From: Oded Gabbay <oded.gabbay@gmail.com>
To: linux-kernel@vger.kernel.org, netdev@vger.kernel.org
Cc: SW_Drivers@habana.ai, gregkh@linuxfoundation.org,
davem@davemloft.net, kuba@kernel.org
Subject: [PATCH 00/15] Adding GAUDI NIC code to habanalabs driver
Date: Thu, 10 Sep 2020 19:11:11 +0300 [thread overview]
Message-ID: <20200910161126.30948-1-oded.gabbay@gmail.com> (raw)
This patch-set adds support for initializing and using the GAUDI NIC ports,
functioning as scale-out interconnect when doing distributed Deep Learning
training. The training can be performed over tens of thousands of GAUDIs
and it is done using the RDMA-over-converged-Ethernet (RoCE) v2 protocol.
Each GAUDI exposes 10x100GbE ports that are designed to scale-out the
inter-GAUDI communication by integrating a complete communication engine
on-die. This native integration allows users to use the same scaling
technology, both inside the server and rack (termed as scale-up), as well
as for scaling across racks (scale-out). The racks can be connected
directly between GAUDI processors, or through any number of standard
Ethernet switches.
The driver exposes the NIC ports to the user as standard Ethernet ports by
registering each port to the networking subsystem. This allows the user to
manage the ports with standard tools such as ifconfig, ethtool, etc. It
also enables us to connect to the Linux networking stack and thus support
standard networking protocols, such as IPv4, IPv6, TCP, etc. In addition,
we can also leverage protocols such as DCB for dynamically configuring
priorities to avoid congestion.
For each NIC port there is a matching QMAN entity. For RoCE, the user
submits workloads to the NIC through the QMAN, same as he does for the
compute engines. For regular Ethernet, the user sends and receives packets
through the standard Ethernet sockets. Those sockets are used only as a
control path. The data path that is used for AI training goes through the
RoCE interface.
It is important to note that there are some limitations and uniqueness
in GAUDI's NIC H/W, compared to other networking adapters that enforced us
to use a less-than-common driver design:
1. The NIC functionality is NOT exposed as different PCI Physical
Functions. There is a single PF which is used for compute and
networking, as the main goal of the NIC ports is to be used as
intra-communication and not as standard network interfaces. This
implies we can't connect different drivers to handle the networking
ports because it is the same device, from the kernel POV, as the
compute. Therefore, we must integrate the networking code into the
main habanalabs driver.
2. Although our communication engine implements RDMA, and the driver code
uses well-known RDMA concepts such as QP context, CQ, WQ, etc., the
GAUDI architecture does NOT support other basic IBverbs concepts, such
as MR and protection domain. Therefore, we can't connect to the standard
IBverb infrastructure in the user-space and kernel (rdma-core library
and infiniband subsystem, respectively) because the standard RDMA s/w
and tools won't work on our H/W. Instead, we added a new IOCTL to the
driver's existing IOCTL API. The new IOCTL exposes the available
NIC control operations to the user (e.g. Create a QP context).
3. The die-on communication engine provides minimal offloading for standard
Ethernet and TCP/IP protocols, as those are only used for control plane.
E.g. the packets are copied rather than using descriptors.
Therefore, the Ethernet performance is quite low compared to standard
Ethernet adapters.
4. There is no virtualization support per port.
Most or all of the above limitations will hopefully be improved in future
ASIC generations.
Patch-set organization:
- Patches 1 & 2 are just adding some auto-generated register header files
and NIC-related definitions to the interface between the driver and the
GAUDI firmware.
- Patch 3 adds initialization of security restrictions on the NIC engines.
- Patch 4 adds initialization of the NIC QMANs. The QMANs are needed to
send RDMA packets through the NIC engines.
- Patches 5-11 adds the NIC driver code. It contains the basic Ethernet
driver and H/W initialization, the NIC PHY driver code and the new NIC
control IOCTL operations.
- Patch 12-14 adds support for debugfs, ethtool and DCB.
- Patch 15 adds the implementation of the high-level init/fini functions
and their calls from the common code. This is the patch that actually
enables the NIC ports and allows the user to work with them.
