rte_rawdev_info_get() */
@@ -21,4 +24,21 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* @internal +* Structure representing a device instance +*/ +struct rte_ioat_rawdev { +struct rte_rawdev *rawdev; +volatile struct rte_ioat_registers *regs; +phys_addr_t status_addr; +phys_addr_t ring_addr; + +unsigned short ring_size; +struct rte_ioat_desc *desc_ring; + +/* to report completions, the device will write status back here */ + volatile uint64_t status __rte_cache_aligned; +}; +#endif
Signature · 1 line
2.21.0
PATCH v2 5/8 raw/ioat: add device info function
In reply to Bruce RichardsonAdd in the "info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.
---
V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)
app/test/test_rawdev.c +11 -0
@@ -29,6 +29,17 @@ REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton);static int test_rawdev_selftest_ioat (void ) { +const int count =rte_rawdev_count (); +int i; + +for (i =0 ; i < count; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; +if (rte_rawdev_info_get (i, &info) ==0 && +strstr (info.driver_name,"ioat" ) !=NULL ) +return 0 ; + } + +printf ("No IOAT rawdev found, skipping tests\n" );return TEST_SKIPPED ; }
doc/guides/rawdevs/ioat_rawdev.rst +34 -0
@@ -83,3 +83,37 @@ parameters need to be passed to create or initialize the device. Once probed successfully, the device will appear as a ``rawdev``, that is a "raw device type" inside DPDK, and can be accessed using APIs from the ``rte_rawdev`` library. + +Using IOAT Rawdev Devices +-------------------------- + +To use the devices from an application, the rawdev API can be used, along +with definitions taken from the device-specific header file +``rte_ioat_rawdev.h``. This header is needed to get the definition of +structure parameters used by some of the rawdev APIs for IOAT rawdev +devices, as well as providing key functions for using the device for memory +copies. + +Getting Device Information +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Basic information about each rawdev device can be got using the +``rte_rawdev_info_get()`` API. For most applications, this API will be +needed to verify that the rawdev in question is of the expected type. For +example, the following code in ``test_ioat_rawdev.c`` is used to identify +the IOAT rawdev device for use for the tests: + +.. code-block:: C + + for (i = 0; i < count && !found; i++) { + struct rte_rawdev_info info = { .dev_private = NULL }; + found = (rte_rawdev_info_get(i, &info) == 0 && + strcmp(info.driver_name, + IOAT_PMD_RAWDEV_NAME_STR) == 0); + } + +When calling the ``rte_rawdev_info_get()`` API for an IOAT rawdev device, +the ``dev_private`` field in the ``rte_rawdev_info`` struct should either +be NULL, or else be set to point to a structure of type +``rte_ioat_rawdev_config``, in which case the size of the configured device +input ring will be returned in that structure.
drivers/raw/ioat/ioat_rawdev.c +11 -0
@@ -24,10 +24,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +} +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_info_get = ioat_dev_info_get, };struct rte_rawdev *rawdev =NULL ;
drivers/raw/ioat/rte_ioat_rawdev.h +11 -0
@@ -24,6 +24,17 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* Configuration structure for an ioat rawdev instance +* +* This structure is to be passed as the ".dev_private" parameter when +* calling the rte_rawdev_get_info() and rte_rawdev_configure() APIs on +* an ioat rawdev instance. +*/ +struct rte_ioat_rawdev_config { +unsigned short ring_size; +}; +/** * @internal * Structure representing a device instance
Signature · 1 line
2.21.0
PATCH v2 6/8 raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonAllow initializing a driver instance. Include selftest to validate these functions.
---
V2: test cases placed in self-test routine
--- app/test/test_rawdev.c | 2 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 ++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 78 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 15 ++++++ 7 files changed, 170 insertions(+), 2 deletions(-) create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c
app/test/test_rawdev.c +1 -1
@@ -36,7 +36,7 @@ test_rawdev_selftest_ioat(void)struct rte_rawdev_info info = { .dev_private =NULL };if (rte_rawdev_info_get (i, &info) ==0 &&strstr (info.driver_name,"ioat" ) !=NULL ) -return 0 ; +return rte_rawdev_selftest (i); }printf ("No IOAT rawdev found, skipping tests\n" );
doc/guides/rawdevs/ioat_rawdev.rst +32 -0
@@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API.
drivers/raw/ioat/Makefile +1 -0
@@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h
drivers/raw/ioat/ioat_rawdev.c +78 -0
95 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c new file +41 -0
41 lines not on this page · raw message
drivers/raw/ioat/meson.build +2 -1
@@ -2,7 +2,8 @@ # Copyright 2019 Intel Corporation build = dpdk_conf.has('RTE_ARCH_X86') -sources = files('ioat_rawdev.c') +sources = files('ioat_rawdev.c', + 'ioat_rawdev_test.c') deps += ['rawdev', 'bus_pci'] install_headers('rte_ioat_rawdev.h',
drivers/raw/ioat/rte_ioat_rawdev.h +15 -0
@@ -14,6 +14,7 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <x86intrin.h> #include <rte_memory.h> #include <rte_ioat_spec.h>
@@ -47,9 +48,23 @@ struct rte_ioat_rawdev {unsigned short ring_size;struct rte_ioat_desc *desc_ring; +__m128i *hdls;/* completion handles for returning to user */ /* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; }; +/** +* @internal +* Structure representing a descriptor for a copy operation +*/ +struct rte_ioat_desc { + uint32_t xfer_size __rte_cache_aligned; +uint32_t desc_control; +phys_addr_t src_addr;/* 64 bits */ +phys_addr_t dest_addr;/* 64 bits */ +phys_addr_t next_desc_addr;/* 64 bits */ +uint64_t op_type_specific[4 ]; +}; +#endif
Signature · 1 line
2.21.0
PATCH v2 7/8 raw/ioat: add statistics functions
In reply to Bruce RichardsonAdd stats functions to track what is happening in the driver, and put unit tests to check those.
---
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 14 +++++++++ drivers/raw/ioat/ioat_rawdev.c | 44 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 39 +++++++++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 6 ++++ 4 files changed, 103 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +14 -0
@@ -149,3 +149,17 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. + +Querying Device Statistics +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The statistics from the IOAT rawdev device can be got via the xstats +functions in the ``rte_rawdev`` library, i.e. +``rte_rawdev_xstats_names_get()``, ``rte_rawdev_xstats_get()`` and +``rte_rawdev_xstats_by_name_get``. The statistics returned for each device +instance are: + +* ``failed_enqueues`` +* ``successful_enqueues`` +* ``copies_started`` +* ``copies_completed``
drivers/raw/ioat/ioat_rawdev.c +44 -0
62 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c +39 -0
@@ -2,6 +2,7 @@* Copyright(c) 2019 Intel Corporation */ +#include <inttypes.h> #include "rte_rawdev.h" #include "rte_ioat_rawdev.h"
@@ -13,6 +14,11 @@ ioat_rawdev_test(uint16_t dev_id)#define IOAT_TEST_RINGSIZE 512struct rte_ioat_rawdev_config p = { .ring_size =-1 };struct rte_rawdev_info info = { .dev_private = &p }; +struct rte_rawdev_xstats_name *snames =NULL ; +uint64_t *stats =NULL ; +unsigned int *ids =NULL ; +unsigned int nb_xstats; +unsigned int i;rte_rawdev_info_get (dev_id, &info);if (p.ring_size !=0 ) {
@@ -37,5 +43,38 @@ ioat_rawdev_test(uint16_t dev_id)printf ("Error with rte_rawdev_start()\n" );return -1 ; } + +/* allocate memory for xstats names and values */ + nb_xstats =rte_rawdev_xstats_names_get (dev_id,NULL ,0 ); + + snames =malloc (sizeof (*snames) * nb_xstats); +if (snames ==NULL ) { +printf ("Error allocating xstat names memory\n" ); +return -1 ; + } +rte_rawdev_xstats_names_get (dev_id, snames, nb_xstats); + + ids =malloc (sizeof (*ids) * nb_xstats); +if (ids ==NULL ) { +printf ("Error allocating xstat ids memory\n" ); +return -1 ; + } +for (i =0 ; i < nb_xstats; i++) + ids[i] = i; + + stats =malloc (sizeof (*stats) * nb_xstats); +if (stats ==NULL ) { +printf ("Error allocating xstat memory\n" ); +return -1 ; + } + +rte_rawdev_xstats_get (dev_id, ids, stats, nb_xstats); +for (i =0 ; i < nb_xstats; i++) +printf ("%s: %" PRIu64" " , snames[i].name, stats[i]); +printf ("\n" ); + +free (snames); +free (stats); +free (ids);return 0 ; }
drivers/raw/ioat/rte_ioat_rawdev.h +6 -0
@@ -50,6 +50,12 @@ struct rte_ioat_rawdev {struct rte_ioat_desc *desc_ring;__m128i *hdls;/* completion handles for returning to user */ +/* some statistics for tracking, if added/changed update xstats fns*/ + uint64_t enqueue_failed __rte_cache_aligned; +uint64_t enqueued; +uint64_t started; +uint64_t completed; +/* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; };
Signature · 1 line
2.21.0
PATCH v2 8/8 raw/ioat: add local API to perform copies
In reply to Bruce RichardsonAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
---
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 100 +++++++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev_test.c | 161 +++++++++++++++++++++++++- drivers/raw/ioat/meson.build | 2 +- drivers/raw/ioat/rte_ioat_rawdev.h | 168 +++++++++++++++++++++++++++- 5 files changed, 427 insertions(+), 5 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +100 -0
@@ -150,6 +150,106 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. +Performing Data Copies +~~~~~~~~~~~~~~~~~~~~~~~ + +To perform data copies using IOAT rawdev devices, the functions +``rte_ioat_enqueue_copy()`` and ``rte_ioat_do_copies()`` should be used. +Once copies have been completed, the completion will be reported back when +the application calls ``rte_ioat_completed_copies()``. + +The ``rte_ioat_enqueue_copy()`` function enqueues a single copy to the +device ring for copying at a later point. The parameters to that function +include the physical addresses of both the source and destination buffers, +as well as two "handles" to be returned to the user when the copy is +completed. These handles can be arbitrary values, but two are provided so +that the library can track handles for both source and destination on +behalf of the user, e.g. virtual addresses for the buffers, or mbuf +pointers if packet data is being copied. + +While the ``rte_ioat_enqueue_copy()`` function enqueues a copy operation on +the device ring, the copy will not actually be performed until after the +application calls the ``rte_ioat_do_copies()`` function. This function +informs the device hardware of the elements enqueued on the ring, and the +device will begin to process them. It is expected that, for efficiency +reasons, a burst of operations will be enqueued to the device via multiple +enqueue calls between calls to the ``rte_ioat_do_copies()`` function. + +The following code from ``test_ioat_rawdev.c`` demonstrates how to enqueue +a burst of copies to the device and start the hardware processing of them: + +.. code-block:: C + + struct rte_mbuf *srcs[32], *dsts[32]; + unsigned int j; + + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data; + + srcs[i] = rte_pktmbuf_alloc(pool); + dsts[i] = rte_pktmbuf_alloc(pool); + srcs[i]->data_len = srcs[i]->pkt_len = length; + dsts[i]->data_len = dsts[i]->pkt_len = length; + src_data = rte_pktmbuf_mtod(srcs[i], char *); + + for (j = 0; j < length; j++) + src_data[j] = rand() & 0xFF; + + if (rte_ioat_enqueue_copy(dev_id, + srcs[i]->buf_iova + srcs[i]->data_off, + dsts[i]->buf_iova + dsts[i]->data_off, + length, + (uintptr_t)srcs[i], + (uintptr_t)dsts[i], + 0 /* nofence */) != 1) { + printf("Error with rte_ioat_enqueue_copy for buffer %u\n", + i); + return -1; + } + } + rte_ioat_do_copies(dev_id); + +To retrieve information about completed copies, the API +``rte_ioat_completed_copies()`` should be used. This API will return to the +application a set of completion handles passed in when the relevant copies +were enqueued. + +The following code from ``test_ioat_rawdev.c`` shows the test code +retrieving information about the completed copies and validating the data +is correct before freeing the data buffers using the returned handles: + +.. code-block:: C + + if (rte_ioat_completed_copies(dev_id, 64, (void *)completed_src, + (void *)completed_dst) != RTE_DIM(srcs)) { + printf("Error with rte_ioat_completed_copies\n"); + return -1; + } + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data, *dst_data; + + if (completed_src[i] != srcs[i]) { + printf("Error with source pointer %u\n", i); + return -1; + } + if (completed_dst[i] != dsts[i]) { + printf("Error with dest pointer %u\n", i); + return -1; + } + + src_data = rte_pktmbuf_mtod(srcs[i], char *); + dst_data = rte_pktmbuf_mtod(dsts[i], char *); + for (j = 0; j < length; j++) + if (src_data[j] != dst_data[j]) { + printf("Error with copy of packet %u, byte %u\n", + i, j); + return -1; + } + rte_pktmbuf_free(srcs[i]); + rte_pktmbuf_free(dsts[i]); + } + + Querying Device Statistics ~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/raw/ioat/Makefile +1 -0
@@ -12,6 +12,7 @@ CFLAGS += $(WERROR_FLAGS) LDLIBS += -lrte_eal -lrte_rawdev LDLIBS += -lrte_pci -lrte_bus_pci +LDLIBS += -lrte_mbuf -lrte_mempool # library version LIBABIVER := 1
drivers/raw/ioat/ioat_rawdev_test.c +158 -3
189 lines not on this page · raw message
drivers/raw/ioat/meson.build +1 -1
@@ -4,7 +4,7 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c', 'ioat_rawdev_test.c') -deps += ['rawdev', 'bus_pci'] +deps += ['rawdev', 'bus_pci', 'mbuf'] install_headers('rte_ioat_rawdev.h', 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_rawdev.h +167 -1
184 lines not on this page · raw message
Signature · 1 line
2.21.0
v3 cover raw/ioat: driver for Intel QuickData Technology
In reply to Bruce RichardsonThis patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
V3:
* removed DPDK-specific structure for the descriptor format and reused the structure in the imported file rte_ioat_spec.h
V2:
* moved tests to rawdev selftest function
* some checkpatch and other small cleanups
* added extra documentation details on supported hardware
* aligned the changes to dpdk-devbind with the changes in the NTB set for consistency
Bruce Richardson (8):
raw/ioat: add initial support for ioat rawdev driver usertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
MAINTAINERS | 6 +- app/test/test_rawdev.c | 19 ++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 265 +++++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 + drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 31 ++ drivers/raw/ioat/ioat_rawdev.c | 313 ++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 235 +++++++++++++++ drivers/raw/ioat/meson.build | 10 + drivers/raw/ioat/rte_ioat_rawdev.h | 230 ++++++++++++++ drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + usertools/dpdk-devbind.py | 10 + 20 files changed, 1448 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_ioat_spec.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
Signature · 1 line
2.21.0
PATCH v3 1/8 raw/ioat: add initial support for ioat rawdev driver
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
---
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
--- MAINTAINERS | 6 +- app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 ++ config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 ++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 +++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 ++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 258 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +5 -1
@@ -1068,6 +1068,11 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/
@@ -1078,7 +1083,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processing -----------------
app/test/test_rawdev.c +8 -0
@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }REGISTER_TEST_COMMAND (rawdev_autotest, test_rawdev_selftest_skeleton); + +static int +test_rawdev_selftest_ioat (void ) +{ +return TEST_SKIPPED ; +} + +REGISTER_TEST_COMMAND (ioat_rawdev_autotest, test_rawdev_selftest_ioat);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=n
config/common_base +5 -0
@@ -747,6 +747,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n # CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y +# +# Compile PMD for Intel IOAT raw device +# +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +1 -0
@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +63 -0
63 lines not on this page · raw message
doc/guides/rel_notes/release_19_08.rst +11 -0
@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================
@@ -99,6 +101,15 @@ New Features Updated ``librte_telemetry`` to fetch the global metrics from the ``librte_metrics`` library. +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of + Intel\ |reg| I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_, + allows data copies to be done by hardware instead + of via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items -------------
drivers/raw/Makefile +1 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright(c) 2019 Intel Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +93 -0
93 lines not on this page · raw message
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright 2019 Intel Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior notice +*/ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() */ +#define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ +#define IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +3 -1
@@ -1,7 +1,9 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat', + 'skeleton_rawdev'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +1 -0
@@ -306,6 +306,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev _LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke endif # CONFIG_RTE_LIBRTE_IFPGA_BUS +_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev endif # CONFIG_RTE_LIBRTE_RAWDEV endif # !CONFIG_RTE_BUILD_SHARED_LIBS
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
Quoted 12 lines · Bruce Richardson
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit testsSigned-off-by Bruce Richardson <[email protected]>---
<snip>
MAINTAINERS M: Nipun Gupta <[email protected]> · line 1086
Quoted 9 lines
NXP DPAA2 QDMA
M: Nipun Gupta <[email protected]>
F:drivers/raw/dpaa2_qdma/MAINTAINERS +0
-1@@ -1078,7 +1083,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processing
Unnecessary whitespace change :)
Quoted 8 lines
MAINTAINERS +0
-0-----------------app/test/test_rawdev.c +0
-0@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }
<snip>
doc/guides/rel_notes/release_19_08.rst line 5
Quoted 10 lines
doc/guides/rel_notes/release_19_08.rst +2
-0@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> +
Is this accidental? This doesn't look related to this patch.
Other than a couple of nitpicks, LGTM
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
Quoted 16 lines · Anatoly
Quoted 12 lines · Bruce Richardson
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit testsSigned-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 11 lines
MAINTAINERS M: Nipun Gupta <[email protected]> · line 1085
Quoted 8 lines
NXP DPAA2 QDMA
M: Nipun Gupta <[email protected]>
F:drivers/raw/dpaa2_qdma/MAINTAINERS +0
-1@@ -1078,7 +1083,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processingUnnecessary whitespace change :)
Thanks, good catch.
Quoted 10 lines
Quoted 7 lines
MAINTAINERS +0
-0-----------------app/test/test_rawdev.c +0
-0@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }<snip>
Quoted 12 lines
doc/guides/rel_notes/release_19_08.rst line 4
Quoted 9 lines
doc/guides/rel_notes/release_19_08.rst +2
-0@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> +Is this accidental? This doesn't look related to this patch.
No, it's not accidental. It's the sphinx / rst include which allows the use of the actual symbols for (c) (R) etc. I make use of those symbols in the official product names in the following documentation. [It's a useful include that I think should always be put in our docs.]
Re: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
In reply to Bruce RichardsonHi Bruce,
-----Original Message-----
From: dev <[email protected]> On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 6:41 PM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
---
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
--- MAINTAINERS | 6 +- app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 ++ config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 ++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 +++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 ++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 258 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +5 -1
@@ -1068,6 +1068,11 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/
@@ -1078,7 +1083,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processing -----------------
app/test/test_rawdev.c +8 -0
@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }REGISTER_TEST_COMMAND (rawdev_autotest, test_rawdev_selftest_skeleton); + +static int +test_rawdev_selftest_ioat (void ) +{ +return TEST_SKIPPED ; +} + +REGISTER_TEST_COMMAND (ioat_rawdev_autotest, test_rawdev_selftest_ioat);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=n
config/common_base +4 -0
@@ -747,6 +747,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n # CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y +# +# Compile PMD for Intel IOAT raw device # +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +1 -0
@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +63 -0
63 lines not on this page · raw message
+ Registers (rev 04)
+ 00:04.6 System peripheral: Intel Corporation Sky Lake-E CBDMA
+ Registers (rev 04)
+ 00:04.7 System peripheral: Intel Corporation Sky Lake-E CBDMA
+ Registers (rev 04)
+
+
+Compilation
+------------
+
+For builds done with ``make``, the driver compilation is enabled by the +``CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV`` build configuration option. This +is enabled by default in builds for x86 platforms, and disabled in +other configurations.
+
+For builds using ``meson`` and ``ninja``, the driver will be built when +the target platform is x86-based.
doc/guides/rel_notes/release_19_08.rst +11 -0
@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================
@@ -99,6 +101,15 @@ New Features Updated ``librte_telemetry`` to fetch the global metrics from the ``librte_metrics`` library. +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of Intel\ |reg| + I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel-techno + logy.html>`_, allows data copies to be done by hardware instead of + via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items -------------
drivers/raw/Makefile +1 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause # Copyright(c) 2019 Intel +Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +16 -0
@@ -0,0 +1,93 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#include <rte_bus_pci.h> +#include <rte_rawdev_pmd.h> + +#include "rte_ioat_rawdev.h" + +/* Dynamic log type identifier */ +int ioat_pmd_logtype; + +static struct rte_pci_driver ioat_pmd_drv; + +#define IOAT_VENDOR_ID 0x8086 +#define IOAT_DEVICE_ID 0x2021
In the second patch, the script dpdk-devbind.py supports CBDMAs in both Broadwell and Skylake. But you only supports Skylake here.
