FlexIO SDK Programming Guide
NVIDIA DOCA FlexIO SDK Programming Guide
This guide provides an overview and configuration instructions for DOCA FlexIO SDK API.
The datapath accelerator (DPA) processor is an auxiliary processor designed to accelerate packet processing and other datapath operations. The FlexIO SDK exposes an API for managing the device and executing native code over it.
The DPA processor is supported on NVIDIA® BlueField®-3 DPUs and later generations.
FlexIO SDK library exposes a few layers of functionality:
libflexio
– library for DPU-side operations. Mainly used for resource management.libflexio_dev
– library for DPA-side operations. Mainly used for data path implementation.libflexio_os
– library for DPA OS-level accesslibflexio_libc
– custom LibC for DPA
A typical application is composed of two parts: One running on the host machine or the DPU target and another running directly over the DPA.
For the driver API reference, refer to the NVIDIA DOCA Driver APIs Reference Manual.
DPA programs cannot create resources. The responsibility of creating resources, such as FlexIO process, thread, outbox and window, as well as queues for packet processing (completion, receive and send), lies on the DPU program. The relevant information should be communicated (copied) to the DPA side and the address of the copied information should be passed as an argument to the running thread.
6.1. Example
DPU side:
- Declare a variable to hold the DPA buffer address.
flexio_uintptr_t app_data_dpa_daddr;
- Allocate a buffer on the DPA side.
flexio_buf_dev_alloc(flexio_process, sizeof(struct my_app_data), &app_data_dpa_daddr);
- Copy application data to the DPA buffer.
flexio_host2dev_memcpy(flexio_process, (uintptr_t)app_data, sizeof(struct my_app_data), app_data_dpa_daddr);
struct my_app_data
should be common between the DPU and DPA applications so the DPA application can access the struct fields.The event handler should get the address to the DPA buffer with the copied data:
flexio_event_handler_create(flexio_process, net_entry_point, app_data_dpa_daddr, NULL, flexio_outbox, &app_ctx.net_event_handler)
DPA side:
__dpa_rpc__ uint64_t event_handler_init(uint64_t thread_arg)
{
struct my_app_data *app_data;
app_data = (my_app_data *)thread_arg;
...
}
As mentioned previously, the DPU program is responsible for allocating buffers on the DPA side (same as resources). The DPU program should allocate device memory in advance for the DPA program needs (e.g., queues data buffer and rings, buffers for the program functionality, etc). The DPU program is also responsible for releasing the allocated memory. For this purpose, the FlexIO SDK API exposes the following memory management functions:
flexio_status flexio_buf_dev_alloc(struct flexio_process *process, size_t buff_bsize, flexio_uintptr_t *dest_daddr_p);
flexio_status flexio_buf_dev_free(flexio_uintptr_t daddr_p);
flexio_status flexio_host2dev_memcpy(struct flexio_process *process, void *src_haddr, size_t buff_bsize, flexio_uintptr_t dest_daddr);
flexio_status flexio_buf_dev_memset(struct flexio_process *process, int value, size_t buff_bsize, flexio_uintptr_t dest_daddr);
7.1. Buffers/Rings For DPA Queues
The FlexIO SDK exposes an API for allocating work queues and completion queues for the DPA. This means that the DPA may have direct access and control over these queues, allowing it to create doorbells and access their memory.
When creating a FlexIO SDK queue, the user must pre-allocate and provide memory buffers for the queue's element buffer/ring. This buffer may be allocated on the DPU or the DPA memory.
To this end, the FlexIO SDK exposes the flexio_qmem
struct, which allows the user to provide the buffer address and type (DPA or DPU). For addresses on the DPU memory, the user must also provide the relevant DUMEM ID. But for addresses on the DPA memory, this is not needed as the relevant process DUMEM ID is used.
7.2. Memory Allocation Best Practices
To optimize process DUMEM memory allocation, it is recommended to use the following allocation sizes (or closest to it):
- Up to 1 page (4KB)
- 26 pages (256KB)
- 211 pages (8MB)
- 216 pages (256MB)
Using these sizes minimizes memory fragmentation over the process DUMEM heap. If other buffer sizes are required, it is recommended to round the allocation up to one of the listed sizes and use it for multiple buffers.
To ensure correctness, certain operations require a barrier. These barriers are exposed in the libflexio_os_mb.h
header file:
flexio_os_dma_to_dev_wr_barrier();
flexio_os_dma_to_dev_wr_order_barrier();
flexio_os_dma_from_dev_rd_barrier();
flexio_os_window_write_barrier();
flexio_os_window_read_barrier();
flexio_os_window_rw_barrier();
flexio_os_all_barrier();
Some cases that require a barrier are handled by the SDK such as outbox and window configuration and DPU memory access through the window.
10.1. Default Window/Outbox
The DPA event handler expects a DPA window and DPA outbox structs on creation. These are used as the default for the event handler thread. The user may choose to set one or both to NULL, in which case there will be no valid default value for one/both of them.
Upon thread invocation on the DPA side, the thread context is set for the provided default IDs. If at any point the outbox/window IDs are changed, then the thread context on the next invocation is restored to the default IDs. This means that the DPA Window MKey must be configured each time the thread is invoked, as it has no default value.
10.2. HART Management
DPA HARTs are the equivalent of a logical core. A DPA program must be assigned a HART in order to execute.
It is possible to set HART affinity for an event handler upon creation. This causes the event handler to execute its DPA program over specific HARTs (or a group of HARTs). DPA supports three types of affinity:
- None – select a HART from all available HARTs
- Strict – select only the specified HART (by ID)
- Group – select a HART from all the HARTs in the specified group (by ID, must be created in advance)
The affinity type and ID, if applicable, are passed to the event handler upon creation using the affinity
field of the flexio_event_handler_attr
struct.