Thanks,
Oded
Omer Shpigelman (15):
habanalabs/gaudi: add NIC H/W and registers definitions
habanalabs/gaudi: add NIC firmware-related definitions
habanalabs/gaudi: add NIC security configuration
habanalabs/gaudi: add support for NIC QMANs
habanalabs/gaudi: add NIC Ethernet support
habanalabs/gaudi: add NIC PHY code
habanalabs/gaudi: allow user to get MAC addresses in INFO IOCTL
habanalabs/gaudi: add a new IOCTL for NIC control operations
habanalabs/gaudi: add CQ control operations
habanalabs/gaudi: add WQ control operations
habanalabs/gaudi: add QP error handling
habanalabs/gaudi: add debugfs entries for the NIC
habanalabs/gaudi: Add ethtool support using coresight
habanalabs/gaudi: support DCB protocol
habanalabs/gaudi: add NIC init/fini calls from common code
.../ABI/testing/debugfs-driver-habanalabs | 69 +
drivers/misc/habanalabs/common/context.c | 1 +
drivers/misc/habanalabs/common/device.c | 24 +-
drivers/misc/habanalabs/common/firmware_if.c | 44 +
drivers/misc/habanalabs/common/habanalabs.h | 33 +-
.../misc/habanalabs/common/habanalabs_drv.c | 11 +
.../misc/habanalabs/common/habanalabs_ioctl.c | 151 +-
drivers/misc/habanalabs/common/pci.c | 1 +
drivers/misc/habanalabs/gaudi/Makefile | 4 +
drivers/misc/habanalabs/gaudi/gaudi.c | 958 +++-
drivers/misc/habanalabs/gaudi/gaudiP.h | 333 +-
.../misc/habanalabs/gaudi/gaudi_coresight.c | 144 +
drivers/misc/habanalabs/gaudi/gaudi_nic.c | 4063 +++++++++++++++++
drivers/misc/habanalabs/gaudi/gaudi_nic.h | 354 ++
.../misc/habanalabs/gaudi/gaudi_nic_dcbnl.c | 108 +
.../misc/habanalabs/gaudi/gaudi_nic_debugfs.c | 402 ++
.../misc/habanalabs/gaudi/gaudi_nic_ethtool.c | 582 +++
drivers/misc/habanalabs/gaudi/gaudi_phy.c | 1272 ++++++
.../misc/habanalabs/gaudi/gaudi_security.c | 3973 ++++++++++++++++
drivers/misc/habanalabs/goya/goya.c | 44 +
.../misc/habanalabs/include/common/cpucp_if.h | 34 +-
.../include/gaudi/asic_reg/gaudi_regs.h | 26 +-
.../include/gaudi/asic_reg/nic0_qm0_masks.h | 800 ++++
.../include/gaudi/asic_reg/nic0_qm0_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic0_qm1_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic0_qpc0_masks.h | 500 ++
.../include/gaudi/asic_reg/nic0_qpc0_regs.h | 710 +++
.../include/gaudi/asic_reg/nic0_qpc1_regs.h | 710 +++
.../include/gaudi/asic_reg/nic0_rxb_regs.h | 508 +++
.../include/gaudi/asic_reg/nic0_rxe0_masks.h | 354 ++
.../include/gaudi/asic_reg/nic0_rxe0_regs.h | 158 +
.../include/gaudi/asic_reg/nic0_rxe1_regs.h | 158 +
.../include/gaudi/asic_reg/nic0_stat_regs.h | 518 +++
.../include/gaudi/asic_reg/nic0_tmr_regs.h | 184 +
.../include/gaudi/asic_reg/nic0_txe0_masks.h | 336 ++
.../include/gaudi/asic_reg/nic0_txe0_regs.h | 264 ++
.../include/gaudi/asic_reg/nic0_txe1_regs.h | 264 ++
.../include/gaudi/asic_reg/nic0_txs0_masks.h | 336 ++
.../include/gaudi/asic_reg/nic0_txs0_regs.h | 214 +
.../include/gaudi/asic_reg/nic0_txs1_regs.h | 214 +
.../include/gaudi/asic_reg/nic1_qm0_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic1_qm1_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic2_qm0_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic2_qm1_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic3_qm0_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic3_qm1_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic4_qm0_regs.h | 834 ++++
.../include/gaudi/asic_reg/nic4_qm1_regs.h | 834 ++++
drivers/misc/habanalabs/include/gaudi/gaudi.h | 12 +
.../habanalabs/include/gaudi/gaudi_fw_if.h | 24 +
.../habanalabs/include/gaudi/gaudi_masks.h | 15 +
.../include/hw_ip/nic/nic_general.h | 13 +
include/uapi/misc/habanalabs.h | 296 +-
53 files changed, 27497 insertions(+), 62 deletions(-)
create mode 100644 drivers/misc/habanalabs/gaudi/gaudi_nic.c
create mode 100644 drivers/misc/habanalabs/gaudi/gaudi_nic.h
create mode 100644 drivers/misc/habanalabs/gaudi/gaudi_nic_dcbnl.c
create mode 100644 drivers/misc/habanalabs/gaudi/gaudi_nic_debugfs.c
create mode 100644 drivers/misc/habanalabs/gaudi/gaudi_nic_ethtool.c
create mode 100644 drivers/misc/habanalabs/gaudi/gaudi_phy.c