Thanks,
Jiayu
+
+#define IOAT_PMD_LOG(level, fmt, args...) rte_log(RTE_LOG_ ## level, \
+ ioat_pmd_logtype, "%s(): " fmt "\n", __func__, ##args)
+
+#define IOAT_PMD_DEBUG(fmt, args...) IOAT_PMD_LOG(DEBUG, fmt, ## args) +#define IOAT_PMD_INFO(fmt, args...) IOAT_PMD_LOG(INFO, fmt, ## args) +#define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) +#define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args)
+
+static int
+ioat_rawdev_create(const char *name, struct rte_pci_device *dev) {
+ RTE_SET_USED(name);
+ RTE_SET_USED(dev);
+ return 0;
+}
+
+static int
+ioat_rawdev_destroy(const char *name)
+{
+ RTE_SET_USED(name);
+ return 0;
+}
+
+static int
+ioat_rawdev_probe(struct rte_pci_driver *drv, struct rte_pci_device +*dev) {
+ char name[32];
+ int ret = 0;
+
+
+ rte_pci_device_name(&dev->addr, name, sizeof(name));
+ IOAT_PMD_INFO("Init %s on NUMA node %d", name, dev->device.numa_node);
+
+ dev->device.driver = &drv->driver;
+ ret = ioat_rawdev_create(name, dev);
+ return ret;
+}
+
+static int
+ioat_rawdev_remove(struct rte_pci_device *dev) {
+ char name[32];
+ int ret;
+
+ rte_pci_device_name(&dev->addr, name, sizeof(name));
+
+ IOAT_PMD_INFO("Closing %s on NUMA node %d",
+ name, dev->device.numa_node);
+
+ ret = ioat_rawdev_destroy(name);
+ return ret;
+}
+
+static const struct rte_pci_id pci_id_ioat_map[] = {
+ { RTE_PCI_DEVICE(IOAT_VENDOR_ID, IOAT_DEVICE_ID) },
+ { .vendor_id = 0, /* sentinel */ },
+};
+
+static struct rte_pci_driver ioat_pmd_drv = {
+ .id_table = pci_id_ioat_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
+ RTE_PCI_DRV_IOVA_AS_VA,
+ .probe = ioat_rawdev_probe,
+ .remove = ioat_rawdev_remove,
+};
+
+RTE_PMD_REGISTER_PCI(IOAT_PMD_RAWDEV_NAME, ioat_pmd_drv); +RTE_PMD_REGISTER_PCI_TABLE(IOAT_PMD_RAWDEV_NAME, pci_id_ioat_map); +RTE_PMD_REGISTER_KMOD_DEP(IOAT_PMD_RAWDEV_NAME, "* igb_uio | +uio_pci_generic");
+
+RTE_INIT(ioat_pmd_init_log)
+{
+ ioat_pmd_logtype = rte_log_register(IOAT_PMD_LOG_NAME);
+ if (ioat_pmd_logtype >= 0)
+ rte_log_set_level(ioat_pmd_logtype, RTE_LOG_INFO); }
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause # Copyright 2019 Intel +Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior +notice */ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() +*/ #define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ #define +IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +3 -1
@@ -1,7 +1,9 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat', + 'skeleton_rawdev'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +1 -0
@@ -306,6 +306,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev _LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke endif # CONFIG_RTE_LIBRTE_IFPGA_BUS +_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev endif # CONFIG_RTE_LIBRTE_RAWDEV endif # !CONFIG_RTE_BUILD_SHARED_LIBS
--
2.21.0
Re: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
Hi Jiayu,
01/07/2019 09:38, Hu, Jiayu:
Hi Bruce,
-----Original Message-----
From: dev <[email protected]> On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 6:41 PM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
Add stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
---
V2: don't create a new file for unit testing, add to existing rawdev test file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
---
MAINTAINERS | 6 +- app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 ++ config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 ++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 +++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 ++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 258 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst
create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +0 -0
@@ -1068,6 +1068,11 @@ M: Tianfei zhang <[email protected]>
F: drivers/raw/ifpga_rawdev/
F: doc/guides/rawdevs/ifpga_rawdev.rst
+IOAT Rawdev
+M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/
+F: doc/guides/rawdevs/ioat_rawdev.rst
+
NXP DPAA2 QDMA
M: Nipun Gupta <[email protected]>
F: drivers/raw/dpaa2_qdma/
@@ -1078,7 +1083,6 @@ M: Nipun Gupta <[email protected]>
F: drivers/raw/dpaa2_cmdif/
F: doc/guides/rawdevs/dpaa2_cmdif.rst
-
Packet processing
-----------------
app/test/test_rawdev.c +0 -0
@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void)
}
REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton); +
+static int
+test_rawdev_selftest_ioat(void)
+{
+ return TEST_SKIPPED;
+}
+
+REGISTER_TEST_COMMAND(ioat_rawdev_autotest, test_rawdev_selftest_ioat);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
CONFIG_RTE_SCHED_VECTOR=n
config/common_base +0 -0
@@ -747,6 +747,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n
#CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y
+#
+# Compile PMD for Intel IOAT raw device #
+CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y
+
#
# Compile librte_ring
#
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +0 -0
@@ -14,3 +14,4 @@ application through rawdev API.
dpaa2_cmdifdpaa2_qdmaifpga_rawdev
+ ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +63 -0
63 lines not on this page · raw message
+ Registers (rev 04)
+ 00:04.6 System peripheral: Intel Corporation Sky Lake-E CBDMA
+ Registers (rev 04)
+ 00:04.7 System peripheral: Intel Corporation Sky Lake-E CBDMA
+ Registers (rev 04)
+
+
+Compilation
+------------
+
+For builds done with ``make``, the driver compilation is enabled by the +``CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV`` build configuration option. This +is enabled by default in builds for x86 platforms, and disabled in +other configurations.
+
+For builds using ``meson`` and ``ninja``, the driver will be built when +the target platform is x86-based.
doc/guides/rel_notes/release_19_08.rst +0 -0
@@ -1,6 +1,8 @@
.. SPDX-License-Identifier: BSD-3-Clause
Copyright 2019 The DPDK contributors
+.. include:: <isonum.txt>
+
DPDK Release 19.08
==================
@@ -99,6 +101,15 @@ New Features
Updated ``librte_telemetry`` to fetch the global metrics from the ``librte_metrics`` library.
+* **Added Intel QuickData Technology PMD**
+
+ The PMD for Intel\ |reg| QuickData Technology, part of Intel\ |reg|
+ I/O Acceleration Technology `(Intel I/OAT)
+ <https://www.intel.com/content/www/us/en/wireless-network/accel-techno
+ logy.html>`_, allows data copies to be done by hardware instead of
+ via software, reducing cycles spent copying large blocks of data in
+ applications.
+
Removed Items
-------------
drivers/raw/Makefile +0 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif
DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat
include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause # Copyright(c) 2019 Intel +Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +16 -0
@@ -0,0 +1,93 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#include <rte_bus_pci.h> +#include <rte_rawdev_pmd.h> + +#include "rte_ioat_rawdev.h" + +/* Dynamic log type identifier */ +int ioat_pmd_logtype; + +static struct rte_pci_driver ioat_pmd_drv; + +#define IOAT_VENDOR_ID 0x8086 +#define IOAT_DEVICE_ID 0x2021
In the second patch, the script dpdk-devbind.py supports CBDMAs in both Broadwell and Skylake. But you only supports Skylake here.
Thanks,
Jiayu
+
+#define IOAT_PMD_LOG(level, fmt, args...) rte_log(RTE_LOG_ ## level, \
+ ioat_pmd_logtype, "%s(): " fmt "\n", __func__, ##args)
+
+#define IOAT_PMD_DEBUG(fmt, args...) IOAT_PMD_LOG(DEBUG, fmt, ## args) +#define IOAT_PMD_INFO(fmt, args...) IOAT_PMD_LOG(INFO, fmt, ## args) +#define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) +#define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args)
+
+static int
+ioat_rawdev_create(const char *name, struct rte_pci_device *dev) {
+ RTE_SET_USED(name);
+ RTE_SET_USED(dev);
+ return 0;
+}
+
+static int
+ioat_rawdev_destroy(const char *name)
+{
+ RTE_SET_USED(name);
+ return 0;
+}
+
+static int
+ioat_rawdev_probe(struct rte_pci_driver *drv, struct rte_pci_device +*dev) {
+ char name[32];
+ int ret = 0;
+
+
+ rte_pci_device_name(&dev->addr, name, sizeof(name));
+ IOAT_PMD_INFO("Init %s on NUMA node %d", name, dev->device.numa_node);
+
+ dev->device.driver = &drv->driver;
+ ret = ioat_rawdev_create(name, dev);
+ return ret;
+}
+
+static int
+ioat_rawdev_remove(struct rte_pci_device *dev) {
+ char name[32];
+ int ret;
+
+ rte_pci_device_name(&dev->addr, name, sizeof(name));
+
+ IOAT_PMD_INFO("Closing %s on NUMA node %d",
+ name, dev->device.numa_node);
+
+ ret = ioat_rawdev_destroy(name);
+ return ret;
+}
+
+static const struct rte_pci_id pci_id_ioat_map[] = {
+ { RTE_PCI_DEVICE(IOAT_VENDOR_ID, IOAT_DEVICE_ID) },
+ { .vendor_id = 0, /* sentinel */ },
+};
+
+static struct rte_pci_driver ioat_pmd_drv = {
+ .id_table = pci_id_ioat_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
+ RTE_PCI_DRV_IOVA_AS_VA,
+ .probe = ioat_rawdev_probe,
+ .remove = ioat_rawdev_remove,
+};
+
+RTE_PMD_REGISTER_PCI(IOAT_PMD_RAWDEV_NAME, ioat_pmd_drv); +RTE_PMD_REGISTER_PCI_TABLE(IOAT_PMD_RAWDEV_NAME, pci_id_ioat_map); +RTE_PMD_REGISTER_KMOD_DEP(IOAT_PMD_RAWDEV_NAME, "* igb_uio | +uio_pci_generic");
+
+RTE_INIT(ioat_pmd_init_log)
+{
+ ioat_pmd_logtype = rte_log_register(IOAT_PMD_LOG_NAME);
+ if (ioat_pmd_logtype >= 0)
+ rte_log_set_level(ioat_pmd_logtype, RTE_LOG_INFO); }
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause # Copyright 2019 Intel +Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior +notice */ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() +*/ #define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ #define +IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +0 -0
@@ -1,7 +1,9 @@
# SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP
-drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma',
+ 'ifpga_rawdev', 'ioat',
+ 'skeleton_rawdev']std_deps = ['rawdev']config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV'driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +0 -0
@@ -306,6 +306,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y)
_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev
_LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke
endif # CONFIG_RTE_LIBRTE_IFPGA_BUS
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev
endif # CONFIG_RTE_LIBRTE_RAWDEV
endif # !CONFIG_RTE_BUILD_SHARED_LIBS
--
2.21.0
Re: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
In reply to Bruce RichardsonSorry for the incorrect format in the last mail.
The comment is repeated below.
Quoted 333 lines
-----Original Message-----
From: dev <[email protected]> On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 6:41 PM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driverAdd stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit testsSigned-off-by Bruce Richardson <[email protected]>---
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci--- MAINTAINERS | 6 +- app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 ++ config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 ++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 +++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 ++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 258 insertions(+), 2 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.mapMAINTAINERS +5
-1@@ -1068,6 +1068,11 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/@@ -1078,7 +1083,6 @@ M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst - Packet processing -----------------app/test/test_rawdev.c +0
-0@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) } REGISTER_TEST_COMMAND(rawdev_autotest,
test_rawdev_selftest_skeleton);app/test/test_rawdev.c +8
-0+ +static int +test_rawdev_selftest_ioat (void ) +{ +return TEST_SKIPPED ; +} + +REGISTER_TEST_COMMAND(ioat_rawdev_autotest,
test_rawdev_selftest_ioat);config/common_armv8a_linux +1
-0@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=nconfig/common_base +0
-0@@ -747,6 +747,11 @@
CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n
#
CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=yconfig/common_base +5
-0+# +# Compile PMD for Intel IOAT raw device +# +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #config/defconfig_arm-armv7a-linuxapp-gcc +1
-0@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=nconfig/defconfig_ppc_64-power8-linuxapp-gcc +1
-0@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=ndoc/guides/rawdevs/index.rst +1
-0@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdevdoc/guides/rawdevs/ioat_rawdev.rst new file +7
-0@@ -0,0 +1,63 @@ +.. SPDX-License-Identifier: BSD-3-Clause + Copyright(c) 2019 Intel Corporation. + +.. include:: <isonum.txt> + +IOAT Rawdev Driver for Intel\ |reg| QuickData Technology +======================================================================
doc/guides/rawdevs/ioat_rawdev.rst new file +5
-0+ +The ``ioat`` rawdev driver provides a poll-mode driver (PMD) for Intel\ |reg| +QuickData Technology, part of Intel\ |reg| I/O Acceleration Technology +`(Intel I/OAT) +<https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_.
+This PMD, when used on supported hardware, allows data copies, for example,
+cloning packet data, to be accelerated by that hardware rather than having todoc/guides/rawdevs/ioat_rawdev.rst new file +6
-0+be done by software, freeing up CPU cycles for other tasks. + +Hardware Requirements +---------------------- + +On Linux, the presence of an Intel\ |reg| QuickData Technology hardwarecan
doc/guides/rawdevs/ioat_rawdev.rst new file +3
-0+be detected by checking the output of the ``lspci`` command, where the +hardware will be often listed as "Crystal Beach DMA" or "CBDMA". For +example, on a system with Intel\ |reg| Xeon\ |reg| CPU E5-2699 v4@2.20GHz,
doc/guides/rawdevs/ioat_rawdev.rst new file +6
-0+lspci shows: + +.. code-block:: console + + # lspci | grep DMA + 00:04.0 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/XeonE3 v4/Xeon D Crystal Beach DMA Channel 0 (rev 01)
+ 00:04.1 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 1 (rev 01)
+ 00:04.2 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 2 (rev 01)
+ 00:04.3 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 3 (rev 01)
+ 00:04.4 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 4 (rev 01)
+ 00:04.5 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 5 (rev 01)
+ 00:04.6 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 6 (rev 01)
+ 00:04.7 System peripheral: Intel Corporation Xeon E7 v4/Xeon E5 v4/Xeon E3 v4/Xeon D Crystal Beach DMA Channel 7 (rev 01)doc/guides/rawdevs/ioat_rawdev.rst new file +8
-0+ +On a system with Intel\ |reg| Xeon\ |reg| Gold 6154 CPU @ 3.00GHz, lspci +shows: + +.. code-block:: console + + # lspci | grep DMA + 00:04.0 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers(rev 04)
+ 00:04.1 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)
+ 00:04.2 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)
+ 00:04.3 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)
+ 00:04.4 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)
+ 00:04.5 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)
+ 00:04.6 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)
+ 00:04.7 System peripheral: Intel Corporation Sky Lake-E CBDMA Registers (rev 04)doc/guides/rawdevs/ioat_rawdev.rst new file +7
-0+ + +Compilation +------------ + +For builds done with ``make``, the driver compilation is enabled by the +``CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV`` build configuration option.This is
doc/guides/rawdevs/ioat_rawdev.rst new file +5
-0+enabled by default in builds for x86 platforms, and disabled in other +configurations. + +For builds using ``meson`` and ``ninja``, the driver will be built when the +target platform is x86-based.doc/guides/rel_notes/release_19_08.rst +11
-0@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================@@ -99,6 +101,15 @@ New Features Updated ``librte_telemetry`` to fetch the global metrics from the ``librte_metrics`` library. +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of + Intel\ |reg| I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel- technology.html>`_, + allows data copies to be done by hardware instead + of via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items-------------
drivers/raw/Makefile +0
-0@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) +=
dpaa2_cmdif
DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) +=dpaa2_qdma
endif
DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) +=ifpga_rawdevdrivers/raw/Makefile +1
-0+DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mkdrivers/raw/ioat/Makefile new file +28
-0@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright(c) 2019 Intel Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mkdrivers/raw/ioat/ioat_rawdev.c new file +6
-0@@ -0,0 +1,93 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#include <rte_bus_pci.h> +#include <rte_rawdev_pmd.h>
drivers/raw/ioat/ioat_rawdev.c line 16
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev.c new file +10
-0+ +#include "rte_ioat_rawdev.h" + +/* Dynamic log type identifier */ +int ioat_pmd_logtype; + +static struct rte_pci_driver ioat_pmd_drv; + +#define IOAT_VENDOR_ID 0x8086 +#define IOAT_DEVICE_ID 0x2021
In the second patch, the script dpdk-devbind.py supports CBDMAs in both Broadwell and Skylake. But you only supports Skylake here.
Thanks,
Jiayu
Quoted 171 lines
drivers/raw/ioat/ioat_rawdev.c new file +5
-0+ +#define IOAT_PMD_LOG(level, fmt, args...) rte_log(RTE_LOG_ ## level, \ + ioat_pmd_logtype, "%s(): " fmt "\n", __func__, ##args) + +#define IOAT_PMD_DEBUG(fmt, args...) IOAT_PMD_LOG(DEBUG, fmt, ##args)
drivers/raw/ioat/ioat_rawdev.c new file +3
-0+#define IOAT_PMD_INFO(fmt, args...) IOAT_PMD_LOG(INFO, fmt, ## args) +#define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) +#define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt,## args)
drivers/raw/ioat/ioat_rawdev.c new file +25
-0+ +static int +ioat_rawdev_create (const char *name,struct rte_pci_device *dev) +{ +RTE_SET_USED (name); +RTE_SET_USED (dev); +return 0 ; +} + +static int +ioat_rawdev_destroy (const char *name) +{ +RTE_SET_USED (name); +return 0 ; +} + +static int +ioat_rawdev_probe (struct rte_pci_driver *drv,struct rte_pci_device *dev) +{ +char name[32 ]; +int ret =0 ; + + +rte_pci_device_name (&dev->addr, name,sizeof (name)); + IOAT_PMD_INFO("Init %s on NUMA node %d", name, dev-Quoted 1 line
device.numa_node);
drivers/raw/ioat/ioat_rawdev.c new file +29
-0+ + dev->device.driver = &drv->driver; + ret =ioat_rawdev_create (name, dev); +return ret; +} + +static int +ioat_rawdev_remove (struct rte_pci_device *dev) +{ +char name[32 ]; +int ret; + +rte_pci_device_name (&dev->addr, name,sizeof (name)); + +IOAT_PMD_INFO ("Closing %s on NUMA node %d" , + name, dev->device.numa_node); + + ret =ioat_rawdev_destroy (name); +return ret; +} + +static const struct rte_pci_id pci_id_ioat_map[] = { + {RTE_PCI_DEVICE (IOAT_VENDOR_ID ,IOAT_DEVICE_ID ) }, + { .vendor_id =0 ,/* sentinel */ }, +}; + +static struct rte_pci_driver ioat_pmd_drv = { + .id_table = pci_id_ioat_map, + .drv_flags = RTE_PCI_DRV_NEED_MAPPING |
RTE_PCI_DRV_INTR_LSC|drivers/raw/ioat/ioat_rawdev.c new file +7
-0+ RTE_PCI_DRV_IOVA_AS_VA, + .probe = ioat_rawdev_probe, + .remove = ioat_rawdev_remove, +}; + +RTE_PMD_REGISTER_PCI (IOAT_PMD_RAWDEV_NAME , ioat_pmd_drv); +RTE_PMD_REGISTER_PCI_TABLE(IOAT_PMD_RAWDEV_NAME,
pci_id_ioat_map); +RTE_PMD_REGISTER_KMOD_DEP(IOAT_PMD_RAWDEV_NAME, "*igb_uio|uio_pci_generic");drivers/raw/ioat/ioat_rawdev.c new file +7
-0+ +RTE_INIT (ioat_pmd_init_log) +{ + ioat_pmd_logtype =rte_log_register (IOAT_PMD_LOG_NAME ); +if (ioat_pmd_logtype >=0 ) +rte_log_set_level (ioat_pmd_logtype,RTE_LOG_INFO ); +}drivers/raw/ioat/meson.build new file +8
-0@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright 2019 Intel Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')drivers/raw/ioat/rte_ioat_rawdev.h new file +14
-0@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior notice
+ */drivers/raw/ioat/rte_ioat_rawdev.h new file +9
-0+ +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() */ +#define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ +#define IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif drivers/raw/ioat/rte_pmd_ioat_version.map new file +4
-0@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};drivers/raw/meson.build +3
-1@@ -1,7 +1,9 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat', + 'skeleton_rawdev'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'mk/rte.app.mk +0
-0@@ -306,6 +306,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -
lrte_pmd_ifpga_rawdev
_LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke
endif #CONFIG_RTE_LIBRTE_IFPGA_BUS
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev
endif #CONFIG_RTE_LIBRTE_RAWDEVendif # !CONFIG_RTE_BUILD_SHARED_LIBS
--
2.21.0
Re: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driver
Quoted 24 lines · Jiayu
Sorry for the incorrect format in the last mail.
The comment is repeated below.Quoted 21 lines
-----Original Message-----
From: dev <[email protected]> On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 6:41 PM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 1/8] raw/ioat: add initial support for ioat rawdev driverAdd stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit testsSigned-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 9 lines
+
drivers/raw/ioat/ioat_rawdev.c line 16
Quoted 4 lines
drivers/raw/ioat/ioat_rawdev.c new file +4
-0+static struct rte_pci_driver ioat_pmd_drv; + +#define IOAT_VENDOR_ID 0x8086 +#define IOAT_DEVICE_ID 0x2021In the second patch, the script
dpdk-devbind.pysupports CBDMAs in both Broadwell and Skylake. But you only supports Skylake here.
Will fix in v4, thanks.
/Bruce
PATCH v3 2/8 usertools/dpdk-devbind.py: add support for IOAT devices
In reply to Bruce RichardsonIn order to allow binding/unbinding of devices for use by the ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.
---
V2: rather than adding a DMA section, add to "misc (rawdev)" section where
other device types, e.g. ntb can also do.
NOTE: this set largely overlaps with the equivalent changes made by the patchset adding NTB support [1]. Since it's unclear which set will be added first, this set is based off the latest head.