Since application execution is divided between the DPU side and the DPA processor services, debugging may be somewhat challenging, especially considering the fact that the DPA side does not have a terminal allowing the use of the C stdio library printf services.
11.1. Using Device Prints API
Another logging option is to use FlexIO SDK infrastructure to write strings from the DPA side to the DPU side console or file. The DPU side's flexio.h
file provides the flexio_print_init
API call for initializing the required infrastructures to support this. Once initialized, the DPA side must have the thread context, which can be obtained by calling flexio_dev_get_thread_ctx
. flexio_dev_print
can then be called to write a string to the DPA side where it is directed to the console or a file, according to user configuration in the init stage.
It is important to call flexio_print_destroy()
when exiting the DPU application to ensure proper clean-up of the print mechanism resources.
Device prints use an internal QP for the communication between the DPA and the DPU. When running over an InfiniBand fabric, the user must ensure that the subnet is well-configured and that the relevant device's port is in active
state.
IMPORTANT! flexio_print_init
should only be called once per process. Calling it multiple times recreates print resources which may cause a resource leak and other malfunctions.
11.1.1. Printf Support
Only limited functionality is implemented for printf. Not all libc printf is supported. Please consult the following list for supported modifiers:
- Formats –
%c
,%s
,%d
,%ld
,%u
,%lu
,%i
,%li
,%x
,%hx
,%hxx
,%lx
,%X
,%lX
,%o
,%lo
,%p
,%%
- Flags –
.
,*
,-
,+
,#
- General supported modifiers:
- "0" padding
- Min/max characters in string
- General unsupported modifiers:
- Floating point modifiers –
%e
,%E
,%f
,%lf
,%LF
- Precision modifiers
- Floating point modifiers –
Please refer to NVIDIA DOCA FlexIO Sample Guide for more information about the API of this DOCA library.
Notice
This document is provided for information purposes only and shall not be regarded as a warranty of a certain functionality, condition, or quality of a product. NVIDIA Corporation nor any of its direct or indirect subsidiaries and affiliates (collectively: “NVIDIA”) make no representations or warranties, expressed or implied, as to the accuracy or completeness of the information contained in this document and assume no responsibility for any errors contained herein. NVIDIA shall have no liability for the consequences or use of such information or for any infringement of patents or other rights of third parties that may result from its use. This document is not a commitment to develop, release, or deliver any Material (defined below), code, or functionality.
NVIDIA reserves the right to make corrections, modifications, enhancements, improvements, and any other changes to this document, at any time without notice.
Customer should obtain the latest relevant information before placing orders and should verify that such information is current and complete.
NVIDIA products are sold subject to the NVIDIA standard terms and conditions of sale supplied at the time of order acknowledgement, unless otherwise agreed in an individual sales agreement signed by authorized representatives of NVIDIA and customer (“Terms of Sale”). NVIDIA hereby expressly objects to applying any customer general terms and conditions with regards to the purchase of the NVIDIA product referenced in this document. No contractual obligations are formed either directly or indirectly by this document.
NVIDIA products are not designed, authorized, or warranted to be suitable for use in medical, military, aircraft, space, or life support equipment, nor in applications where failure or malfunction of the NVIDIA product can reasonably be expected to result in personal injury, death, or property or environmental damage. NVIDIA accepts no liability for inclusion and/or use of NVIDIA products in such equipment or applications and therefore such inclusion and/or use is at customer’s own risk.
NVIDIA makes no representation or warranty that products based on this document will be suitable for any specified use. Testing of all parameters of each product is not necessarily performed by NVIDIA. It is customer’s sole responsibility to evaluate and determine the applicability of any information contained in this document, ensure the product is suitable and fit for the application planned by customer, and perform the necessary testing for the application in order to avoid a default of the application or the product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA product and may result in additional or different conditions and/or requirements beyond those contained in this document. NVIDIA accepts no liability related to any default, damage, costs, or problem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is contrary to this document or (ii) customer product designs.
No license, either expressed or implied, is granted under any NVIDIA patent right, copyright, or other NVIDIA intellectual property right under this document. Information published by NVIDIA regarding third-party products or services does not constitute a license from NVIDIA to use such products or services or a warranty or endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual property rights of the third party, or a license from NVIDIA under the patents or other intellectual property rights of NVIDIA.
Reproduction of information in this document is permissible only if approved in advance by NVIDIA in writing, reproduced without alteration and in full compliance with all applicable export laws and regulations, and accompanied by all associated conditions, limitations, and notices.
THIS DOCUMENT AND ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, “MATERIALS”) ARE BEING PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. TO THE EXTENT NOT PROHIBITED BY LAW, IN NO EVENT WILL NVIDIA BE LIABLE FOR ANY DAMAGES, INCLUDING WITHOUT LIMITATION ANY DIRECT, INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE, OR CONSEQUENTIAL DAMAGES, HOWEVER CAUSED AND REGARDLESS OF THE THEORY OF LIABILITY, ARISING OUT OF ANY USE OF THIS DOCUMENT, EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards customer for the products described herein shall be limited in accordance with the Terms of Sale for the product.
Trademarks
NVIDIA, the NVIDIA logo, and Mellanox are trademarks and/or registered trademarks of Mellanox Technologies Ltd. and/or NVIDIA Corporation in the U.S. and in other countries. The registered trademark Linux® is used pursuant to a sublicense from the Linux Foundation, the exclusive licensee of Linus Torvalds, owner of the mark on a world¬wide basis. Other company and product names may be trademarks of the respective companies with which they are associated.
Copyright
© 2022 NVIDIA Corporation & affiliates. All rights reserved.