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_qm0_masks.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_qm0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_qm1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_qpc0_masks.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_qpc0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_qpc1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_rxb_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_rxe0_masks.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_rxe0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_rxe1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_stat_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_tmr_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_txe0_masks.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_txe0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_txe1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_txs0_masks.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_txs0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic0_txs1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic1_qm0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic1_qm1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic2_qm0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic2_qm1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic3_qm0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic3_qm1_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic4_qm0_regs.h
create mode 100644 drivers/misc/habanalabs/include/gaudi/asic_reg/nic4_qm1_regs.h
create mode 100644 drivers/misc/habanalabs/include/hw_ip/nic/nic_general.h
--
2.17.1
next reply other threads:[~2020-09-10 16:16 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-10 16:11 Oded Gabbay [this message]
2020-09-10 16:11 ` [PATCH 02/15] habanalabs/gaudi: add NIC firmware-related definitions Oded Gabbay
2020-09-10 16:11 ` [PATCH 03/15] habanalabs/gaudi: add NIC security configuration Oded Gabbay
2020-09-10 16:11 ` [PATCH 04/15] habanalabs/gaudi: add support for NIC QMANs Oded Gabbay
2020-09-10 16:11 ` [PATCH 05/15] habanalabs/gaudi: add NIC Ethernet support Oded Gabbay
2020-09-10 20:03 ` Jakub Kicinski
2020-09-10 20:18 ` Oded Gabbay
2020-09-14 9:52 ` Omer Shpigelman
2020-09-14 16:47 ` Jakub Kicinski
2020-09-10 16:11 ` [PATCH 06/15] habanalabs/gaudi: add NIC PHY code Oded Gabbay
2020-09-10 16:11 ` [PATCH 07/15] habanalabs/gaudi: allow user to get MAC addresses in INFO IOCTL Oded Gabbay
2020-09-10 16:11 ` [PATCH 08/15] habanalabs/gaudi: add a new IOCTL for NIC control operations Oded Gabbay
2020-09-10 16:11 ` [PATCH 09/15] habanalabs/gaudi: add CQ " Oded Gabbay
2020-09-10 16:11 ` [PATCH 10/15] habanalabs/gaudi: add WQ " Oded Gabbay
2020-09-10 16:11 ` [PATCH 11/15] habanalabs/gaudi: add QP error handling Oded Gabbay
2020-09-10 16:11 ` [PATCH 12/15] habanalabs/gaudi: add debugfs entries for the NIC Oded Gabbay
2020-09-10 20:01 ` Jakub Kicinski
2020-09-10 20:10 ` Oded Gabbay
2020-09-10 20:16 ` Jakub Kicinski
2020-09-10 20:17 ` Oded Gabbay
2020-09-10 20:30 ` Jakub Kicinski
2020-09-10 20:33 ` Oded Gabbay
2020-09-14 13:48 ` Omer Shpigelman
2020-09-14 16:50 ` Jakub Kicinski
2020-09-15 12:57 ` Oded Gabbay
2020-09-16 16:38 ` Jakub Kicinski
2020-09-10 16:11 ` [PATCH 13/15] habanalabs/gaudi: Add ethtool support using coresight Oded Gabbay
2020-09-10 20:19 ` Andrew Lunn
2020-09-10 20:22 ` Oded Gabbay
2020-09-10 16:11 ` [PATCH 14/15] habanalabs/gaudi: support DCB protocol Oded Gabbay
2020-09-10 16:11 ` [PATCH 15/15] habanalabs/gaudi: add NIC init/fini calls from common code Oded Gabbay
2020-09-10 20:01 ` [PATCH 00/15] Adding GAUDI NIC code to habanalabs driver Jakub Kicinski
2020-09-10 20:16 ` Oded Gabbay
2020-09-10 20:25 ` Andrew Lunn
2020-09-10 20:30 ` Oded Gabbay
2020-09-10 20:38 ` Andrew Lunn
2020-09-10 20:52 ` Oded Gabbay
2020-09-11 6:22 ` Greg Kroah-Hartman
2020-09-10 20:28 ` Jakub Kicinski
2020-09-10 20:32 ` Oded Gabbay
2020-09-10 21:05 ` Florian Fainelli
2020-09-10 21:15 ` Oded Gabbay
2020-09-10 21:23 ` Florian Fainelli
-- strict thread matches above, loose matches on Subject: below --
2020-09-10 15:03 Oded Gabbay
2020-09-10 15:54 ` Greg KH
2020-09-10 15:59 ` Oded Gabbay
2020-09-10 16:08 ` Greg KH
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200910161126.30948-1-oded.gabbay@gmail.com \
--to=oded.gabbay@gmail.com \
--cc=SW_Drivers@habana.ai \
--cc=davem@davemloft.net \
--cc=gregkh@linuxfoundation.org \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).