[1] http://patches.dpdk.org/patch/55127/
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++++++++++ usertools/dpdk-devbind.py | 10 ++++++++++ 2 files changed, 21 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -61,3 +61,14 @@ configurations. For builds using ``meson`` and ``ninja``, the driver will be built when the target platform is x86-based. + +Device Setup +------------- + +The Intel\ |reg| QuickData Technology HW devices will need to be bound to a +user-space IO driver for use. The script ``dpdk-devbind.py`` script +included with DPDK can be used to view the state of the devices and to bind +them to a suitable DPDK-supported kernel driver. When querying the status +of the devices, they will appear under the category of "Misc (rawdev) +devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be +used to see the state of those devices alone.
usertools/dpdk-devbind.py +10 -0
@@ -36,11 +36,17 @@ octeontx2_npa = {'Class': '08', 'Vendor': '177d', 'Device': 'a0fb,a0fc', 'SVendor': None, 'SDevice': None} +intel_ioat_bdw = {'Class': '08', 'Vendor': '8086', 'Device': '6f20,6f21,6f22,6f23,6f24,6f25,6f26,6f27,6f2e,6f2f', + 'SVendor': None, 'SDevice': None} +intel_ioat_skx = {'Class': '08', 'Vendor': '8086', 'Device': '2021', + 'SVendor': None, 'SDevice': None} + network_devices = [network_class, cavium_pkx, avp_vnic, ifpga_class] crypto_devices = [encryption_class, intel_processor_class] eventdev_devices = [cavium_sso, cavium_tim, octeontx2_sso] mempool_devices = [cavium_fpa, octeontx2_npa] compress_devices = [cavium_zip] +misc_devices = [intel_ioat_bdw, intel_ioat_skx] # global dict ethernet devices present. Dictionary indexed by PCI address. # Each device within this is itself a dictionary of device properties
@@ -595,6 +601,8 @@ def show_status(): if status_dev == "compress" or status_dev == "all": show_device_status(compress_devices , "Compress") + if status_dev == "misc" or status_dev == "all": + show_device_status(misc_devices, "Misc (rawdev)") def parse_args(): '''Parses the command-line arguments given by the user and takes the
@@ -670,6 +678,7 @@ def do_arg_actions(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) show_status()
@@ -690,6 +699,7 @@ def main(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) do_arg_actions() if __name__ == "__main__":
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 2/8] usertools/dpdk-devbind.py: add support for IOAT devices
Quoted 7 lines · Bruce Richardson
In order to allow binding/unbinding of devices for use by the
ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.Signed-off-by Bruce Richardson <[email protected]>---
Signature · 2 lines
Thanks, Anatoly
PATCH v3 3/8 raw/ioat: add register definition file
In reply to Bruce RichardsonAdd in the list of registers for the device. File is taken from the SPDK project:
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
--- drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++++++++++++++ 3 files changed, 304 insertions(+), 1 deletion(-) create mode 100644 drivers/raw/ioat/rte_ioat_spec.h
drivers/raw/ioat/Makefile +1 -0
@@ -24,5 +24,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h +SYMLINK-y-include += rte_ioat_spec.h include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/meson.build +2 -1
@@ -5,4 +5,5 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c') deps += ['rawdev', 'bus_pci'] -install_headers('rte_ioat_rawdev.h') +install_headers('rte_ioat_rawdev.h', + 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_spec.h new file +301 -0
301 lines not on this page · raw message
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 3/8] raw/ioat: add register definition file
Quoted 7 lines · Bruce Richardson
Add in the list of registers for the device. File is taken from the SPDK project:
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
Signed-off-by Bruce Richardson <[email protected]>---
The indentation is slightly inconsistent across the file, but LGTM
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 3/8] raw/ioat: add register definition file
Quoted 4 lines · Anatoly
Quoted 3 lines · Bruce Richardson
Add in the list of registers for the device. File is taken from the SPDK project:
Quoted 9 lines
Quoted 4 lines
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
Signed-off-by Bruce Richardson <[email protected]>---
The indentation is slightly inconsistent across the file, but LGTM
Acked-by Anatoly Burakov <[email protected]>
I've tried to keep the delta to the original file to a minimum, so didn't look at whitespace. I think it's ok as-is.
PATCH v3 4/8 raw/ioat: create device on probe and destroy on release
In reply to Bruce RichardsonAdd the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 ++++ drivers/raw/ioat/ioat_rawdev.c | 93 +++++++++++++++++++++++++++++- drivers/raw/ioat/rte_ioat_rawdev.h | 20 +++++++ 3 files changed, 121 insertions(+), 3 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -72,3 +72,14 @@ them to a suitable DPDK-supported kernel driver. When querying the status of the devices, they will appear under the category of "Misc (rawdev) devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be used to see the state of those devices alone. + +Device Probing and Initialization +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Once bound to a suitable kernel device driver, the HW devices will be found +as part of the PCI scan done at application initialization time. No vdev +parameters need to be passed to create or initialize the device. + +Once probed successfully, the device will appear as a ``rawdev``, that is a +"raw device type" inside DPDK, and can be accessed using APIs from the +``rte_rawdev`` library.
drivers/raw/ioat/ioat_rawdev.c +90 -3
111 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +20 -0
@@ -14,6 +14,9 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <rte_memory.h> +#include <rte_ioat_spec.h> +/** Name of the device driver */ #define IOAT_PMD_RAWDEV_NAME rawdev_ioat/** String reported as the device driver name by rte_rawdev_info_get() */
@@ -21,4 +24,21 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* @internal +* Structure representing a device instance +*/ +struct rte_ioat_rawdev { +struct rte_rawdev *rawdev; +volatile struct rte_ioat_registers *regs; +phys_addr_t status_addr; +phys_addr_t ring_addr; + +unsigned short ring_size; +struct rte_ioat_desc *desc_ring; + +/* to report completions, the device will write status back here */ + volatile uint64_t status __rte_cache_aligned; +}; +#endif
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
Quoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 7 lines
ioat_rawdev_create(const char *name, struct rte_pci_device *dev) {drivers/raw/ioat/ioat_rawdev.c +3
-2-RTE_SET_USED (name); -RTE_SET_USED (dev); +static const struct rte_rawdev_ops ioat_rawdev_ops = { + }; +
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 42
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev.c +10
-0+struct rte_rawdev *rawdev =NULL ; +struct rte_ioat_rawdev *ioat =NULL ; +int ret =0 ; +int retry =0 ; + +if (!name) { +IOAT_PMD_ERR ("Invalid name of the device!" ); + ret = -EINVAL ; + goto cleanup; + }
Is checking dev not necessary here?
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 50
Quoted 8 lines
drivers/raw/ioat/ioat_rawdev.c +8
-0+ +/* Allocate device structure */ + rawdev =rte_rawdev_pmd_allocate (name,sizeof (struct rte_ioat_rawdev ), + dev->device.numa_node); +if (rawdev ==NULL ) { +IOAT_PMD_ERR ("Unable to allocate raw device" ); + ret = -EINVAL ; + goto cleanup;
EINVAL is supposed to be used to indicate invalid arguments. Inability to allocate is not an "invalid arguments" condition. Does rte_rawdev_pmd_allocate() set its own errno value? If so, perhaps it would be worth passing it on? If not, perhaps -ENOMEM would be a better return value?
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
Quoted 21 lines · Anatoly
Quoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 11 lines
ioat_rawdev_create(const char *name, struct rte_pci_device *dev) {drivers/raw/ioat/ioat_rawdev.c +7
-2-RTE_SET_USED (name); -RTE_SET_USED (dev); +static const struct rte_rawdev_ops ioat_rawdev_ops = { + }; + +struct rte_rawdev *rawdev =NULL ; +struct rte_ioat_rawdev *ioat =NULL ; +int ret =0 ; +int retry =0 ;
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 42
Quoted 6 lines
drivers/raw/ioat/ioat_rawdev.c +6
-0+ +if (!name) { +IOAT_PMD_ERR ("Invalid name of the device!" ); + ret = -EINVAL ; + goto cleanup; + }Is checking
devnot necessary here?
No, the only place it's called from is already using the PCI device structure. I don't think the probe function can ever be called with a NULL parameter.
Quoted 6 lines
Quoted 6 lines
drivers/raw/ioat/ioat_rawdev.c +6
-0+ +/* Allocate device structure */ + rawdev =rte_rawdev_pmd_allocate (name,sizeof (struct rte_ioat_rawdev ), + dev->device.numa_node); +if (rawdev ==NULL ) { +IOAT_PMD_ERR ("Unable to allocate raw device" );
Quoted 9 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+ ret = -EINVAL; + goto cleanup;EINVAL is supposed to be used to indicate invalid arguments. Inability to allocate is not an "invalid arguments" condition. Does
rte_rawdev_pmd_allocate()set its own errno value? If so, perhaps it would be worth passing it on? If not, perhaps-ENOMEMwould be a better return value?
Fixing in v4.
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
In reply to Bruce RichardsonQuoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 63
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+ rawdev->driver_name = dev->device.driver->name; + + ioat = rawdev->dev_private; + ioat->rawdev = rawdev; + ioat->regs = dev->mem_resource[0 ].addr; + ioat->ring_size =0 ; + ioat->desc_ring =NULL ; + ioat->status_addr =rte_malloc_virt2iova (ioat) + + offsetof(struct rte_ioat_rawdev , status);
While reviewing other patch, i remembered that i've seen this here. You can't make any guarantees about IOVA addresses in rte_malloc-allocated memory. Are you sure you don't require IOVA-contiguous memory here?
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
Quoted 10 lines · Anatoly
Quoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 14 lines
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 63
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+ rawdev->driver_name = dev->device.driver->name; + + ioat = rawdev->dev_private; + ioat->rawdev = rawdev; + ioat->regs = dev->mem_resource[0 ].addr; + ioat->ring_size =0 ; + ioat->desc_ring =NULL ; + ioat->status_addr =rte_malloc_virt2iova (ioat) + + offsetof(struct rte_ioat_rawdev , status);While reviewing other patch, i remembered that i've seen this here. You can't make any guarantees about IOVA addresses in rte_malloc-allocated memory. Are you sure you don't require IOVA-contiguous memory here?
Presumably we can guarantee that for structures less than 1 page in size, this will work? I believe the device structure should be within that page limit.
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
Quoted 25 lines · Bruce Richardson
Quoted 10 lines · Anatoly
Quoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 14 lines
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 63
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+ rawdev->driver_name = dev->device.driver->name; + + ioat = rawdev->dev_private; + ioat->rawdev = rawdev; + ioat->regs = dev->mem_resource[0 ].addr; + ioat->ring_size =0 ; + ioat->desc_ring =NULL ; + ioat->status_addr =rte_malloc_virt2iova (ioat) + + offsetof(struct rte_ioat_rawdev , status);While reviewing other patch, i remembered that i've seen this here. You can't make any guarantees about IOVA addresses in rte_malloc-allocated memory. Are you sure you don't require IOVA-contiguous memory here?
Quoted 4 lines
Presumably we can guarantee that for structures less than 1 page in size, this will work? I believe the device structure should be within that page limit.
No, we can't. That would only be true if you were allocating IOVA-contiguous memory. Otherwise there's nothing stopping the allocator to allocate even a few kilobytes across page boundary.
You can only ever guarantee that *one cache line* will not cross the page boundary with rte_malloc. With rte_memzone and IOVA_CONTIG flag, you'll be able to guarantee IOVA-contiguousness in all cases (or allocation failure).
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
Quoted 31 lines · Anatoly
Quoted 25 lines · Bruce Richardson
Quoted 10 lines · Anatoly
Quoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 14 lines
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 63
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+ rawdev->driver_name = dev->device.driver->name; + + ioat = rawdev->dev_private; + ioat->rawdev = rawdev; + ioat->regs = dev->mem_resource[0 ].addr; + ioat->ring_size =0 ; + ioat->desc_ring =NULL ; + ioat->status_addr =rte_malloc_virt2iova (ioat) + + offsetof(struct rte_ioat_rawdev , status);While reviewing other patch, i remembered that i've seen this here. You can't make any guarantees about IOVA addresses in rte_malloc-allocated memory. Are you sure you don't require IOVA-contiguous memory here?
Quoted 4 lines
Presumably we can guarantee that for structures less than 1 page in size, this will work? I believe the device structure should be within that page limit.
Quoted 8 lines
No, we can't. That would only be true if you were allocating IOVA-contiguous memory. Otherwise there's nothing stopping the allocator to allocate even a few kilobytes across page boundary.
You can only ever guarantee that *one cache line* will not cross the page boundary with
rte_malloc. Withrte_memzoneandIOVA_CONTIGflag, you'll be able to guarantee IOVA-contiguousness in all cases (or allocation failure).
Ok, so I either need to move this field to the start of the structure, i.e. have offset zero, or else use contiguous allocation. Will fix in next version.
/Bruce
Re: [dpdk-dev] [PATCH v3 4/8] raw/ioat: create device on probe and destroy on release
Quoted 40 lines · Bruce Richardson
Quoted 31 lines · Anatoly
Quoted 25 lines · Bruce Richardson
Quoted 10 lines · Anatoly
Quoted 6 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 14 lines
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 63
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+ rawdev->driver_name = dev->device.driver->name; + + ioat = rawdev->dev_private; + ioat->rawdev = rawdev; + ioat->regs = dev->mem_resource[0 ].addr; + ioat->ring_size =0 ; + ioat->desc_ring =NULL ; + ioat->status_addr =rte_malloc_virt2iova (ioat) + + offsetof(struct rte_ioat_rawdev , status);While reviewing other patch, i remembered that i've seen this here. You can't make any guarantees about IOVA addresses in rte_malloc-allocated memory. Are you sure you don't require IOVA-contiguous memory here?
Quoted 4 lines
Presumably we can guarantee that for structures less than 1 page in size, this will work? I believe the device structure should be within that page limit.
Quoted 8 lines
No, we can't. That would only be true if you were allocating IOVA-contiguous memory. Otherwise there's nothing stopping the allocator to allocate even a few kilobytes across page boundary.
You can only ever guarantee that *one cache line* will not cross the page boundary with
rte_malloc. Withrte_memzoneandIOVA_CONTIGflag, you'll be able to guarantee IOVA-contiguousness in all cases (or allocation failure).
Quoted 6 lines
Ok, so I either need to move this field to the start of the structure, i.e. have offset zero, or else use contiguous allocation. Will fix in next version.
/Bruce
The latter is probably more explicit in intention, i'd rather the code not rely on details of rte_malloc implementation :)
Signature · 2 lines
Thanks, Anatoly
PATCH v3 5/8 raw/ioat: add device info function
In reply to Bruce RichardsonAdd in the "info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.
---
V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)
app/test/test_rawdev.c +11 -0
@@ -29,6 +29,17 @@ REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton);static int test_rawdev_selftest_ioat (void ) { +const int count =rte_rawdev_count (); +int i; + +for (i =0 ; i < count; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; +if (rte_rawdev_info_get (i, &info) ==0 && +strstr (info.driver_name,"ioat" ) !=NULL ) +return 0 ; + } + +printf ("No IOAT rawdev found, skipping tests\n" );return TEST_SKIPPED ; }
doc/guides/rawdevs/ioat_rawdev.rst +34 -0
@@ -83,3 +83,37 @@ parameters need to be passed to create or initialize the device. Once probed successfully, the device will appear as a ``rawdev``, that is a "raw device type" inside DPDK, and can be accessed using APIs from the ``rte_rawdev`` library. + +Using IOAT Rawdev Devices +-------------------------- + +To use the devices from an application, the rawdev API can be used, along +with definitions taken from the device-specific header file +``rte_ioat_rawdev.h``. This header is needed to get the definition of +structure parameters used by some of the rawdev APIs for IOAT rawdev +devices, as well as providing key functions for using the device for memory +copies. + +Getting Device Information +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Basic information about each rawdev device can be got using the +``rte_rawdev_info_get()`` API. For most applications, this API will be +needed to verify that the rawdev in question is of the expected type. For +example, the following code in ``test_ioat_rawdev.c`` is used to identify +the IOAT rawdev device for use for the tests: + +.. code-block:: C + + for (i = 0; i < count && !found; i++) { + struct rte_rawdev_info info = { .dev_private = NULL }; + found = (rte_rawdev_info_get(i, &info) == 0 && + strcmp(info.driver_name, + IOAT_PMD_RAWDEV_NAME_STR) == 0); + } + +When calling the ``rte_rawdev_info_get()`` API for an IOAT rawdev device, +the ``dev_private`` field in the ``rte_rawdev_info`` struct should either +be NULL, or else be set to point to a structure of type +``rte_ioat_rawdev_config``, in which case the size of the configured device +input ring will be returned in that structure.
drivers/raw/ioat/ioat_rawdev.c +11 -0
@@ -24,10 +24,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +} +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_info_get = ioat_dev_info_get, };struct rte_rawdev *rawdev =NULL ;
drivers/raw/ioat/rte_ioat_rawdev.h +11 -0
@@ -24,6 +24,17 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* Configuration structure for an ioat rawdev instance +* +* This structure is to be passed as the ".dev_private" parameter when +* calling the rte_rawdev_get_info() and rte_rawdev_configure() APIs on +* an ioat rawdev instance. +*/ +struct rte_ioat_rawdev_config { +unsigned short ring_size; +}; +/** * @internal * Structure representing a device instance
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 5/8] raw/ioat: add device info function
Quoted 22 lines · Bruce Richardson
Add in the "
info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.Signed-off-by Bruce Richardson <[email protected]>---
V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)app/test/test_rawdev.c +0
-0@@ -29,6 +29,17 @@ REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton); static int
app/test/test_rawdev.c REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton); · line 39
Quoted 10 lines
app/test/test_rawdev.c +8
-0test_rawdev_selftest_ioat (void) { +const int count =rte_rawdev_count (); +int i; + +for (i =0 ; i < count; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; +if (rte_rawdev_info_get (i, &info) ==0 && +strstr (info.driver_name,"ioat" ) !=NULL ) +return 0 ;
TEST_SUCCESS? Also, didn't "ioat" have a macro for its name?
Quoted 19 lines
app/test/test_rawdev.c +3
-0+ } + +printf ("No IOAT rawdev found, skipping tests\n" );return TEST_SKIPPED ; }doc/guides/rawdevs/ioat_rawdev.rst +5
-0@@ -83,3 +83,37 @@ parameters need to be passed to create or initialize the device. Once probed successfully, the device will appear as a ``rawdev``, that is a "raw device type" inside DPDK, and can be accessed using APIs from the ``rte_rawdev`` library. + +Using IOAT Rawdev Devices +-------------------------- + +To use the devices from an application, the rawdev API can be used, along
doc/guides/rawdevs/ioat_rawdev.rst parameters need to be passed to create or initialize the device. · line 100
Quoted 10 lines
doc/guides/rawdevs/ioat_rawdev.rst +10
-0+with definitions taken from the device-specific header file +``rte_ioat_rawdev.h``. This header is needed to get the definition of +structure parameters used by some of the rawdev APIs for IOAT rawdev +devices, as well as providing key functions for using the device for memory +copies. + +Getting Device Information +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Basic information about each rawdev device can be got using the
According to certain unreliable sources [1], usage of "can be got" has not been used since the end of 19th century. I didn't think you were *that* old, Bruce!
doc/guides/rawdevs/ioat_rawdev.rst parameters need to be passed to create or initialize the device. · line 113
Quoted 13 lines
doc/guides/rawdevs/ioat_rawdev.rst +13
-0+``rte_rawdev_info_get()`` API. For most applications, this API will be +needed to verify that the rawdev in question is of the expected type. For +example, the following code in ``test_ioat_rawdev.c`` is used to identify +the IOAT rawdev device for use for the tests: + +.. code-block:: C + + for (i = 0; i < count && !found; i++) { + struct rte_rawdev_info info = { .dev_private = NULL }; + found = (rte_rawdev_info_get(i, &info) == 0 && + strcmp(info.driver_name, + IOAT_PMD_RAWDEV_NAME_STR) == 0); + }
The code here doesn't match the actual implementation in the autotest.
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 5/8] raw/ioat: add device info function
Quoted 27 lines · Anatoly
Quoted 26 lines · Bruce Richardson
Add in the "
info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.Signed-off-by Bruce Richardson <[email protected]>---
V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)app/test/test_rawdev.c +2
-0@@ -29,6 +29,17 @@ REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton);static int test_rawdev_selftest_ioat (void ) { +const int count =rte_rawdev_count (); +int i;
Quoted 9 lines
app/test/test_rawdev.c REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton); · line 39
Quoted 6 lines
app/test/test_rawdev.c +6
-0+ +for (i =0 ; i < count; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; +if (rte_rawdev_info_get (i, &info) ==0 && +strstr (info.driver_name,"ioat" ) !=NULL ) +return 0 ;
TEST_SUCCESS? Also, didn't "ioat" have a macro for its name?
Ack on the TEST_SUCCESS part.
For the seond point, yes, there is a macro for the name, but to get it one has to include the header file for the driver, which will not be available if the driver is not being built. Therefore, I just use a basic check for the "ioat" name, rather than cluttering up the code with lots of conditionals for handling cases where the driver won't be built.
/Bruce
PATCH v3 6/8 raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonAllow initializing a driver instance. Include selftest to validate these functions.
---
V3: don't add a new descriptor format struct, reuse from rte_ioat_spec.h
V2: test cases placed in self-test routine
--- app/test/test_rawdev.c | 2 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 ++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 78 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 4 +- 7 files changed, 158 insertions(+), 3 deletions(-) create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c
app/test/test_rawdev.c +1 -1
@@ -36,7 +36,7 @@ test_rawdev_selftest_ioat(void)struct rte_rawdev_info info = { .dev_private =NULL };if (rte_rawdev_info_get (i, &info) ==0 &&strstr (info.driver_name,"ioat" ) !=NULL ) -return 0 ; +return rte_rawdev_selftest (i); }printf ("No IOAT rawdev found, skipping tests\n" );
doc/guides/rawdevs/ioat_rawdev.rst +32 -0
@@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API.
drivers/raw/ioat/Makefile +1 -0
@@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h
drivers/raw/ioat/ioat_rawdev.c +78 -0
95 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c new file +41 -0
41 lines not on this page · raw message
drivers/raw/ioat/meson.build +2 -1
@@ -2,7 +2,8 @@ # Copyright 2019 Intel Corporation build = dpdk_conf.has('RTE_ARCH_X86') -sources = files('ioat_rawdev.c') +sources = files('ioat_rawdev.c', + 'ioat_rawdev_test.c') deps += ['rawdev', 'bus_pci'] install_headers('rte_ioat_rawdev.h',
drivers/raw/ioat/rte_ioat_rawdev.h +3 -1
@@ -14,6 +14,7 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <x86intrin.h> #include <rte_memory.h> #include <rte_ioat_spec.h>
@@ -46,7 +47,8 @@ struct rte_ioat_rawdev {phys_addr_t ring_addr;unsigned short ring_size; -struct rte_ioat_desc *desc_ring; +struct rte_ioat_generic_hw_desc *desc_ring; +__m128i *hdls;/* completion handles for returning to user */ /* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned;
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 6/8] raw/ioat: add configure, start and stop functions
Quoted 25 lines · Bruce Richardson
Allow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]>---
V3: don't add a new descriptor format struct, reuse from
rte_ioat_spec.h
V2: test cases placed in self-test routine--- app/test/test_rawdev.c | 2 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 ++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 78 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 4 +- 7 files changed, 158 insertions(+), 3 deletions(-)create mode 100644
drivers/raw/ioat/ioat_rawdev_test.capp/test/test_rawdev.c +0
-0
app/test/test_rawdev.c test_rawdev_selftest_ioat(void) · line 39
Quoted 5 lines
app/test/test_rawdev.c +1
-1@@ -36,7 +36,7 @@ test_rawdev_selftest_ioat(void)struct rte_rawdev_info info = { .dev_private =NULL };if (rte_rawdev_info_get (i, &info) ==0 &&strstr (info.driver_name,"ioat" ) !=NULL ) -return 0 ; +return rte_rawdev_selftest (i);
Even though it doesn't matter in practice, technically, we can't pass a raw return value to the test caller. It should be TEST_SUCCESS or TEST_FAILURE.
Quoted 23 lines
app/test/test_rawdev.c +0
-0}printf ("No IOAT rawdev found, skipping tests\n" );doc/guides/rawdevs/ioat_rawdev.rst +12
-0@@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096.
<snip>
Quoted 7 lines
doc/guides/rawdevs/ioat_rawdev.rst +7
-0+ if (params->ring_size > 4096 || params->ring_size < 64 || + !rte_is_power_of_2(params->ring_size)) + return -EINVAL; + + ioat->ring_size = params->ring_size; + if (ioat->desc_ring != NULL) { + rte_free(ioat->desc_ring);
doc/guides/rawdevs/ioat_rawdev.rst
Quoted 10 lines
doc/guides/rawdevs/ioat_rawdev.rst +10
-0+ ioat->desc_ring = NULL; + } + + /* allocate one block of memory for both descriptors + * and completion handles. + */ + ioat->desc_ring = rte_zmalloc_socket(NULL, + (DESC_SZ + COMPLETION_SZ) * ioat->ring_size, + 0, /* alignment, default to 64Byte */ + dev->device->numa_node);
Using rte_zmalloc for hardware structures seems suspect. Do you not need IOVA-contiguous memory here?
doc/guides/rawdevs/ioat_rawdev.rst
Quoted 11 lines
doc/guides/rawdevs/ioat_rawdev.rst +11
-0+ if (ioat->desc_ring == NULL) + return -ENOMEM; + ioat->hdls = (void *)&ioat->desc_ring[ioat->ring_size]; + + ioat->ring_addr = rte_malloc_virt2iova(ioat->desc_ring); + + /* configure descriptor ring - each one points to next */ + for (i = 0; i < ioat->ring_size; i++) { + ioat->desc_ring[i].next = ioat->ring_addr + + (((i + 1) % ioat->ring_size) * DESC_SZ); + }
OK, this *definitely* looks suspect :) with rte_zmalloc(), there's no guarantee that the entire allocated block resides on the same physical page, so you can't assume IOVA addresses will be contiguous either, unless you only intend to operate in IOVA as VA mode (which i didn't notice).
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 6/8] raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonQuoted 10 lines
-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 11:41 AM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 6/8] raw/ioat: add configure, start and stop functions
<snip>
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 119
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev.c +6
-0+int ioat_rawdev_test (uint16_t dev_id); +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_configure = ioat_dev_configure, + .dev_start = ioat_dev_start, + .dev_stop = ioat_dev_stop, .dev_info_get = ioat_dev_info_get, + .dev_selftest = ioat_rawdev_test,
Build fail for ./devtools/test-build.sh x86_64-native-linux-gcc+next+shared
/drivers/raw/ioat/ioat_rawdev.c: In function 'ioat_rawdev_create':drivers/raw/ioat/ioat_rawdev.c:163:20: error: initialization of 'int (*)(void)' from incompatible pointer type 'int (*)(uint16_t)' {aka 'int (*)(short unsigned int)'} [-Werror=incompatible-pointer-types]
163 | .dev_selftest = ioat_rawdev_test,
Thanks,
Reshma
Re: [dpdk-dev] [PATCH v3 6/8] raw/ioat: add configure, start and stop functions
Quoted 13 lines · Reshma
Quoted 10 lines
-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 11:41 AM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 6/8] raw/ioat: add configure, start and stop functions<snip>
Quoted 16 lines
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 119
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev.c +6
-0+int ioat_rawdev_test (uint16_t dev_id); +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_configure = ioat_dev_configure, + .dev_start = ioat_dev_start, + .dev_stop = ioat_dev_stop, .dev_info_get = ioat_dev_info_get, + .dev_selftest = ioat_rawdev_test,Build fail for
./devtools/test-build.shx86_64-native-linux-gcc+next+shared
/drivers/raw/ioat/ioat_rawdev.c: In function 'ioat_rawdev_create':drivers/raw/ioat/ioat_rawdev.c:163:20: error: initialization of 'int (*)(void)' from incompatible pointer type 'int (*)(uint16_t)' {aka 'int (*)(short unsigned int)'} [-Werror=incompatible-pointer-types]
163 | .dev_selftest =ioat_rawdev_test,
Apologies, I forgot to put in the cover letter that this depends upon the rawdev patchset I previously upstreamed:
PATCH v3 7/8 raw/ioat: add statistics functions
In reply to Bruce RichardsonAdd stats functions to track what is happening in the driver, and put unit tests to check those.
---
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 14 +++++++++ drivers/raw/ioat/ioat_rawdev.c | 44 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 39 +++++++++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 6 ++++ 4 files changed, 103 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +14 -0
@@ -149,3 +149,17 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. + +Querying Device Statistics +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The statistics from the IOAT rawdev device can be got via the xstats +functions in the ``rte_rawdev`` library, i.e. +``rte_rawdev_xstats_names_get()``, ``rte_rawdev_xstats_get()`` and +``rte_rawdev_xstats_by_name_get``. The statistics returned for each device +instance are: + +* ``failed_enqueues`` +* ``successful_enqueues`` +* ``copies_started`` +* ``copies_completed``
drivers/raw/ioat/ioat_rawdev.c +44 -0
62 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c +39 -0
@@ -2,6 +2,7 @@* Copyright(c) 2019 Intel Corporation */ +#include <inttypes.h> #include "rte_rawdev.h" #include "rte_ioat_rawdev.h"
@@ -13,6 +14,11 @@ ioat_rawdev_test(uint16_t dev_id)#define IOAT_TEST_RINGSIZE 512struct rte_ioat_rawdev_config p = { .ring_size =-1 };struct rte_rawdev_info info = { .dev_private = &p }; +struct rte_rawdev_xstats_name *snames =NULL ; +uint64_t *stats =NULL ; +unsigned int *ids =NULL ; +unsigned int nb_xstats; +unsigned int i;rte_rawdev_info_get (dev_id, &info);if (p.ring_size !=0 ) {
@@ -37,5 +43,38 @@ ioat_rawdev_test(uint16_t dev_id)printf ("Error with rte_rawdev_start()\n" );return -1 ; } + +/* allocate memory for xstats names and values */ + nb_xstats =rte_rawdev_xstats_names_get (dev_id,NULL ,0 ); + + snames =malloc (sizeof (*snames) * nb_xstats); +if (snames ==NULL ) { +printf ("Error allocating xstat names memory\n" ); +return -1 ; + } +rte_rawdev_xstats_names_get (dev_id, snames, nb_xstats); + + ids =malloc (sizeof (*ids) * nb_xstats); +if (ids ==NULL ) { +printf ("Error allocating xstat ids memory\n" ); +return -1 ; + } +for (i =0 ; i < nb_xstats; i++) + ids[i] = i; + + stats =malloc (sizeof (*stats) * nb_xstats); +if (stats ==NULL ) { +printf ("Error allocating xstat memory\n" ); +return -1 ; + } + +rte_rawdev_xstats_get (dev_id, ids, stats, nb_xstats); +for (i =0 ; i < nb_xstats; i++) +printf ("%s: %" PRIu64" " , snames[i].name, stats[i]); +printf ("\n" ); + +free (snames); +free (stats); +free (ids);return 0 ; }
drivers/raw/ioat/rte_ioat_rawdev.h +6 -0
@@ -50,6 +50,12 @@ struct rte_ioat_rawdev {struct rte_ioat_generic_hw_desc *desc_ring;__m128i *hdls;/* completion handles for returning to user */ +/* some statistics for tracking, if added/changed update xstats fns*/ + uint64_t enqueue_failed __rte_cache_aligned; +uint64_t enqueued; +uint64_t started; +uint64_t completed; +/* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; };
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 7/8] raw/ioat: add statistics functions
Quoted 5 lines · Bruce Richardson
Add stats functions to track what is happening in the driver, and put unit tests to check those.
Signed-off-by Bruce Richardson <[email protected]>---
<snip>
Quoted 4 lines
drivers/raw/ioat/ioat_rawdev_test.c +4
-0+/* allocate memory for xstats names and values */ + nb_xstats =rte_rawdev_xstats_names_get (dev_id,NULL ,0 ); + + snames =malloc (sizeof (*snames) * nb_xstats);
drivers/raw/ioat/ioat_rawdev_test.c ioat_rawdev_test(uint16_t dev_id) · line 60
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev_test.c +10
-0+if (snames ==NULL ) { +printf ("Error allocating xstat names memory\n" ); +return -1 ; + } +rte_rawdev_xstats_names_get (dev_id, snames, nb_xstats); + + ids =malloc (sizeof (*ids) * nb_xstats); +if (ids ==NULL ) { +printf ("Error allocating xstat ids memory\n" ); +return -1 ;
Leaking snames here.
drivers/raw/ioat/ioat_rawdev_test.c ioat_rawdev_test(uint16_t dev_id) · line 68
Quoted 8 lines
drivers/raw/ioat/ioat_rawdev_test.c +8
-0+ } +for (i =0 ; i < nb_xstats; i++) + ids[i] = i; + + stats =malloc (sizeof (*stats) * nb_xstats); +if (stats ==NULL ) { +printf ("Error allocating xstat memory\n" ); +return -1 ;
Leaking snames and ids here. Perhaps a goto and a ret value wouldn't hurt :)
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v3 7/8] raw/ioat: add statistics functions
In reply to Bruce RichardsonQuoted 6 lines
-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 11:41 AM
To: [email protected]
<snip>
Quoted 7 lines
drivers/raw/ioat/ioat_rawdev.c +7
-0+static int +ioat_xstats_get (const struct rte_rawdev *dev,const unsigned int ids[], +uint64_t values[],unsigned int n) +{ +const struct rte_ioat_rawdev *ioat = dev->dev_private; +unsigned int i; +
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 131
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev.c +10
-0+for (i =0 ; i < n; i++) { +switch (ids[i]) { +case 0 : values[i] = ioat->enqueue_failed;break ; +case 1 : values[i] = ioat->enqueued;break ; +case 2 : values[i] = ioat->started;break ; +case 3 : values[i] = ioat->completed;break ; +default : values[i] =0 ;break ; + } + } +return n;
Should this return n or how many actually filled ?
Thanks,
Reshma
Re: [dpdk-dev] [PATCH v3 7/8] raw/ioat: add statistics functions
Quoted 21 lines · Reshma
Quoted 6 lines
-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 11:41 AM
To: [email protected]<snip>
Quoted 11 lines
drivers/raw/ioat/ioat_rawdev.c +11
-0+static int +ioat_xstats_get (const struct rte_rawdev *dev,const unsigned int ids[], +uint64_t values[],unsigned int n) +{ +const struct rte_ioat_rawdev *ioat = dev->dev_private; +unsigned int i; + +for (i =0 ; i < n; i++) { +switch (ids[i]) { +case 0 : values[i] = ioat->enqueue_failed;break ; +case 1 : values[i] = ioat->enqueued;break ;
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 131
Quoted 6 lines
drivers/raw/ioat/ioat_rawdev.c +6
-0+case 2 : values[i] = ioat->started;break ; +case 3 : values[i] = ioat->completed;break ; +default : values[i] =0 ;break ; + } + } +return n;Should this return n or how many actually filled ?
They should both be "n", since we never break out of the loop early.
/Bruce
PATCH v3 8/8 raw/ioat: add local API to perform copies
In reply to Bruce RichardsonAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
---
V3: updated to use descriptor format in rte_ioat_spec.h
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 100 ++++++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev_test.c | 161 +++++++++++++++++++++++++- drivers/raw/ioat/meson.build | 2 +- drivers/raw/ioat/rte_ioat_rawdev.h | 169 +++++++++++++++++++++++++++- 5 files changed, 428 insertions(+), 5 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +100 -0
@@ -150,6 +150,106 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. +Performing Data Copies +~~~~~~~~~~~~~~~~~~~~~~~ + +To perform data copies using IOAT rawdev devices, the functions +``rte_ioat_enqueue_copy()`` and ``rte_ioat_do_copies()`` should be used. +Once copies have been completed, the completion will be reported back when +the application calls ``rte_ioat_completed_copies()``. + +The ``rte_ioat_enqueue_copy()`` function enqueues a single copy to the +device ring for copying at a later point. The parameters to that function +include the physical addresses of both the source and destination buffers, +as well as two "handles" to be returned to the user when the copy is +completed. These handles can be arbitrary values, but two are provided so +that the library can track handles for both source and destination on +behalf of the user, e.g. virtual addresses for the buffers, or mbuf +pointers if packet data is being copied. + +While the ``rte_ioat_enqueue_copy()`` function enqueues a copy operation on +the device ring, the copy will not actually be performed until after the +application calls the ``rte_ioat_do_copies()`` function. This function +informs the device hardware of the elements enqueued on the ring, and the +device will begin to process them. It is expected that, for efficiency +reasons, a burst of operations will be enqueued to the device via multiple +enqueue calls between calls to the ``rte_ioat_do_copies()`` function. + +The following code from ``test_ioat_rawdev.c`` demonstrates how to enqueue +a burst of copies to the device and start the hardware processing of them: + +.. code-block:: C + + struct rte_mbuf *srcs[32], *dsts[32]; + unsigned int j; + + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data; + + srcs[i] = rte_pktmbuf_alloc(pool); + dsts[i] = rte_pktmbuf_alloc(pool); + srcs[i]->data_len = srcs[i]->pkt_len = length; + dsts[i]->data_len = dsts[i]->pkt_len = length; + src_data = rte_pktmbuf_mtod(srcs[i], char *); + + for (j = 0; j < length; j++) + src_data[j] = rand() & 0xFF; + + if (rte_ioat_enqueue_copy(dev_id, + srcs[i]->buf_iova + srcs[i]->data_off, + dsts[i]->buf_iova + dsts[i]->data_off, + length, + (uintptr_t)srcs[i], + (uintptr_t)dsts[i], + 0 /* nofence */) != 1) { + printf("Error with rte_ioat_enqueue_copy for buffer %u\n", + i); + return -1; + } + } + rte_ioat_do_copies(dev_id); + +To retrieve information about completed copies, the API +``rte_ioat_completed_copies()`` should be used. This API will return to the +application a set of completion handles passed in when the relevant copies +were enqueued. + +The following code from ``test_ioat_rawdev.c`` shows the test code +retrieving information about the completed copies and validating the data +is correct before freeing the data buffers using the returned handles: + +.. code-block:: C + + if (rte_ioat_completed_copies(dev_id, 64, (void *)completed_src, + (void *)completed_dst) != RTE_DIM(srcs)) { + printf("Error with rte_ioat_completed_copies\n"); + return -1; + } + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data, *dst_data; + + if (completed_src[i] != srcs[i]) { + printf("Error with source pointer %u\n", i); + return -1; + } + if (completed_dst[i] != dsts[i]) { + printf("Error with dest pointer %u\n", i); + return -1; + } + + src_data = rte_pktmbuf_mtod(srcs[i], char *); + dst_data = rte_pktmbuf_mtod(dsts[i], char *); + for (j = 0; j < length; j++) + if (src_data[j] != dst_data[j]) { + printf("Error with copy of packet %u, byte %u\n", + i, j); + return -1; + } + rte_pktmbuf_free(srcs[i]); + rte_pktmbuf_free(dsts[i]); + } + + Querying Device Statistics ~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/raw/ioat/Makefile +1 -0
@@ -12,6 +12,7 @@ CFLAGS += $(WERROR_FLAGS) LDLIBS += -lrte_eal -lrte_rawdev LDLIBS += -lrte_pci -lrte_bus_pci +LDLIBS += -lrte_mbuf -lrte_mempool # library version LIBABIVER := 1
drivers/raw/ioat/ioat_rawdev_test.c +158 -3
189 lines not on this page · raw message
drivers/raw/ioat/meson.build +1 -1
@@ -4,7 +4,7 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c', 'ioat_rawdev_test.c') -deps += ['rawdev', 'bus_pci'] +deps += ['rawdev', 'bus_pci', 'mbuf'] install_headers('rte_ioat_rawdev.h', 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_rawdev.h +168 -1
185 lines not on this page · raw message
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v3 8/8] raw/ioat: add local API to perform copies
Quoted 27 lines · Bruce Richardson
Add local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
Signed-off-by Bruce Richardson <[email protected]>---
V3: updated to use descriptor format in
rte_ioat_spec.h
V2: test cases moved to self-test routine--- doc/guides/rawdevs/ioat_rawdev.rst | 100 ++++++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev_test.c | 161 +++++++++++++++++++++++++- drivers/raw/ioat/meson.build | 2 +- drivers/raw/ioat/rte_ioat_rawdev.h | 169 +++++++++++++++++++++++++++- 5 files changed, 428 insertions(+), 5 deletions(-)doc/guides/rawdevs/ioat_rawdev.rst +1
-0@@ -150,6 +150,106 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. +Performing Data Copies
doc/guides/rawdevs/ioat_rawdev.rst The following code shows how the device is configured in · line 163
Quoted 10 lines
doc/guides/rawdevs/ioat_rawdev.rst +10
-0+~~~~~~~~~~~~~~~~~~~~~~~ + +To perform data copies using IOAT rawdev devices, the functions +``rte_ioat_enqueue_copy()`` and ``rte_ioat_do_copies()`` should be used. +Once copies have been completed, the completion will be reported back when +the application calls ``rte_ioat_completed_copies()``. + +The ``rte_ioat_enqueue_copy()`` function enqueues a single copy to the +device ring for copying at a later point. The parameters to that function +include the physical addresses of both the source and destination buffers,
Physical or IOVA?
Quoted 8 lines
doc/guides/rawdevs/ioat_rawdev.rst +8
-0+as well as two "handles" to be returned to the user when the copy is +completed. These handles can be arbitrary values, but two are provided so +that the library can track handles for both source and destination on +behalf of the user, e.g. virtual addresses for the buffers, or mbuf +pointers if packet data is being copied. + +While the ``rte_ioat_enqueue_copy()`` function enqueues a copy operation on +the device ring, the copy will not actually be performed until after the
<snip>
doc/guides/rawdevs/ioat_rawdev.rst
Quoted 12 lines
doc/guides/rawdevs/ioat_rawdev.rst +8
-0+ rte_mempool_free(pool); free(snames); free(stats); free(ids); return 0; + +err: + rte_mempool_free(pool); + free(snames); + free(stats); + free(ids); + return -1;
The goto is added a commit too late :D
Otherwise, LGTM
Signature · 2 lines
Thanks, Anatoly
Quoted 9 lines
-----Original Message-----
From: dev [mailto:[email protected]] On Behalf Of Bruce Richardson
Sent: Thursday, June 27, 2019 11:41 AM
To: [email protected]
Cc: [email protected]; [email protected]; Richardson, Bruce <[email protected]>
Subject: [dpdk-dev] [PATCH v3 0/8] raw/ioat: driver for Intel QuickData Technology
Quoted 8 lines
This patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
Applied (release notes conflict - but that's par for the course close to integration) Build each patch with meson (git rebase -x is awesome :)
Ran unit tests:
$ ioat_rawdev_probe(): Init 0000:00:04.0 on NUMA node 0 $ failed_enqueues: 0 successful_enqueues: 3300 copies_started: 3300 copies_completed: 3300
All seems fine to me;
v4 cover raw/ioat: driver for Intel QuickData Technology
In reply to Bruce RichardsonThis patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
V4:
* changed memory management to use contiguous memzones instead of malloc memory
* added missing device ids for BDX platforms
* other misc cleanup following review (see individual patches logs)
V3:
* removed DPDK-specific structure for the descriptor format and reused the structure in the imported file rte_ioat_spec.h
V2:
* moved tests to rawdev selftest function
* some checkpatch and other small cleanups
* added extra documentation details on supported hardware
* aligned the changes to dpdk-devbind with the changes in the NTB set for consistency
Bruce Richardson (9):
rawdev: allow devices to skip extra memory allocation raw/ioat: add initial support for ioat rawdev driver usertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
MAINTAINERS | 5 + app/test/test_rawdev.c | 20 ++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 265 +++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 + drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 31 ++ drivers/raw/ioat/ioat_rawdev.c | 356 ++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 235 +++++++++++++ drivers/raw/ioat/meson.build | 10 + drivers/raw/ioat/rte_ioat_rawdev.h | 234 +++++++++++++ drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- lib/librte_rawdev/rte_rawdev.c | 11 +- mk/rte.app.mk | 1 + usertools/dpdk-devbind.py | 10 + 21 files changed, 1502 insertions(+), 6 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_ioat_spec.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
Signature · 1 line
2.21.0
PATCH v4 1/9 rawdev: allow devices to skip extra memory allocation
Some device drivers want to allocate their own private memory, and should be allowed to do so. Therefore skip memory allocation and associated error checks if zero-length private memory is requested.
While adjusting the code for new indent level, fix incorrect error message.
--- lib/librte_rawdev/rte_rawdev.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
lib/librte_rawdev/rte_rawdev.c +6 -5
@@ -496,16 +496,17 @@ rte_rawdev_pmd_allocate(const char *name, size_t dev_priv_size, int socket_id) rawdev = &rte_rawdevs[dev_id]; - rawdev->dev_private =rte_zmalloc_socket ("rawdev private" , +if (dev_priv_size >0 ) { + rawdev->dev_private =rte_zmalloc_socket ("rawdev private" , dev_priv_size,RTE_CACHE_LINE_SIZE , socket_id); -if (!rawdev->dev_private) { -RTE_RDEV_ERR ("Unable to allocate memory to Skeleton dev" ); -return NULL ; +if (!rawdev->dev_private) { +RTE_RDEV_ERR ("Unable to allocate memory for rawdev" ); +return NULL ; + } } - rawdev->dev_id = dev_id; rawdev->socket_id = socket_id; rawdev->started =0 ;
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v4 1/9] rawdev: allow devices to skip extra memory allocation
Re: [dpdk-dev] [PATCH v4 1/9] rawdev: allow devices to skip extra memory allocation
In reply to Bruce RichardsonQuoted 10 lines
-----Original Message-----
From: Bruce Richardson <[email protected]>
Sent: Monday, July 1, 2019 9:26 PM
To: [email protected]
Cc: [email protected]; [email protected]; [email protected]; Bruce Richardson <[email protected]>; Shreyansh Jain <[email protected]>; Hemant Agrawal <[email protected]>
Subject: [PATCH v4 1/9] rawdev: allow devices to skip extra memory allocation
Quoted 5 lines
Some device drivers want to allocate their own private memory, and should
be allowed to do so. Therefore skip memory allocation and associated error
checks if zero-length private memory is requested.
Agree with this - rawdev was intended for flexibility and this (allowing them their own memory) is definitely better way ahead. Thanks for proposing. But, I think the kind of caveat should also be added to the header declaring this function:
Probably something like this:
--->8--- lib/librte_rawdev/rte_rawdev_pmd.h ---
/**
* Allocates a new rawdev slot for an raw device and returns the pointer
* to that slot for the driver to use.
*
* @param name
* Unique identifier name for each device
* @param dev_private_size
* Private data allocated within rte_rawdev object.
* <b>Set to 0 to disable internal allocation and allow for self-allocation</b>
* @param socket_id
* Socket to allocate resources on.
* @return
* - Slot in the rte_dev_devices array for a new device;
*/
struct rte_rawdev *
rte_rawdev_pmd_allocate(const char *name, size_t dev_private_size,
int socket_id);
--->8---
Quoted 8 lines
While adjusting the code for new indent level, fix incorrect error message.
Cc Shreyansh Jain <[email protected]>Cc Hemant Agrawal <[email protected]>Signed-off-by Bruce Richardson <[email protected]>---
If you can update the header, please use my ACK in next version.
[...]
Re: [dpdk-dev] [PATCH v4 1/9] rawdev: allow devices to skip extra memory allocation
Quoted 45 lines · Shreyansh Jain
Quoted 10 lines
-----Original Message-----
From: Bruce Richardson <[email protected]>
Sent: Monday, July 1, 2019 9:26 PM
To: [email protected]
Cc: [email protected]; [email protected]; [email protected]; Bruce Richardson <[email protected]>; Shreyansh Jain <[email protected]>; Hemant Agrawal <[email protected]>
Subject: [PATCH v4 1/9] rawdev: allow devices to skip extra memory allocationQuoted 5 lines
Some device drivers want to allocate their own private memory, and should
be allowed to do so. Therefore skip memory allocation and associated error
checks if zero-length private memory is requested.Agree with this - rawdev was intended for flexibility and this (allowing them their own memory) is definitely better way ahead. Thanks for proposing. But, I think the kind of caveat should also be added to the header declaring this function:
Probably something like this:
--->8---
lib/librte_rawdev/rte_rawdev_pmd.h---
/**
* Allocates a new rawdev slot for an raw device and returns the pointer
* to that slot for the driver to use.
*
* @param name
* Unique identifier name for each device
* @paramdev_private_size
* Private data allocated withinrte_rawdevobject.
* <b>Set to 0 to disable internal allocation and allow for self-allocation</b>
* @paramsocket_id
* Socket to allocate resources on.
* @return
* - Slot in therte_dev_devicesarray for a new device;
*/
structrte_rawdev*
rte_rawdev_pmd_allocate(const char *name,size_tdev_private_size,int socket_id);--->8---
Quoted 4 lines
While adjusting the code for new indent level, fix incorrect error message.
Quoted 9 lines
Quoted 4 lines
Cc Shreyansh Jain <[email protected]>Cc Hemant Agrawal <[email protected]>Signed-off-by Bruce Richardson <[email protected]>---
If you can update the header, please use my ACK in next version.
Acked-by Shreyansh Jain <[email protected]>
Thanks Shreyansh, will do.
/Bruce
PATCH v4 2/9 raw/ioat: add initial support for ioat rawdev driver
In reply to Bruce RichardsonAdd stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
---
V4: add in missing device IDs
V3: remove unneeded whitespace change
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
--- MAINTAINERS | 5 + app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 +++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 ++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++ drivers/raw/ioat/ioat_rawdev.c | 113 ++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 +++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 278 insertions(+), 1 deletion(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +5 -0
@@ -1074,6 +1074,11 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/
app/test/test_rawdev.c +8 -0
@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }REGISTER_TEST_COMMAND (rawdev_autotest, test_rawdev_selftest_skeleton); + +static int +test_rawdev_selftest_ioat (void ) +{ +return TEST_SKIPPED ; +} + +REGISTER_TEST_COMMAND (ioat_rawdev_autotest, test_rawdev_selftest_ioat);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=n
config/common_base +5 -0
@@ -747,6 +747,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n # CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y +# +# Compile PMD for Intel IOAT raw device +# +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -54,3 +54,4 @@ CONFIG_RTE_LIBRTE_QEDE_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -56,3 +56,4 @@ CONFIG_RTE_LIBRTE_ENIC_PMD=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +1 -0
@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +63 -0
63 lines not on this page · raw message
doc/guides/rel_notes/release_19_08.rst +11 -0
@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================
@@ -118,6 +120,15 @@ New Features Added telemetry mode to l3fwd-power application to report application level busyness, empty and full polls of rte_eth_rx_burst(). +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of + Intel\ |reg| I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_, + allows data copies to be done by hardware instead + of via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items -------------
drivers/raw/Makefile +1 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright(c) 2019 Intel Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +113 -0
113 lines not on this page · raw message
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright 2019 Intel Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior notice +*/ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() */ +#define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ +#define IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +3 -1
@@ -1,7 +1,9 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat', + 'skeleton_rawdev'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +1 -0
@@ -306,6 +306,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev _LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke endif # CONFIG_RTE_LIBRTE_IFPGA_BUS +_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev endif # CONFIG_RTE_LIBRTE_RAWDEV endif # !CONFIG_RTE_BUILD_SHARED_LIBS
Signature · 1 line
2.21.0
PATCH v4 3/9 usertools/dpdk-devbind.py: add support for IOAT devices
In reply to Bruce RichardsonIn order to allow binding/unbinding of devices for use by the ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.
---
V2: rather than adding a DMA section, add to "misc (rawdev)" section where
other device types, e.g. ntb can also do.
NOTE: this set largely overlaps with the equivalent changes made by the patchset adding NTB support [1]. Since it's unclear which set will be added first, this set is based off the latest head.
[1] http://patches.dpdk.org/patch/55127/
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++++++++++ usertools/dpdk-devbind.py | 10 ++++++++++ 2 files changed, 21 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -61,3 +61,14 @@ configurations. For builds using ``meson`` and ``ninja``, the driver will be built when the target platform is x86-based. + +Device Setup +------------- + +The Intel\ |reg| QuickData Technology HW devices will need to be bound to a +user-space IO driver for use. The script ``dpdk-devbind.py`` script +included with DPDK can be used to view the state of the devices and to bind +them to a suitable DPDK-supported kernel driver. When querying the status +of the devices, they will appear under the category of "Misc (rawdev) +devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be +used to see the state of those devices alone.
usertools/dpdk-devbind.py +10 -0
@@ -36,11 +36,17 @@ octeontx2_npa = {'Class': '08', 'Vendor': '177d', 'Device': 'a0fb,a0fc', 'SVendor': None, 'SDevice': None} +intel_ioat_bdw = {'Class': '08', 'Vendor': '8086', 'Device': '6f20,6f21,6f22,6f23,6f24,6f25,6f26,6f27,6f2e,6f2f', + 'SVendor': None, 'SDevice': None} +intel_ioat_skx = {'Class': '08', 'Vendor': '8086', 'Device': '2021', + 'SVendor': None, 'SDevice': None} + network_devices = [network_class, cavium_pkx, avp_vnic, ifpga_class] crypto_devices = [encryption_class, intel_processor_class] eventdev_devices = [cavium_sso, cavium_tim, octeontx2_sso] mempool_devices = [cavium_fpa, octeontx2_npa] compress_devices = [cavium_zip] +misc_devices = [intel_ioat_bdw, intel_ioat_skx] # global dict ethernet devices present. Dictionary indexed by PCI address. # Each device within this is itself a dictionary of device properties
@@ -595,6 +601,8 @@ def show_status(): if status_dev == "compress" or status_dev == "all": show_device_status(compress_devices , "Compress") + if status_dev == "misc" or status_dev == "all": + show_device_status(misc_devices, "Misc (rawdev)") def parse_args(): '''Parses the command-line arguments given by the user and takes the
@@ -670,6 +678,7 @@ def do_arg_actions(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) show_status()
@@ -690,6 +699,7 @@ def main(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) do_arg_actions() if __name__ == "__main__":
Signature · 1 line
2.21.0
PATCH v4 4/9 raw/ioat: add register definition file
In reply to Bruce RichardsonAdd in the list of registers for the device. File is taken from the SPDK project:
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
--- drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++++++++++++++ 3 files changed, 304 insertions(+), 1 deletion(-) create mode 100644 drivers/raw/ioat/rte_ioat_spec.h
drivers/raw/ioat/Makefile +1 -0
@@ -24,5 +24,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h +SYMLINK-y-include += rte_ioat_spec.h include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/meson.build +2 -1
@@ -5,4 +5,5 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c') deps += ['rawdev', 'bus_pci'] -install_headers('rte_ioat_rawdev.h') +install_headers('rte_ioat_rawdev.h', + 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_spec.h new file +301 -0
301 lines not on this page · raw message
Signature · 1 line
2.21.0
PATCH v4 5/9 raw/ioat: create device on probe and destroy on release
In reply to Bruce RichardsonAdd the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
---
V4:
* changed error code for pmd allocation to ENOMEM instead of EINVAL
* Use contiguous memory for device structure allocation
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++ drivers/raw/ioat/ioat_rawdev.c | 112 ++++++++++++++++++++++++++++- drivers/raw/ioat/rte_ioat_rawdev.h | 24 +++++++ 3 files changed, 144 insertions(+), 3 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -72,3 +72,14 @@ them to a suitable DPDK-supported kernel driver. When querying the status of the devices, they will appear under the category of "Misc (rawdev) devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be used to see the state of those devices alone. + +Device Probing and Initialization +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Once bound to a suitable kernel device driver, the HW devices will be found +as part of the PCI scan done at application initialization time. No vdev +parameters need to be passed to create or initialize the device. + +Once probed successfully, the device will appear as a ``rawdev``, that is a +"raw device type" inside DPDK, and can be accessed using APIs from the +``rte_rawdev`` library.
drivers/raw/ioat/ioat_rawdev.c +109 -3
130 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +24 -0
@@ -14,6 +14,10 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <rte_memory.h> +#include <rte_memzone.h> +#include <rte_ioat_spec.h> +/** Name of the device driver */ #define IOAT_PMD_RAWDEV_NAME rawdev_ioat/** String reported as the device driver name by rte_rawdev_info_get() */
@@ -21,4 +25,24 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* @internal +* Structure representing a device instance +*/ +struct rte_ioat_rawdev { +struct rte_rawdev *rawdev; +const struct rte_memzone *mz; +const struct rte_memzone *desc_mz; + +volatile struct rte_ioat_registers *regs; +phys_addr_t status_addr; +phys_addr_t ring_addr; + +unsigned short ring_size; +struct rte_ioat_generic_hw_desc *desc_ring; + +/* to report completions, the device will write status back here */ + volatile uint64_t status __rte_cache_aligned; +}; +#endif
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v4 5/9] raw/ioat: create device on probe and destroy on release
Quoted 11 lines · Bruce Richardson
Add the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4:
* changed error code for pmd allocation to ENOMEM instead of EINVAL
* Use contiguous memory for device structure allocation
---
Signature · 2 lines
Thanks, Anatoly
PATCH v4 6/9 raw/ioat: add device info function
In reply to Bruce RichardsonAdd in the "info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.
---
V4: Use TEST_SUCCESS in place of 0, when returning from test case.
Minor documentation updates.
V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)
app/test/test_rawdev.c +11 -0
@@ -29,6 +29,17 @@ REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton);static int test_rawdev_selftest_ioat (void ) { +const int count =rte_rawdev_count (); +int i; + +for (i =0 ; i < count; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; +if (rte_rawdev_info_get (i, &info) ==0 && +strstr (info.driver_name,"ioat" ) !=NULL ) +return TEST_SUCCESS ; + } + +printf ("No IOAT rawdev found, skipping tests\n" );return TEST_SKIPPED ; }
doc/guides/rawdevs/ioat_rawdev.rst +34 -0
@@ -83,3 +83,37 @@ parameters need to be passed to create or initialize the device. Once probed successfully, the device will appear as a ``rawdev``, that is a "raw device type" inside DPDK, and can be accessed using APIs from the ``rte_rawdev`` library. + +Using IOAT Rawdev Devices +-------------------------- + +To use the devices from an application, the rawdev API can be used, along +with definitions taken from the device-specific header file +``rte_ioat_rawdev.h``. This header is needed to get the definition of +structure parameters used by some of the rawdev APIs for IOAT rawdev +devices, as well as providing key functions for using the device for memory +copies. + +Getting Device Information +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Basic information about each rawdev device can be queried using the +``rte_rawdev_info_get()`` API. For most applications, this API will be +needed to verify that the rawdev in question is of the expected type. For +example, the following code snippet can be used to identify an IOAT +rawdev device for use by an application: + +.. code-block:: C + + for (i = 0; i < count && !found; i++) { + struct rte_rawdev_info info = { .dev_private = NULL }; + found = (rte_rawdev_info_get(i, &info) == 0 && + strcmp(info.driver_name, + IOAT_PMD_RAWDEV_NAME_STR) == 0); + } + +When calling the ``rte_rawdev_info_get()`` API for an IOAT rawdev device, +the ``dev_private`` field in the ``rte_rawdev_info`` struct should either +be NULL, or else be set to point to a structure of type +``rte_ioat_rawdev_config``, in which case the size of the configured device +input ring will be returned in that structure.
drivers/raw/ioat/ioat_rawdev.c +11 -0
@@ -34,10 +34,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +} +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_info_get = ioat_dev_info_get, };struct rte_rawdev *rawdev =NULL ;
drivers/raw/ioat/rte_ioat_rawdev.h +11 -0
@@ -25,6 +25,17 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* Configuration structure for an ioat rawdev instance +* +* This structure is to be passed as the ".dev_private" parameter when +* calling the rte_rawdev_get_info() and rte_rawdev_configure() APIs on +* an ioat rawdev instance. +*/ +struct rte_ioat_rawdev_config { +unsigned short ring_size; +}; +/** * @internal * Structure representing a device instance
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v4 6/9] raw/ioat: add device info function
Quoted 37 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Monday, July 1, 2019 11:56 PM
To: [email protected]
Cc: [email protected]; [email protected]; Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v4 6/9] raw/ioat: add device info functionAdd in the "
info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: UseTEST_SUCCESSin place of 0, when returning from test case.
Minor documentation updates.V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)drivers/raw/ioat/ioat_rawdev.c +0
-0@@ -34,10 +34,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args)
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 46
Quoted 10 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +}
The CBDMA user may want to know the maximum copy length that the device supports and the NUMA location of the device. So can ioat_dev_info_get() return more device information?
Thanks,
Jiayu
Re: [dpdk-dev] [PATCH v4 6/9] raw/ioat: add device info function
Quoted 38 lines · Jiayu
Quoted 38 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Monday, July 1, 2019 11:56 PM
To: [email protected]
Cc: [email protected]; [email protected]; Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v4 6/9] raw/ioat: add device info functionAdd in the "
info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: UseTEST_SUCCESSin place of 0, when returning from test case.
Minor documentation updates.V2: Test case is placed in driver self-test routine
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)drivers/raw/ioat/ioat_rawdev.c +0
-0@@ -34,10 +34,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args)
Quoted 14 lines
drivers/raw/ioat/ioat_rawdev.c static struct rte_pci_driver ioat_pmd_drv; · line 45
Quoted 9 lines
drivers/raw/ioat/ioat_rawdev.c +9
-0+static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +}The CBDMA user may want to know the maximum copy length that the device supports and the NUMA location of the device. So can
ioat_dev_info_get()return more device information?
The NUMA node id is already part of the rawdev info structure returned to the user.
As for maximum copy length, yes, that could be added. However, since it's actually a characteristic of the hardware, it could be just a define inside the header file rather than needing to be returned per instance. Unlike with ethdevs and cryptodevs, rawdevs always have a device-specific dependency so using defines from the header should work fine, and not everything needs to come via dev_info.
Regards,
/Bruce
Re: [dpdk-dev] [PATCH v4 6/9] raw/ioat: add device info function
In reply to Bruce RichardsonQuoted 4 lines · Bruce Richardson
Add in the "
info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.
Quoted 8 lines
Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: UseTEST_SUCCESSin place of 0, when returning from test case.
Minor documentation updates.V2: Test case is placed in driver self-test routine
---
Signature · 2 lines
Thanks, Anatoly
PATCH v4 7/9 raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonAllow initializing a driver instance. Include selftest to validate these functions.
---
V4: Guarantee correct SUCCESS/FAILURE return values from test function
Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory.
V3: don't add a new descriptor format struct, reuse from rte_ioat_spec.h
V2: test cases placed in self-test routine
--- app/test/test_rawdev.c | 3 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 +++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 82 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 2 + 7 files changed, 162 insertions(+), 2 deletions(-) create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c
app/test/test_rawdev.c +2 -1
@@ -36,7 +36,8 @@ test_rawdev_selftest_ioat(void)struct rte_rawdev_info info = { .dev_private =NULL };if (rte_rawdev_info_get (i, &info) ==0 &&strstr (info.driver_name,"ioat" ) !=NULL ) -return TEST_SUCCESS ; +return rte_rawdev_selftest (i) ==0 ? +TEST_SUCCESS :TEST_FAILED ; }printf ("No IOAT rawdev found, skipping tests\n" );
doc/guides/rawdevs/ioat_rawdev.rst +32 -0
@@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API.
drivers/raw/ioat/Makefile +1 -0
@@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h
drivers/raw/ioat/ioat_rawdev.c +82 -0
105 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c new file +41 -0
41 lines not on this page · raw message
drivers/raw/ioat/meson.build +2 -1
@@ -2,7 +2,8 @@ # Copyright 2019 Intel Corporation build = dpdk_conf.has('RTE_ARCH_X86') -sources = files('ioat_rawdev.c') +sources = files('ioat_rawdev.c', + 'ioat_rawdev_test.c') deps += ['rawdev', 'bus_pci'] install_headers('rte_ioat_rawdev.h',
drivers/raw/ioat/rte_ioat_rawdev.h +2 -0
@@ -14,6 +14,7 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <x86intrin.h> #include <rte_memory.h> #include <rte_memzone.h> #include <rte_ioat_spec.h>
@@ -51,6 +52,7 @@ struct rte_ioat_rawdev {unsigned short ring_size;struct rte_ioat_generic_hw_desc *desc_ring; +__m128i *hdls;/* completion handles for returning to user */ /* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned;
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v4 7/9] raw/ioat: add configure, start and stop functions
Quoted 13 lines · Bruce Richardson
Allow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: Guarantee correct SUCCESS/FAILURE return values from test function
Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory.
V3: don't add a new descriptor format struct, reuse fromrte_ioat_spec.h
V2: test cases placed in self-test routine
---
<snip>
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 125
Quoted 11 lines
static void ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) {drivers/raw/ioat/ioat_rawdev.c +2
-0@@ -44,11 +119,17 @@ ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) cfg->ring_size = ioat->ring_size; } +int ioat_rawdev_test (uint16_t dev_id); +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev)
...and...
drivers/raw/ioat/ioat_rawdev_test.c line 9
Quoted 3 lines
drivers/raw/ioat/ioat_rawdev_test.c new file +3
-0+ +int ioat_rawdev_test (uint16_t dev_id);/* pre-define to keep compiler happy */ +
This looks like you're trying to avoid introducing a header file? (to my great shame, i don't quite get how would that even work...)
Otherwise,
Signature · 2 lines
Thanks, Anatoly
Re: [dpdk-dev] [PATCH v4 7/9] raw/ioat: add configure, start and stop functions
Quoted 30 lines · Anatoly
Quoted 13 lines · Bruce Richardson
Allow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: Guarantee correct SUCCESS/FAILURE return values from test function
Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory.
V3: don't add a new descriptor format struct, reuse fromrte_ioat_spec.h
V2: test cases placed in self-test routine
---<snip>
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 124
Quoted 10 lines
static void ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) {drivers/raw/ioat/ioat_rawdev.c +2
-0@@ -44,11 +119,17 @@ ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) cfg->ring_size = ioat->ring_size; } +int ioat_rawdev_test (uint16_t dev_id); +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev)...and...
Quoted 7 lines
drivers/raw/ioat/ioat_rawdev_test.c line 9
Quoted 3 lines
drivers/raw/ioat/ioat_rawdev_test.c new file +3
-0+ +int ioat_rawdev_test (uint16_t dev_id);/* pre-define to keep compiler happy */ +This looks like you're trying to avoid introducing a header file? (to my great shame, i don't quite get how would that even work...)
Yes, yes I am. I fail to see the need to create a whole new header file just for a single function prototype, which can be easily stuck in the C file instead. If the scope of internal data or prototypes needed grows beyond this, then we can create an internal header to store it, but I don't foresee that happening any time soon, since things are pretty self-contained apart from these test cases.
Quoted 7 lines
Otherwise,
Acked-by Anatoly Burakov <[email protected]>Signature · 2 lines
Thanks, Anatoly
PATCH v4 8/9 raw/ioat: add statistics functions
In reply to Bruce RichardsonAdd stats functions to track what is happening in the driver, and put unit tests to check those.
---
V4: pulled in memory leak fixes from next patch
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 14 +++++++++ drivers/raw/ioat/ioat_rawdev.c | 44 ++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 45 +++++++++++++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 6 ++++ 4 files changed, 109 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +14 -0
@@ -149,3 +149,17 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. + +Querying Device Statistics +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The statistics from the IOAT rawdev device can be got via the xstats +functions in the ``rte_rawdev`` library, i.e. +``rte_rawdev_xstats_names_get()``, ``rte_rawdev_xstats_get()`` and +``rte_rawdev_xstats_by_name_get``. The statistics returned for each device +instance are: + +* ``failed_enqueues`` +* ``successful_enqueues`` +* ``copies_started`` +* ``copies_completed``
drivers/raw/ioat/ioat_rawdev.c +44 -0
62 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c +45 -0
62 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +6 -0
@@ -54,6 +54,12 @@ struct rte_ioat_rawdev {struct rte_ioat_generic_hw_desc *desc_ring;__m128i *hdls;/* completion handles for returning to user */ +/* some statistics for tracking, if added/changed update xstats fns*/ + uint64_t enqueue_failed __rte_cache_aligned; +uint64_t enqueued; +uint64_t started; +uint64_t completed; +/* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; };
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v4 8/9] raw/ioat: add statistics functions
Quoted 9 lines · Bruce Richardson
Add stats functions to track what is happening in the driver, and put unit tests to check those.
Signed-off-by Bruce Richardson <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: pulled in memory leak fixes from next patch
V2: test cases moved to self-test routine
---
Signature · 2 lines
Thanks, Anatoly
PATCH v4 9/9 raw/ioat: add local API to perform copies
In reply to Bruce RichardsonAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
---
V3: updated to use descriptor format in rte_ioat_spec.h
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 100 ++++++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev_test.c | 155 ++++++++++++++++++++++++- drivers/raw/ioat/meson.build | 2 +- drivers/raw/ioat/rte_ioat_rawdev.h | 169 +++++++++++++++++++++++++++- 5 files changed, 422 insertions(+), 5 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +100 -0
@@ -150,6 +150,106 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. +Performing Data Copies +~~~~~~~~~~~~~~~~~~~~~~~ + +To perform data copies using IOAT rawdev devices, the functions +``rte_ioat_enqueue_copy()`` and ``rte_ioat_do_copies()`` should be used. +Once copies have been completed, the completion will be reported back when +the application calls ``rte_ioat_completed_copies()``. + +The ``rte_ioat_enqueue_copy()`` function enqueues a single copy to the +device ring for copying at a later point. The parameters to that function +include the physical addresses of both the source and destination buffers, +as well as two "handles" to be returned to the user when the copy is +completed. These handles can be arbitrary values, but two are provided so +that the library can track handles for both source and destination on +behalf of the user, e.g. virtual addresses for the buffers, or mbuf +pointers if packet data is being copied. + +While the ``rte_ioat_enqueue_copy()`` function enqueues a copy operation on +the device ring, the copy will not actually be performed until after the +application calls the ``rte_ioat_do_copies()`` function. This function +informs the device hardware of the elements enqueued on the ring, and the +device will begin to process them. It is expected that, for efficiency +reasons, a burst of operations will be enqueued to the device via multiple +enqueue calls between calls to the ``rte_ioat_do_copies()`` function. + +The following code from ``test_ioat_rawdev.c`` demonstrates how to enqueue +a burst of copies to the device and start the hardware processing of them: + +.. code-block:: C + + struct rte_mbuf *srcs[32], *dsts[32]; + unsigned int j; + + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data; + + srcs[i] = rte_pktmbuf_alloc(pool); + dsts[i] = rte_pktmbuf_alloc(pool); + srcs[i]->data_len = srcs[i]->pkt_len = length; + dsts[i]->data_len = dsts[i]->pkt_len = length; + src_data = rte_pktmbuf_mtod(srcs[i], char *); + + for (j = 0; j < length; j++) + src_data[j] = rand() & 0xFF; + + if (rte_ioat_enqueue_copy(dev_id, + srcs[i]->buf_iova + srcs[i]->data_off, + dsts[i]->buf_iova + dsts[i]->data_off, + length, + (uintptr_t)srcs[i], + (uintptr_t)dsts[i], + 0 /* nofence */) != 1) { + printf("Error with rte_ioat_enqueue_copy for buffer %u\n", + i); + return -1; + } + } + rte_ioat_do_copies(dev_id); + +To retrieve information about completed copies, the API +``rte_ioat_completed_copies()`` should be used. This API will return to the +application a set of completion handles passed in when the relevant copies +were enqueued. + +The following code from ``test_ioat_rawdev.c`` shows the test code +retrieving information about the completed copies and validating the data +is correct before freeing the data buffers using the returned handles: + +.. code-block:: C + + if (rte_ioat_completed_copies(dev_id, 64, (void *)completed_src, + (void *)completed_dst) != RTE_DIM(srcs)) { + printf("Error with rte_ioat_completed_copies\n"); + return -1; + } + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data, *dst_data; + + if (completed_src[i] != srcs[i]) { + printf("Error with source pointer %u\n", i); + return -1; + } + if (completed_dst[i] != dsts[i]) { + printf("Error with dest pointer %u\n", i); + return -1; + } + + src_data = rte_pktmbuf_mtod(srcs[i], char *); + dst_data = rte_pktmbuf_mtod(dsts[i], char *); + for (j = 0; j < length; j++) + if (src_data[j] != dst_data[j]) { + printf("Error with copy of packet %u, byte %u\n", + i, j); + return -1; + } + rte_pktmbuf_free(srcs[i]); + rte_pktmbuf_free(dsts[i]); + } + + Querying Device Statistics ~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/raw/ioat/Makefile +1 -0
@@ -12,6 +12,7 @@ CFLAGS += $(WERROR_FLAGS) LDLIBS += -lrte_eal -lrte_rawdev LDLIBS += -lrte_pci -lrte_bus_pci +LDLIBS += -lrte_mbuf -lrte_mempool # library version LIBABIVER := 1
drivers/raw/ioat/ioat_rawdev_test.c +152 -3
187 lines not on this page · raw message
drivers/raw/ioat/meson.build +1 -1
@@ -4,7 +4,7 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c', 'ioat_rawdev_test.c') -deps += ['rawdev', 'bus_pci'] +deps += ['rawdev', 'bus_pci', 'mbuf'] install_headers('rte_ioat_rawdev.h', 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_rawdev.h +168 -1
186 lines not on this page · raw message
Signature · 1 line
2.21.0
Quoted 9 lines · Bruce Richardson
This patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
Apologies, forgot to mention this set has a dependency on:
http://patches.dpdk.org/project/dpdk/list/?series=5120
/Bruce
v5 cover raw/ioat: driver for Intel QuickData Technology
In reply to Bruce RichardsonThis patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
V5:
* updated doxygen comment for internal rawdev pmd API
V4:
* changed memory management to use contiguous memzones instead of malloc memory
* added missing device ids for BDX platforms
* other misc cleanup following review (see individual patches logs)
V3:
* removed DPDK-specific structure for the descriptor format and reused the structure in the imported file rte_ioat_spec.h
V2:
* moved tests to rawdev selftest function
* some checkpatch and other small cleanups
* added extra documentation details on supported hardware
* aligned the changes to dpdk-devbind with the changes in the NTB set for consistency
Bruce Richardson (9):
rawdev: allow devices to skip extra memory allocation raw/ioat: add initial support for ioat rawdev driver usertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
MAINTAINERS | 5 + app/test/test_rawdev.c | 20 ++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 265 +++++++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 + drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 31 ++ drivers/raw/ioat/ioat_rawdev.c | 356 ++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 235 +++++++++++++ drivers/raw/ioat/meson.build | 10 + drivers/raw/ioat/rte_ioat_rawdev.h | 234 +++++++++++++ drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- lib/librte_rawdev/rte_rawdev.c | 11 +- lib/librte_rawdev/rte_rawdev_pmd.h | 4 +- mk/rte.app.mk | 1 + usertools/dpdk-devbind.py | 10 + 22 files changed, 1505 insertions(+), 7 deletions(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_ioat_spec.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
Signature · 1 line
2.21.0
PATCH v5 1/9 rawdev: allow devices to skip extra memory allocation
Some device drivers want to allocate their own private memory, and should be allowed to do so. Therefore skip memory allocation and associated error checks if zero-length private memory is requested.
While adjusting the code for new indent level, fix incorrect error message.
---
V5: Added additional info to doxygen for allocate API
--- lib/librte_rawdev/rte_rawdev.c | 11 ++++++----- lib/librte_rawdev/rte_rawdev_pmd.h | 4 +++- 2 files changed, 9 insertions(+), 6 deletions(-)
lib/librte_rawdev/rte_rawdev.c +6 -5
@@ -496,16 +496,17 @@ rte_rawdev_pmd_allocate(const char *name, size_t dev_priv_size, int socket_id) rawdev = &rte_rawdevs[dev_id]; - rawdev->dev_private =rte_zmalloc_socket ("rawdev private" , +if (dev_priv_size >0 ) { + rawdev->dev_private =rte_zmalloc_socket ("rawdev private" , dev_priv_size,RTE_CACHE_LINE_SIZE , socket_id); -if (!rawdev->dev_private) { -RTE_RDEV_ERR ("Unable to allocate memory to Skeleton dev" ); -return NULL ; +if (!rawdev->dev_private) { +RTE_RDEV_ERR ("Unable to allocate memory for rawdev" ); +return NULL ; + } } - rawdev->dev_id = dev_id; rawdev->socket_id = socket_id; rawdev->started =0 ;
lib/librte_rawdev/rte_rawdev_pmd.h +3 -1
@@ -568,7 +568,9 @@ struct rte_rawdev_ops {* @param name * Unique identifier name for each device * @param dev_private_size -* Private data allocated within rte_rawdev object. +* Size of private data memory allocated within rte_rawdev object. +* Set to 0 to disable internal memory allocation and allow for +* self-allocation. * @param socket_id * Socket to allocate resources on. * @return
Signature · 1 line
2.21.0
PATCH v5 2/9 raw/ioat: add initial support for ioat rawdev driver
In reply to Bruce RichardsonAdd stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit tests
---
V4: add in missing device IDs
V3: remove unneeded whitespace change
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
--- MAINTAINERS | 5 + app/test/test_rawdev.c | 8 ++ config/common_armv8a_linux | 1 + config/common_base | 5 + config/defconfig_arm-armv7a-linuxapp-gcc | 1 + config/defconfig_ppc_64-power8-linuxapp-gcc | 1 + doc/guides/rawdevs/index.rst | 1 + doc/guides/rawdevs/ioat_rawdev.rst | 63 +++++++++++ doc/guides/rel_notes/release_19_08.rst | 11 ++ drivers/raw/Makefile | 1 + drivers/raw/ioat/Makefile | 28 +++++ drivers/raw/ioat/ioat_rawdev.c | 113 ++++++++++++++++++++ drivers/raw/ioat/meson.build | 8 ++ drivers/raw/ioat/rte_ioat_rawdev.h | 24 +++++ drivers/raw/ioat/rte_pmd_ioat_version.map | 4 + drivers/raw/meson.build | 4 +- mk/rte.app.mk | 1 + 17 files changed, 278 insertions(+), 1 deletion(-) create mode 100644 doc/guides/rawdevs/ioat_rawdev.rst create mode 100644 drivers/raw/ioat/Makefile create mode 100644 drivers/raw/ioat/ioat_rawdev.c create mode 100644 drivers/raw/ioat/meson.build create mode 100644 drivers/raw/ioat/rte_ioat_rawdev.h create mode 100644 drivers/raw/ioat/rte_pmd_ioat_version.map
MAINTAINERS +5 -0
@@ -1084,6 +1084,11 @@ M: Tianfei zhang <[email protected]> F: drivers/raw/ifpga_rawdev/ F: doc/guides/rawdevs/ifpga_rawdev.rst +IOAT Rawdev +M: Bruce Richardson <[email protected]> +F: drivers/raw/ioat/ +F: doc/guides/rawdevs/ioat_rawdev.rst + NXP DPAA2 QDMA M: Nipun Gupta <[email protected]> F: drivers/raw/dpaa2_qdma/
app/test/test_rawdev.c +8 -0
@@ -25,3 +25,11 @@ test_rawdev_selftest_skeleton(void) }REGISTER_TEST_COMMAND (rawdev_autotest, test_rawdev_selftest_skeleton); + +static int +test_rawdev_selftest_ioat (void ) +{ +return TEST_SKIPPED ; +} + +REGISTER_TEST_COMMAND (ioat_rawdev_autotest, test_rawdev_selftest_ioat);
config/common_armv8a_linux +1 -0
@@ -34,5 +34,6 @@ CONFIG_RTE_ARCH_ARM64_MEMCPY=n CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n CONFIG_RTE_SCHED_VECTOR=n
config/common_base +5 -0
@@ -752,6 +752,11 @@ CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV=n # CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV=y +# +# Compile PMD for Intel IOAT raw device +# +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=y + # # Compile librte_ring #
config/defconfig_arm-armv7a-linuxapp-gcc +1 -0
@@ -55,3 +55,4 @@ CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_NFP_PMD=n CONFIG_RTE_LIBRTE_HINIC_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
config/defconfig_ppc_64-power8-linuxapp-gcc +1 -0
@@ -57,3 +57,4 @@ CONFIG_RTE_LIBRTE_FM10K_PMD=n CONFIG_RTE_LIBRTE_SFC_EFX_PMD=n CONFIG_RTE_LIBRTE_AVP_PMD=n CONFIG_RTE_LIBRTE_HINIC_PMD=n +CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV=n
doc/guides/rawdevs/index.rst +1 -0
@@ -14,3 +14,4 @@ application through rawdev API. dpaa2_cmdif dpaa2_qdma ifpga_rawdev + ioat_rawdev
doc/guides/rawdevs/ioat_rawdev.rst new file +63 -0
63 lines not on this page · raw message
doc/guides/rel_notes/release_19_08.rst +11 -0
@@ -1,6 +1,8 @@ .. SPDX-License-Identifier: BSD-3-Clause Copyright 2019 The DPDK contributors +.. include:: <isonum.txt> + DPDK Release 19.08 ==================
@@ -126,6 +128,15 @@ New Features Added telemetry mode to l3fwd-power application to report application level busyness, empty and full polls of rte_eth_rx_burst(). +* **Added Intel QuickData Technology PMD** + + The PMD for Intel\ |reg| QuickData Technology, part of + Intel\ |reg| I/O Acceleration Technology `(Intel I/OAT) + <https://www.intel.com/content/www/us/en/wireless-network/accel-technology.html>`_, + allows data copies to be done by hardware instead + of via software, reducing cycles spent copying large blocks of data in + applications. + Removed Items -------------
drivers/raw/Makefile +1 -0
@@ -10,5 +10,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_CMDIF_RAWDEV) += dpaa2_cmdif DIRS-$(CONFIG_RTE_LIBRTE_PMD_DPAA2_QDMA_RAWDEV) += dpaa2_qdma endif DIRS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += ifpga_rawdev +DIRS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat include $(RTE_SDK)/mk/rte.subdir.mk
drivers/raw/ioat/Makefile new file +28 -0
@@ -0,0 +1,28 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright(c) 2019 Intel Corporation + +include $(RTE_SDK)/mk/rte.vars.mk + +# library name +LIB = librte_pmd_ioat_rawdev.a + +# build flags +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +LDLIBS += -lrte_eal -lrte_rawdev +LDLIBS += -lrte_pci -lrte_bus_pci + +# library version +LIBABIVER := 1 + +# versioning export map +EXPORT_MAP := rte_pmd_ioat_version.map + +# library source files +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c + +# export include files +SYMLINK-y-include += rte_ioat_rawdev.h + +include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/ioat_rawdev.c new file +113 -0
113 lines not on this page · raw message
drivers/raw/ioat/meson.build new file +8 -0
@@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright 2019 Intel Corporation + +build = dpdk_conf.has('RTE_ARCH_X86') +sources = files('ioat_rawdev.c') +deps += ['rawdev', 'bus_pci'] + +install_headers('rte_ioat_rawdev.h')
drivers/raw/ioat/rte_ioat_rawdev.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#ifndef _RTE_IOAT_RAWDEV_H_ +#define _RTE_IOAT_RAWDEV_H_ + +/** +* @file rte_ioat_rawdev.h +* +* Definitions for using the ioat rawdev device driver +* +* @warning +* @b EXPERIMENTAL: these structures and APIs may change without prior notice +*/ + +/** Name of the device driver */ +#define IOAT_PMD_RAWDEV_NAME rawdev_ioat +/** String reported as the device driver name by rte_rawdev_info_get() */ +#define IOAT_PMD_RAWDEV_NAME_STR "rawdev_ioat" +/** Name used to adjust the log level for this driver */ +#define IOAT_PMD_LOG_NAME "rawdev.ioat" + +#endif
drivers/raw/ioat/rte_pmd_ioat_version.map new file +4 -0
@@ -0,0 +1,4 @@ +DPDK_19.08 { + + local: *; +};
drivers/raw/meson.build +3 -1
@@ -1,7 +1,9 @@ # SPDX-License-Identifier: BSD-3-Clause # Copyright 2018 NXP -drivers = ['skeleton_rawdev', 'dpaa2_cmdif', 'dpaa2_qdma', 'ifpga_rawdev'] +drivers = ['dpaa2_cmdif', 'dpaa2_qdma', + 'ifpga_rawdev', 'ioat', + 'skeleton_rawdev'] std_deps = ['rawdev'] config_flag_fmt = 'RTE_LIBRTE_PMD_@0@_RAWDEV' driver_name_fmt = 'rte_pmd_@0@'
mk/rte.app.mk +1 -0
@@ -307,6 +307,7 @@ ifeq ($(CONFIG_RTE_LIBRTE_IFPGA_BUS),y) _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IFPGA_RAWDEV) += -lrte_pmd_ifpga_rawdev _LDLIBS-$(CONFIG_RTE_LIBRTE_IPN3KE_PMD) += -lrte_pmd_ipn3ke endif # CONFIG_RTE_LIBRTE_IFPGA_BUS +_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += -lrte_pmd_ioat_rawdev endif # CONFIG_RTE_LIBRTE_RAWDEV endif # !CONFIG_RTE_BUILD_SHARED_LIBS
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 2/9] raw/ioat: add initial support for ioat rawdev driver
Quoted 24 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:12 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 2/9] raw/ioat: add initial support for ioat rawdev driverAdd stubs for ioat rawdev driver support in DPDK, specifically:
* makefile and meson build hooks
* initial public header file
* rawdev main C file, with probe and release functions
* release note update announcing the driver
* initial documentation for the new section in the rawdev doc
* unit test stubs for device unit testsSigned-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>
Quoted 9 lines
---
V4: add in missing device IDs
V3: remove unneeded whitespace change
V2: don't create a new file for unit testing, add to existing rawdev test
file, and place test cases in the driver selftest routine (added later in set)
Add new section in document about identifying hardware using lspci
---
PATCH v5 3/9 usertools/dpdk-devbind.py: add support for IOAT devices
In reply to Bruce RichardsonIn order to allow binding/unbinding of devices for use by the ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.
---
V2: rather than adding a DMA section, add to "misc (rawdev)" section where
other device types, e.g. ntb can also do.
NOTE: this set largely overlaps with the equivalent changes made by the patchset adding NTB support [1]. Since it's unclear which set will be added first, this set is based off the latest head.
[1] http://patches.dpdk.org/patch/55127/
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++++++++++ usertools/dpdk-devbind.py | 10 ++++++++++ 2 files changed, 21 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -61,3 +61,14 @@ configurations. For builds using ``meson`` and ``ninja``, the driver will be built when the target platform is x86-based. + +Device Setup +------------- + +The Intel\ |reg| QuickData Technology HW devices will need to be bound to a +user-space IO driver for use. The script ``dpdk-devbind.py`` script +included with DPDK can be used to view the state of the devices and to bind +them to a suitable DPDK-supported kernel driver. When querying the status +of the devices, they will appear under the category of "Misc (rawdev) +devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be +used to see the state of those devices alone.
usertools/dpdk-devbind.py +10 -0
@@ -36,11 +36,17 @@ octeontx2_npa = {'Class': '08', 'Vendor': '177d', 'Device': 'a0fb,a0fc', 'SVendor': None, 'SDevice': None} +intel_ioat_bdw = {'Class': '08', 'Vendor': '8086', 'Device': '6f20,6f21,6f22,6f23,6f24,6f25,6f26,6f27,6f2e,6f2f', + 'SVendor': None, 'SDevice': None} +intel_ioat_skx = {'Class': '08', 'Vendor': '8086', 'Device': '2021', + 'SVendor': None, 'SDevice': None} + network_devices = [network_class, cavium_pkx, avp_vnic, ifpga_class] crypto_devices = [encryption_class, intel_processor_class] eventdev_devices = [cavium_sso, cavium_tim, octeontx2_sso] mempool_devices = [cavium_fpa, octeontx2_npa] compress_devices = [cavium_zip] +misc_devices = [intel_ioat_bdw, intel_ioat_skx] # global dict ethernet devices present. Dictionary indexed by PCI address. # Each device within this is itself a dictionary of device properties
@@ -595,6 +601,8 @@ def show_status(): if status_dev == "compress" or status_dev == "all": show_device_status(compress_devices , "Compress") + if status_dev == "misc" or status_dev == "all": + show_device_status(misc_devices, "Misc (rawdev)") def parse_args(): '''Parses the command-line arguments given by the user and takes the
@@ -670,6 +678,7 @@ def do_arg_actions(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) show_status()
@@ -690,6 +699,7 @@ def main(): get_device_details(eventdev_devices) get_device_details(mempool_devices) get_device_details(compress_devices) + get_device_details(misc_devices) do_arg_actions() if __name__ == "__main__":
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 3/9] usertools/dpdk-devbind.py: add support for IOAT devices
Quoted 17 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:12 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 3/9]usertools/dpdk-devbind.py: add support for IOAT devicesIn order to allow binding/unbinding of devices for use by the
ioat_rawdev, we need to update the devbind script to add a new class of device, and add device ids for the specific HW instances.
Quoted 9 lines
Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V2: rather than adding a DMA section, add to "misc (rawdev)" section where
other device types, e.g. ntb can also do.
PATCH v5 4/9 raw/ioat: add register definition file
In reply to Bruce RichardsonAdd in the list of registers for the device. File is taken from the SPDK project:
https://github.com/spdk/spdk/blob/master/include/spdk/ioat_spec.h
--- drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_spec.h | 301 +++++++++++++++++++++++++++++++ 3 files changed, 304 insertions(+), 1 deletion(-) create mode 100644 drivers/raw/ioat/rte_ioat_spec.h
drivers/raw/ioat/Makefile +1 -0
@@ -24,5 +24,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h +SYMLINK-y-include += rte_ioat_spec.h include $(RTE_SDK)/mk/rte.lib.mk
drivers/raw/ioat/meson.build +2 -1
@@ -5,4 +5,5 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c') deps += ['rawdev', 'bus_pci'] -install_headers('rte_ioat_rawdev.h') +install_headers('rte_ioat_rawdev.h', + 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_spec.h new file +301 -0
301 lines not on this page · raw message
Signature · 1 line
2.21.0
PATCH v5 5/9 raw/ioat: create device on probe and destroy on release
In reply to Bruce RichardsonAdd the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
---
V4:
* changed error code for pmd allocation to ENOMEM instead of EINVAL
* Use contiguous memory for device structure allocation
--- doc/guides/rawdevs/ioat_rawdev.rst | 11 +++ drivers/raw/ioat/ioat_rawdev.c | 112 ++++++++++++++++++++++++++++- drivers/raw/ioat/rte_ioat_rawdev.h | 24 +++++++ 3 files changed, 144 insertions(+), 3 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +11 -0
@@ -72,3 +72,14 @@ them to a suitable DPDK-supported kernel driver. When querying the status of the devices, they will appear under the category of "Misc (rawdev) devices", i.e. the command ``dpdk-devbind.py --status-dev misc`` can be used to see the state of those devices alone. + +Device Probing and Initialization +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Once bound to a suitable kernel device driver, the HW devices will be found +as part of the PCI scan done at application initialization time. No vdev +parameters need to be passed to create or initialize the device. + +Once probed successfully, the device will appear as a ``rawdev``, that is a +"raw device type" inside DPDK, and can be accessed using APIs from the +``rte_rawdev`` library.
drivers/raw/ioat/ioat_rawdev.c +109 -3
130 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +24 -0
@@ -14,6 +14,10 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <rte_memory.h> +#include <rte_memzone.h> +#include <rte_ioat_spec.h> +/** Name of the device driver */ #define IOAT_PMD_RAWDEV_NAME rawdev_ioat/** String reported as the device driver name by rte_rawdev_info_get() */
@@ -21,4 +25,24 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* @internal +* Structure representing a device instance +*/ +struct rte_ioat_rawdev { +struct rte_rawdev *rawdev; +const struct rte_memzone *mz; +const struct rte_memzone *desc_mz; + +volatile struct rte_ioat_registers *regs; +phys_addr_t status_addr; +phys_addr_t ring_addr; + +unsigned short ring_size; +struct rte_ioat_generic_hw_desc *desc_ring; + +/* to report completions, the device will write status back here */ + volatile uint64_t status __rte_cache_aligned; +}; +#endif
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 5/9] raw/ioat: create device on probe and destroy on release
Quoted 19 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:12 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 5/9] raw/ioat: create device on probe and destroy on releaseAdd the create/destroy driver functions so that we can actually allocate a rawdev and destroy it when done. No rawdev API functions are actually implemented at this point.
Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
PATCH v5 6/9 raw/ioat: add device info function
In reply to Bruce RichardsonAdd in the "info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.
---
V2: Test case is placed in driver self-test routine
V4: Use TEST_SUCCESS in place of 0, when returning from test case.
Minor documentation updates.
--- app/test/test_rawdev.c | 11 ++++++++++ doc/guides/rawdevs/ioat_rawdev.rst | 34 ++++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev.c | 11 ++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 11 ++++++++++ 4 files changed, 67 insertions(+)
app/test/test_rawdev.c +11 -0
@@ -29,6 +29,17 @@ REGISTER_TEST_COMMAND(rawdev_autotest, test_rawdev_selftest_skeleton);static int test_rawdev_selftest_ioat (void ) { +const int count =rte_rawdev_count (); +int i; + +for (i =0 ; i < count; i++) { +struct rte_rawdev_info info = { .dev_private =NULL }; +if (rte_rawdev_info_get (i, &info) ==0 && +strstr (info.driver_name,"ioat" ) !=NULL ) +return TEST_SUCCESS ; + } + +printf ("No IOAT rawdev found, skipping tests\n" );return TEST_SKIPPED ; }
doc/guides/rawdevs/ioat_rawdev.rst +34 -0
@@ -83,3 +83,37 @@ parameters need to be passed to create or initialize the device. Once probed successfully, the device will appear as a ``rawdev``, that is a "raw device type" inside DPDK, and can be accessed using APIs from the ``rte_rawdev`` library. + +Using IOAT Rawdev Devices +-------------------------- + +To use the devices from an application, the rawdev API can be used, along +with definitions taken from the device-specific header file +``rte_ioat_rawdev.h``. This header is needed to get the definition of +structure parameters used by some of the rawdev APIs for IOAT rawdev +devices, as well as providing key functions for using the device for memory +copies. + +Getting Device Information +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Basic information about each rawdev device can be queried using the +``rte_rawdev_info_get()`` API. For most applications, this API will be +needed to verify that the rawdev in question is of the expected type. For +example, the following code snippet can be used to identify an IOAT +rawdev device for use by an application: + +.. code-block:: C + + for (i = 0; i < count && !found; i++) { + struct rte_rawdev_info info = { .dev_private = NULL }; + found = (rte_rawdev_info_get(i, &info) == 0 && + strcmp(info.driver_name, + IOAT_PMD_RAWDEV_NAME_STR) == 0); + } + +When calling the ``rte_rawdev_info_get()`` API for an IOAT rawdev device, +the ``dev_private`` field in the ``rte_rawdev_info`` struct should either +be NULL, or else be set to point to a structure of type +``rte_ioat_rawdev_config``, in which case the size of the configured device +input ring will be returned in that structure.
drivers/raw/ioat/ioat_rawdev.c +11 -0
@@ -34,10 +34,21 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +static void +ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) +{ +struct rte_ioat_rawdev_config *cfg = dev_info; +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (cfg !=NULL ) + cfg->ring_size = ioat->ring_size; +} +static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_info_get = ioat_dev_info_get, };struct rte_rawdev *rawdev =NULL ;
drivers/raw/ioat/rte_ioat_rawdev.h +11 -0
@@ -25,6 +25,17 @@/** Name used to adjust the log level for this driver */ #define IOAT_PMD_LOG_NAME "rawdev.ioat" +/** +* Configuration structure for an ioat rawdev instance +* +* This structure is to be passed as the ".dev_private" parameter when +* calling the rte_rawdev_get_info() and rte_rawdev_configure() APIs on +* an ioat rawdev instance. +*/ +struct rte_ioat_rawdev_config { +unsigned short ring_size; +}; +/** * @internal * Structure representing a device instance
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 6/9] raw/ioat: add device info function
Quoted 21 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:12 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 6/9] raw/ioat: add device info functionAdd in the "
info_get" function to the driver, to allow us to query the device. This allows us to have the unit test pick up the presence of supported hardware or not.Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V2: Test case is placed in driver self-test routine
V4: UseTEST_SUCCESSin place of 0, when returning from test case.
PATCH v5 7/9 raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonAllow initializing a driver instance. Include selftest to validate these functions.
---
V4: Guarantee correct SUCCESS/FAILURE return values from test function
Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory.
V3: don't add a new descriptor format struct, reuse from rte_ioat_spec.h
V2: test cases placed in self-test routine
--- app/test/test_rawdev.c | 3 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 +++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 82 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 2 + 7 files changed, 162 insertions(+), 2 deletions(-) create mode 100644 drivers/raw/ioat/ioat_rawdev_test.c
app/test/test_rawdev.c +2 -1
@@ -36,7 +36,8 @@ test_rawdev_selftest_ioat(void)struct rte_rawdev_info info = { .dev_private =NULL };if (rte_rawdev_info_get (i, &info) ==0 &&strstr (info.driver_name,"ioat" ) !=NULL ) -return TEST_SUCCESS ; +return rte_rawdev_selftest (i) ==0 ? +TEST_SUCCESS :TEST_FAILED ; }printf ("No IOAT rawdev found, skipping tests\n" );
doc/guides/rawdevs/ioat_rawdev.rst +32 -0
@@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API.
drivers/raw/ioat/Makefile +1 -0
@@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h
drivers/raw/ioat/ioat_rawdev.c +82 -0
105 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c new file +41 -0
41 lines not on this page · raw message
drivers/raw/ioat/meson.build +2 -1
@@ -2,7 +2,8 @@ # Copyright 2019 Intel Corporation build = dpdk_conf.has('RTE_ARCH_X86') -sources = files('ioat_rawdev.c') +sources = files('ioat_rawdev.c', + 'ioat_rawdev_test.c') deps += ['rawdev', 'bus_pci'] install_headers('rte_ioat_rawdev.h',
drivers/raw/ioat/rte_ioat_rawdev.h +2 -0
@@ -14,6 +14,7 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <x86intrin.h> #include <rte_memory.h> #include <rte_memzone.h> #include <rte_ioat_spec.h>
@@ -51,6 +52,7 @@ struct rte_ioat_rawdev {unsigned short ring_size;struct rte_ioat_generic_hw_desc *desc_ring; +__m128i *hdls;/* completion handles for returning to user */ /* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned;
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 7/9] raw/ioat: add configure, start and stop functions
Quoted 18 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:12 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 7/9] raw/ioat: add configure, start and stop functionsAllow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
Re: [dpdk-dev] [PATCH v5 7/9] raw/ioat: add configure, start and stop functions
In reply to Bruce RichardsonQuoted 177 lines · Bruce Richardson
Allow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: Guarantee correct SUCCESS/FAILURE return values from test function
Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory.
V3: don't add a new descriptor format struct, reuse fromrte_ioat_spec.h
V2: test cases placed in self-test routine--- app/test/test_rawdev.c | 3 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 +++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 82 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 2 + 7 files changed, 162 insertions(+), 2 deletions(-) create mode 100644 drivers/raw/ioat/ioat_rawdev_test.capp/test/test_rawdev.c +2
-1@@ -36,7 +36,8 @@ test_rawdev_selftest_ioat(void)struct rte_rawdev_info info = { .dev_private =NULL };if (rte_rawdev_info_get (i, &info) ==0 &&strstr (info.driver_name,"ioat" ) !=NULL ) -return TEST_SUCCESS ; +return rte_rawdev_selftest (i) ==0 ? +TEST_SUCCESS :TEST_FAILED ; }printf ("No IOAT rawdev found, skipping tests\n" );doc/guides/rawdevs/ioat_rawdev.rst +32
-0@@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API.drivers/raw/ioat/Makefile +1
-0@@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.hdrivers/raw/ioat/ioat_rawdev.c +75
-0@@ -34,6 +34,81 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +#define DESC_SZ sizeof(struct rte_ioat_generic_hw_desc) +#define COMPLETION_SZ sizeof(__m128i) + +static int +ioat_dev_configure (const struct rte_rawdev *dev,rte_rawdev_obj_t config) +{ +struct rte_ioat_rawdev_config *params = config; +struct rte_ioat_rawdev *ioat = dev->dev_private; +char mz_name[RTE_MEMZONE_NAMESIZE ]; +unsigned short i; + +if (dev->started) +return -EBUSY ; + +if (params ==NULL ) +return -EINVAL ; + +if (params->ring_size >4096 || params->ring_size <64 || + !rte_is_power_of_2 (params->ring_size)) +return -EINVAL ; + + ioat->ring_size = params->ring_size; +if (ioat->desc_ring !=NULL ) { +rte_memzone_free (ioat->desc_mz); + ioat->desc_ring =NULL ; + ioat->desc_mz =NULL ; + } + +/* allocate one block of memory for both descriptors +* and completion handles. +*/ +snprintf (mz_name,sizeof (mz_name),"rawdev%u_desc_ring" , dev->dev_id); + ioat->desc_mz =rte_memzone_reserve (mz_name, + (DESC_SZ +COMPLETION_SZ ) * ioat->ring_size, + dev->device->numa_node,RTE_MEMZONE_IOVA_CONTIG ); +if (ioat->desc_mz ==NULL ) +return -ENOMEM ; + ioat->desc_ring = ioat->desc_mz->addr; + ioat->hdls = (void *)&ioat->desc_ring[ioat->ring_size]; + + ioat->ring_addr = ioat->desc_mz->iova; + +/* configure descriptor ring - each one points to next */ +for (i =0 ; i < ioat->ring_size; i++) { + ioat->desc_ring[i].next = ioat->ring_addr + + (((i +1 ) % ioat->ring_size) *DESC_SZ ); + } + +return 0 ; +} + +static int +ioat_dev_start (struct rte_rawdev *dev) +{ +struct rte_ioat_rawdev *ioat = dev->dev_private; + +if (ioat->ring_size ==0 || ioat->desc_ring ==NULL ) +return -EBUSY ; + +/* inform hardware of where the descriptor ring is */ + ioat->regs->chainaddr = ioat->ring_addr; +/* inform hardware of where to write the status/completions */ + ioat->regs->chancmp = ioat->status_addr; + +/* prime the status register to be set to the last element */ + ioat->status = ioat->ring_addr + ((ioat->ring_size -1 ) *DESC_SZ ); +return 0 ; +} + +static void +ioat_dev_stop (struct rte_rawdev *dev) +{ +RTE_SET_USED (dev); +} +static void ioat_dev_info_get (struct rte_rawdev *dev,rte_rawdev_obj_t dev_info) {
drivers/raw/ioat/ioat_rawdev.c ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) · line 123
Quoted 5 lines
drivers/raw/ioat/ioat_rawdev.c +2
-0@@ -44,11 +119,17 @@ ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info) cfg->ring_size = ioat->ring_size; } +extern int ioat_rawdev_test (uint16_t dev_id); +
This signature doesn't match with the rte_rawdev_ops definition (which should be int(*func)(void)).
It will generate an error as such:
../drivers/raw/ioat/ioat_rawdev.c:176:20: error: initialization from incompatible pointer type [-Werror=incompatible-pointer-types]
.dev_selftest = ioat_rawdev_test,
^../drivers/raw/ioat/ioat_rawdev.c:176:20: note: (near initialization for ‘ioat_rawdev_ops.dev_selftest’)
Quoted 101 lines
drivers/raw/ioat/ioat_rawdev.c +5
-0static int ioat_rawdev_create (const char *name,struct rte_pci_device *dev) {static const struct rte_rawdev_ops ioat_rawdev_ops = { + .dev_configure = ioat_dev_configure, + .dev_start = ioat_dev_start, + .dev_stop = ioat_dev_stop, .dev_info_get = ioat_dev_info_get, + .dev_selftest = ioat_rawdev_test, };struct rte_rawdev *rawdev =NULL ;@@ -154,6 +235,7 @@ ioat_rawdev_destroy(const char *name)if (rdev->dev_private !=NULL ) {struct rte_ioat_rawdev *ioat = rdev->dev_private; rdev->dev_private =NULL ; +rte_memzone_free (ioat->desc_mz);rte_memzone_free (ioat->mz); }drivers/raw/ioat/ioat_rawdev_test.c new file +41
-0@@ -0,0 +1,41 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright(c) 2019 Intel Corporation +*/ + +#include "rte_rawdev.h" +#include "rte_ioat_rawdev.h" + +int ioat_rawdev_test (uint16_t dev_id);/* pre-define to keep compiler happy */ + +int +ioat_rawdev_test (uint16_t dev_id) +{ +#define IOAT_TEST_RINGSIZE 512 +struct rte_ioat_rawdev_config p = { .ring_size =-1 }; +struct rte_rawdev_info info = { .dev_private = &p }; + +rte_rawdev_info_get (dev_id, &info); +if (p.ring_size !=0 ) { +printf ("Error, initial ring size is non-zero (%d)\n" , + (int )p.ring_size); +return -1 ; + } + + p.ring_size =IOAT_TEST_RINGSIZE ; +if (rte_rawdev_configure (dev_id, &info) !=0 ) { +printf ("Error with rte_rawdev_configure()\n" ); +return -1 ; + } +rte_rawdev_info_get (dev_id, &info); +if (p.ring_size !=IOAT_TEST_RINGSIZE ) { +printf ("Error, ring size is not %d (%d)\n" , +IOAT_TEST_RINGSIZE , (int )p.ring_size); +return -1 ; + } + +if (rte_rawdev_start (dev_id) !=0 ) { +printf ("Error with rte_rawdev_start()\n" ); +return -1 ; + } +return 0 ; +}drivers/raw/ioat/meson.build +2
-1@@ -2,7 +2,8 @@ # Copyright 2019 Intel Corporation build = dpdk_conf.has('RTE_ARCH_X86') -sources = files('ioat_rawdev.c') +sources = files('ioat_rawdev.c', + 'ioat_rawdev_test.c') deps += ['rawdev', 'bus_pci'] install_headers('rte_ioat_rawdev.h',drivers/raw/ioat/rte_ioat_rawdev.h +2
-0@@ -14,6 +14,7 @@* @b EXPERIMENTAL: these structures and APIs may change without prior notice */ +#include <x86intrin.h> #include <rte_memory.h> #include <rte_memzone.h> #include <rte_ioat_spec.h> @@ -51,6 +52,7 @@ struct rte_ioat_rawdev {unsigned short ring_size;struct rte_ioat_generic_hw_desc *desc_ring; +__m128i *hdls;/* completion handles for returning to user */ /* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned;
Re: [dpdk-dev] [PATCH v5 7/9] raw/ioat: add configure, start and stop functions
Quoted 121 lines · Aaron Conole
Quoted 114 lines · Bruce Richardson
Allow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]> Acked-by:Anatoly Burakov <[email protected]> Tested-by: Harry van Haaren <[email protected]> ---
V4: Guarantee correct SUCCESS/FAILURE return values from test function Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory. V3: don't add a new descriptor format struct, reuse from
rte_ioat_spec.hV2: test cases placed in self-test routine ---app/test/test_rawdev.c | 3 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 +++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 82 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 2 + 7 files changed, 162insertions(+), 2 deletions(-) create mode 100644
drivers/raw/ioat/ioat_rawdev_test.capp/test/test_rawdev.c index renamed from app/test/test_rawdev.c +0
-0
623117af9..524a9d5f3100644 ---a/app/test/test_rawdev.c+++b/app/test/test_rawdev.c@@ -36,7 +36,8 @@test_rawdev_selftest_ioat(void)structrte_rawdev_infoinfo = { .dev_private = NULL }; if (rte_rawdev_info_get(i, &info)== 0 &&strstr(info.driver_name, "ioat")!= NULL) - returnTEST_SUCCESS; + returnrte_rawdev_selftest(i)== 0
? +TEST_SUCCESS:TEST_FAILED;
}app/test/test_rawdev.c index renamed from app/test/test_rawdev.c +1
-0printf("No IOAT rawdev found, skipping tests\n"); diff --git a/doc/guides/rawdevs/ioat_rawdev.rst b/doc/guides/rawdevs/ioat_rawdev.rst index efed9c64c..a0594d2cb 100644 --- a/doc/guides/rawdevs/ioat_rawdev.rst +++ b/doc/guides/rawdevs/ioat_rawdev.rst @@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API. diff --git a/drivers/raw/ioat/Makefile b/drivers/raw/ioat/Makefile index 1e10938f3..b1af9c666 100644 --- a/drivers/raw/ioat/Makefile +++ b/drivers/raw/ioat/Makefile @@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h diff --git a/drivers/raw/ioat/ioat_rawdev.c b/drivers/raw/ioat/ioat_rawdev.c index 2bfe2544d..0c91b3579 100644 --- a/drivers/raw/ioat/ioat_rawdev.c +++ b/drivers/raw/ioat/ioat_rawdev.c @@ -34,6 +34,81 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +#define DESC_SZ sizeof(struct rte_ioat_generic_hw_desc) +#define
COMPLETION_SZsizeof(__m128i)+ +static int +ioat_dev_configure(const
structrte_rawdev*dev,rte_rawdev_obj_tconfig) +{ + structrte_ioat_rawdev_config*params = config; + structrte_ioat_rawdev
*ioat =dev->dev_private; + char mz_name[RTE_MEMZONE_NAMESIZE]; +
unsigned short i; + + if (dev->started) + return-EBUSY; + + if (params == NULL) + return-EINVAL; + + if
(params->ring_size> 4096 ||params->ring_size< 64 || +
!rte_is_power_of_2(params->ring_size)) + return-EINVAL; + +ioat->ring_size=params->ring_size; + if (ioat->desc_ring!=
NULL) { +rte_memzone_free(ioat->desc_mz); +ioat->desc_ring= NULL; +ioat->desc_mz= NULL; + } + + /*
allocate one block of memory for both descriptors + * and completion
handles. + */ + snprintf(mz_name,sizeof(mz_name),
"rawdev%u_desc_ring",dev->dev_id); +ioat->desc_mz=
rte_memzone_reserve(mz_name, + (DESC_SZ+COMPLETION_SZ) *ioat->ring_size, +dev->device->numa_node,RTE_MEMZONE_IOVA_CONTIG); + if (ioat->desc_mz
== NULL) + return-ENOMEM; +ioat->desc_ring=ioat->desc_mz->addr; +ioat->hdls= (void
*)&ioat->desc_ring[ioat->ring_size]; + +ioat->ring_addr=ioat->desc_mz->iova; + + /* configure descriptor ring - each one
points to next */ + for (i = 0; i <ioat->ring_size; i++) { +
ioat->desc_ring[i].next =ioat->ring_addr+ +
(((i + 1) %ioat->ring_size) *DESC_SZ); + } + + return 0; +} +
+static int +ioat_dev_start(structrte_rawdev*dev) +{ + structrte_ioat_rawdev*ioat =dev->dev_private; + + if (ioat->ring_size
== 0 ||ioat->desc_ring== NULL) + return-EBUSY; + + /*
inform hardware of where the descriptor ring is */ +ioat->regs->chainaddr=ioat->ring_addr; + /* inform hardware of where
to write the status/completions */ +ioat->regs->chancmp=ioat->status_addr; + + /* prime the status register to be set to
the last element */ +ioat->status=ioat->ring_addr+
((ioat->ring_size- 1) *DESC_SZ); + return 0; +} + +static void
+ioat_dev_stop(structrte_rawdev*dev) +{ +RTE_SET_USED(dev); +} +
static voidioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info){ @@ -44,11 +119,17 @@ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info)cfg->ring_size=ioat->ring_size; }Quoted 1 line
+extern int
ioat_rawdev_test(uint16_t dev_id); +This signature doesn't match with the
rte_rawdev_opsdefinition (which should be int(*func)(void)).
Quoted 8 lines
It will generate an error as such:
../drivers/raw/ioat/ioat_rawdev.c:176:20: error: initialization from incompatible pointer type [-Werror=incompatible-pointer-types] .dev_selftest =ioat_rawdev_test, ^../drivers/raw/ioat/ioat_rawdev.c:176:20: note: (near initialization for ‘ioat_rawdev_ops.dev_selftest’)
That was changed by commit 863fd2930bbe ("rawdev: pass the device id as parameter to selftest") which was part of the patchset on which this set depended.
/Bruce
Re: [dpdk-dev] [PATCH v5 7/9] raw/ioat: add configure, start and stop functions
Quoted 130 lines · Bruce Richardson
Quoted 121 lines · Aaron Conole
Quoted 114 lines · Bruce Richardson
Allow initializing a driver instance. Include selftest to validate these functions.
Signed-off-by Bruce Richardson <[email protected]> Acked-by:Anatoly Burakov <[email protected]> Tested-by: Harry van Haaren <[email protected]> ---
V4: Guarantee correct SUCCESS/FAILURE return values from test function Use memzone rather than malloc for ring allocation so we can guarantee contiguous memory. V3: don't add a new descriptor format struct, reuse from
rte_ioat_spec.hV2: test cases placed in self-test routine ---app/test/test_rawdev.c | 3 +- doc/guides/rawdevs/ioat_rawdev.rst | 32 +++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev.c | 82 +++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 41 +++++++++++++++ drivers/raw/ioat/meson.build | 3 +- drivers/raw/ioat/rte_ioat_rawdev.h | 2 + 7 files changed, 162insertions(+), 2 deletions(-) create mode 100644
drivers/raw/ioat/ioat_rawdev_test.capp/test/test_rawdev.c index renamed from app/test/test_rawdev.c +0
-0
623117af9..524a9d5f3100644 ---a/app/test/test_rawdev.c+++b/app/test/test_rawdev.c@@ -36,7 +36,8 @@test_rawdev_selftest_ioat(void)structrte_rawdev_infoinfo = { .dev_private = NULL }; if (rte_rawdev_info_get(i, &info)== 0 &&strstr(info.driver_name, "ioat")!= NULL) - returnTEST_SUCCESS; + returnrte_rawdev_selftest(i)== 0
? +TEST_SUCCESS:TEST_FAILED;
}app/test/test_rawdev.c index renamed from app/test/test_rawdev.c +1
-0printf("No IOAT rawdev found, skipping tests\n"); diff --git a/doc/guides/rawdevs/ioat_rawdev.rst b/doc/guides/rawdevs/ioat_rawdev.rst index efed9c64c..a0594d2cb 100644 --- a/doc/guides/rawdevs/ioat_rawdev.rst +++ b/doc/guides/rawdevs/ioat_rawdev.rst @@ -117,3 +117,35 @@ the ``dev_private`` field in the ``rte_rawdev_info`` struct should either be NULL, or else be set to point to a structure of type ``rte_ioat_rawdev_config``, in which case the size of the configured device input ring will be returned in that structure. + +Device Configuration +~~~~~~~~~~~~~~~~~~~~~ + +Configuring an IOAT rawdev device is done using the +``rte_rawdev_configure()`` API, which takes the same structure parameters +as the, previously referenced, ``rte_rawdev_info_get()`` API. The main +difference is that, because the parameter is used as input rather than +output, the ``dev_private`` structure element cannot be NULL, and must +point to a valid ``rte_ioat_rawdev_config`` structure, containing the ring +size to be used by the device. The ring size must be a power of two, +between 64 and 4096. + +The following code shows how the device is configured in +``test_ioat_rawdev.c``: + +.. code-block:: C + + #define IOAT_TEST_RINGSIZE 512 + struct rte_ioat_rawdev_config p = { .ring_size = -1 }; + struct rte_rawdev_info info = { .dev_private = &p }; + + /* ... */ + + p.ring_size = IOAT_TEST_RINGSIZE; + if (rte_rawdev_configure(dev_id, &info) != 0) { + printf("Error with rte_rawdev_configure()\n"); + return -1; + } + +Once configured, the device can then be made ready for use by calling the +``rte_rawdev_start()`` API. diff --git a/drivers/raw/ioat/Makefile b/drivers/raw/ioat/Makefile index 1e10938f3..b1af9c666 100644 --- a/drivers/raw/ioat/Makefile +++ b/drivers/raw/ioat/Makefile @@ -21,6 +21,7 @@ EXPORT_MAP := rte_pmd_ioat_version.map # library source files SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev.c +SRCS-$(CONFIG_RTE_LIBRTE_PMD_IOAT_RAWDEV) += ioat_rawdev_test.c # export include files SYMLINK-y-include += rte_ioat_rawdev.h diff --git a/drivers/raw/ioat/ioat_rawdev.c b/drivers/raw/ioat/ioat_rawdev.c index 2bfe2544d..0c91b3579 100644 --- a/drivers/raw/ioat/ioat_rawdev.c +++ b/drivers/raw/ioat/ioat_rawdev.c @@ -34,6 +34,81 @@ static struct rte_pci_driver ioat_pmd_drv; #define IOAT_PMD_ERR(fmt, args...) IOAT_PMD_LOG(ERR, fmt, ## args) #define IOAT_PMD_WARN(fmt, args...) IOAT_PMD_LOG(WARNING, fmt, ## args) +#define DESC_SZ sizeof(struct rte_ioat_generic_hw_desc) +#define
COMPLETION_SZsizeof(__m128i)+ +static int +ioat_dev_configure(const
structrte_rawdev*dev,rte_rawdev_obj_tconfig) +{ + structrte_ioat_rawdev_config*params = config; + structrte_ioat_rawdev
*ioat =dev->dev_private; + char mz_name[RTE_MEMZONE_NAMESIZE]; +
unsigned short i; + + if (dev->started) + return-EBUSY; + + if (params == NULL) + return-EINVAL; + + if
(params->ring_size> 4096 ||params->ring_size< 64 || +
!rte_is_power_of_2(params->ring_size)) + return-EINVAL; + +ioat->ring_size=params->ring_size; + if (ioat->desc_ring!=
NULL) { +rte_memzone_free(ioat->desc_mz); +ioat->desc_ring= NULL; +ioat->desc_mz= NULL; + } + + /*
allocate one block of memory for both descriptors + * and completion
handles. + */ + snprintf(mz_name,sizeof(mz_name),
"rawdev%u_desc_ring",dev->dev_id); +ioat->desc_mz=
rte_memzone_reserve(mz_name, + (DESC_SZ+COMPLETION_SZ) *ioat->ring_size, +dev->device->numa_node,RTE_MEMZONE_IOVA_CONTIG); + if (ioat->desc_mz
== NULL) + return-ENOMEM; +ioat->desc_ring=ioat->desc_mz->addr; +ioat->hdls= (void
*)&ioat->desc_ring[ioat->ring_size]; + +ioat->ring_addr=ioat->desc_mz->iova; + + /* configure descriptor ring - each one
points to next */ + for (i = 0; i <ioat->ring_size; i++) { +
ioat->desc_ring[i].next =ioat->ring_addr+ +
(((i + 1) %ioat->ring_size) *DESC_SZ); + } + + return 0; +} +
+static int +ioat_dev_start(structrte_rawdev*dev) +{ + structrte_ioat_rawdev*ioat =dev->dev_private; + + if (ioat->ring_size
== 0 ||ioat->desc_ring== NULL) + return-EBUSY; + + /*
inform hardware of where the descriptor ring is */ +ioat->regs->chainaddr=ioat->ring_addr; + /* inform hardware of where
to write the status/completions */ +ioat->regs->chancmp=ioat->status_addr; + + /* prime the status register to be set to
the last element */ +ioat->status=ioat->ring_addr+
((ioat->ring_size- 1) *DESC_SZ); + return 0; +} + +static void
+ioat_dev_stop(structrte_rawdev*dev) +{ +RTE_SET_USED(dev); +} +
static voidioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info){ @@ -44,11 +119,17 @@ioat_dev_info_get(struct rte_rawdev *dev, rte_rawdev_obj_t dev_info)cfg->ring_size=ioat->ring_size; }Quoted 1 line
+extern int
ioat_rawdev_test(uint16_t dev_id); +This signature doesn't match with the
rte_rawdev_opsdefinition (which should be int(*func)(void)).Quoted 8 lines
It will generate an error as such:
../drivers/raw/ioat/ioat_rawdev.c:176:20: error: initialization from incompatible pointer type [-Werror=incompatible-pointer-types] .dev_selftest =ioat_rawdev_test, ^../drivers/raw/ioat/ioat_rawdev.c:176:20: note: (near initialization for ‘ioat_rawdev_ops.dev_selftest’)
Quoted 3 lines
That was changed by commit
863fd2930bbe("rawdev: pass the device id as parameter to selftest") which was part of the patchset on which this set depended.
Strange - the robot didn't pick that commit up. Was this submitted before that commit was live?
Quoted 1 line
/Bruce
PATCH v5 8/9 raw/ioat: add statistics functions
In reply to Bruce RichardsonAdd stats functions to track what is happening in the driver, and put unit tests to check those.
---
V4: pulled in memory leak fixes from next patch
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 14 +++++++++ drivers/raw/ioat/ioat_rawdev.c | 44 ++++++++++++++++++++++++++++ drivers/raw/ioat/ioat_rawdev_test.c | 45 +++++++++++++++++++++++++++++ drivers/raw/ioat/rte_ioat_rawdev.h | 6 ++++ 4 files changed, 109 insertions(+)
doc/guides/rawdevs/ioat_rawdev.rst +14 -0
@@ -149,3 +149,17 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. + +Querying Device Statistics +~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The statistics from the IOAT rawdev device can be got via the xstats +functions in the ``rte_rawdev`` library, i.e. +``rte_rawdev_xstats_names_get()``, ``rte_rawdev_xstats_get()`` and +``rte_rawdev_xstats_by_name_get``. The statistics returned for each device +instance are: + +* ``failed_enqueues`` +* ``successful_enqueues`` +* ``copies_started`` +* ``copies_completed``
drivers/raw/ioat/ioat_rawdev.c +44 -0
62 lines not on this page · raw message
drivers/raw/ioat/ioat_rawdev_test.c +45 -0
62 lines not on this page · raw message
drivers/raw/ioat/rte_ioat_rawdev.h +6 -0
@@ -54,6 +54,12 @@ struct rte_ioat_rawdev {struct rte_ioat_generic_hw_desc *desc_ring;__m128i *hdls;/* completion handles for returning to user */ +/* some statistics for tracking, if added/changed update xstats fns*/ + uint64_t enqueue_failed __rte_cache_aligned; +uint64_t enqueued; +uint64_t started; +uint64_t completed; +/* to report completions, the device will write status back here */ volatile uint64_t status __rte_cache_aligned; };
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 8/9] raw/ioat: add statistics functions
Quoted 19 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:12 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 8/9] raw/ioat: add statistics functionsAdd stats functions to track what is happening in the driver, and put unit tests to check those.
Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V4: pulled in memory leak fixes from next patch
V2: test cases moved to self-test routine
PATCH v5 9/9 raw/ioat: add local API to perform copies
In reply to Bruce RichardsonAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
---
V3: updated to use descriptor format in rte_ioat_spec.h
V2: test cases moved to self-test routine
--- doc/guides/rawdevs/ioat_rawdev.rst | 100 ++++++++++++++++ drivers/raw/ioat/Makefile | 1 + drivers/raw/ioat/ioat_rawdev_test.c | 155 ++++++++++++++++++++++++- drivers/raw/ioat/meson.build | 2 +- drivers/raw/ioat/rte_ioat_rawdev.h | 169 +++++++++++++++++++++++++++- 5 files changed, 422 insertions(+), 5 deletions(-)
doc/guides/rawdevs/ioat_rawdev.rst +100 -0
@@ -150,6 +150,106 @@ The following code shows how the device is configured in Once configured, the device can then be made ready for use by calling the ``rte_rawdev_start()`` API. +Performing Data Copies +~~~~~~~~~~~~~~~~~~~~~~~ + +To perform data copies using IOAT rawdev devices, the functions +``rte_ioat_enqueue_copy()`` and ``rte_ioat_do_copies()`` should be used. +Once copies have been completed, the completion will be reported back when +the application calls ``rte_ioat_completed_copies()``. + +The ``rte_ioat_enqueue_copy()`` function enqueues a single copy to the +device ring for copying at a later point. The parameters to that function +include the physical addresses of both the source and destination buffers, +as well as two "handles" to be returned to the user when the copy is +completed. These handles can be arbitrary values, but two are provided so +that the library can track handles for both source and destination on +behalf of the user, e.g. virtual addresses for the buffers, or mbuf +pointers if packet data is being copied. + +While the ``rte_ioat_enqueue_copy()`` function enqueues a copy operation on +the device ring, the copy will not actually be performed until after the +application calls the ``rte_ioat_do_copies()`` function. This function +informs the device hardware of the elements enqueued on the ring, and the +device will begin to process them. It is expected that, for efficiency +reasons, a burst of operations will be enqueued to the device via multiple +enqueue calls between calls to the ``rte_ioat_do_copies()`` function. + +The following code from ``test_ioat_rawdev.c`` demonstrates how to enqueue +a burst of copies to the device and start the hardware processing of them: + +.. code-block:: C + + struct rte_mbuf *srcs[32], *dsts[32]; + unsigned int j; + + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data; + + srcs[i] = rte_pktmbuf_alloc(pool); + dsts[i] = rte_pktmbuf_alloc(pool); + srcs[i]->data_len = srcs[i]->pkt_len = length; + dsts[i]->data_len = dsts[i]->pkt_len = length; + src_data = rte_pktmbuf_mtod(srcs[i], char *); + + for (j = 0; j < length; j++) + src_data[j] = rand() & 0xFF; + + if (rte_ioat_enqueue_copy(dev_id, + srcs[i]->buf_iova + srcs[i]->data_off, + dsts[i]->buf_iova + dsts[i]->data_off, + length, + (uintptr_t)srcs[i], + (uintptr_t)dsts[i], + 0 /* nofence */) != 1) { + printf("Error with rte_ioat_enqueue_copy for buffer %u\n", + i); + return -1; + } + } + rte_ioat_do_copies(dev_id); + +To retrieve information about completed copies, the API +``rte_ioat_completed_copies()`` should be used. This API will return to the +application a set of completion handles passed in when the relevant copies +were enqueued. + +The following code from ``test_ioat_rawdev.c`` shows the test code +retrieving information about the completed copies and validating the data +is correct before freeing the data buffers using the returned handles: + +.. code-block:: C + + if (rte_ioat_completed_copies(dev_id, 64, (void *)completed_src, + (void *)completed_dst) != RTE_DIM(srcs)) { + printf("Error with rte_ioat_completed_copies\n"); + return -1; + } + for (i = 0; i < RTE_DIM(srcs); i++) { + char *src_data, *dst_data; + + if (completed_src[i] != srcs[i]) { + printf("Error with source pointer %u\n", i); + return -1; + } + if (completed_dst[i] != dsts[i]) { + printf("Error with dest pointer %u\n", i); + return -1; + } + + src_data = rte_pktmbuf_mtod(srcs[i], char *); + dst_data = rte_pktmbuf_mtod(dsts[i], char *); + for (j = 0; j < length; j++) + if (src_data[j] != dst_data[j]) { + printf("Error with copy of packet %u, byte %u\n", + i, j); + return -1; + } + rte_pktmbuf_free(srcs[i]); + rte_pktmbuf_free(dsts[i]); + } + + Querying Device Statistics ~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/raw/ioat/Makefile +1 -0
@@ -12,6 +12,7 @@ CFLAGS += $(WERROR_FLAGS) LDLIBS += -lrte_eal -lrte_rawdev LDLIBS += -lrte_pci -lrte_bus_pci +LDLIBS += -lrte_mbuf -lrte_mempool # library version LIBABIVER := 1
drivers/raw/ioat/ioat_rawdev_test.c +152 -3
187 lines not on this page · raw message
drivers/raw/ioat/meson.build +1 -1
@@ -4,7 +4,7 @@ build = dpdk_conf.has('RTE_ARCH_X86') sources = files('ioat_rawdev.c', 'ioat_rawdev_test.c') -deps += ['rawdev', 'bus_pci'] +deps += ['rawdev', 'bus_pci', 'mbuf'] install_headers('rte_ioat_rawdev.h', 'rte_ioat_spec.h')
drivers/raw/ioat/rte_ioat_rawdev.h +168 -1
186 lines not on this page · raw message
Signature · 1 line
2.21.0
Re: [dpdk-dev] [PATCH v5 9/9] raw/ioat: add local API to perform copies
Quoted 22 lines
-----Original Message-----
From: Richardson, Bruce
Sent: Tuesday, July 2, 2019 10:13 PM
To: [email protected]
Cc: Hu, Jiayu <[email protected]>; Richardson, Bruce <[email protected]>; Burakov, Anatoly <[email protected]>; Van Haaren, Harry <[email protected]>
Subject: [PATCH v5 9/9] raw/ioat: add local API to perform copiesAdd local APIs to trigger data copies, and retrieve handle values once those copies are completed. Included are unit tests to validate the data is copies correctly.
Signed-off-by Bruce Richardson <[email protected]>Acked-by Anatoly Burakov <[email protected]>Tested-by Harry van Haaren <[email protected]>---
V3: updated to use descriptor format in
rte_ioat_spec.h
V2: test cases moved to self-test routine
---
Thomas Monjalon
Applied02/07/2019 16:12, Bruce Richardson:
Quoted 8 lines
This patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
[...]
Quoted 10 lines
Bruce Richardson (9):
rawdev: allow devices to skip extra memory allocation raw/ioat: add initial support for ioat rawdev driverusertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
Applied, thanks
Quoted 30 lines · Bruce Richardson
This patch series adds support for the Intel QuickData Technology device, part of the Intel I/O Acceleration Technology (Intel I/OAT). It is a raw device for allowing hardware DMA i.e. data copies in hardware.
Performing the copies in hardware can provide performance improvements for applications where the average copy size is reasonably large, e.g. 1k packets. For smaller packets, e.g. 64-256 bytes, offloading the copy may reduce performance due to the overhead of using hardware.
V5:
* updated doxygen comment for internal rawdev pmd APIV4:
* changed memory management to use contiguous memzones instead of malloc memory
* added missing device ids for BDX platforms
* other misc cleanup following review (see individual patches logs)V3:
* removed DPDK-specific structure for the descriptor format and reused the structure in the imported filerte_ioat_spec.hV2:
* moved tests to rawdev selftest function
* some checkpatch and other small cleanups
* added extra documentation details on supported hardware
* aligned the changes to dpdk-devbind with the changes in the NTB set for consistency
Quoted 10 lines
Bruce Richardson (9):
rawdev: allow devices to skip extra memory allocation raw/ioat: add initial support for ioat rawdev driverusertools/dpdk-devbind.py: add support for IOAT devices raw/ioat: add register definition file raw/ioat: create device on probe and destroy on release raw/ioat: add device info function
raw/ioat: add configure, start and stop functions raw/ioat: add statistics functions
raw/ioat: add local API to perform copies
Hi Bruce, Thomas,
Getting following build error with icc, both for 32-bit & 64-bit:
.../dpdk/drivers/raw/ioat/ioat_rawdev_test.c(28): (col. 9) error #13203: No EMMS instruction before call to function