- 04 Jun, 2021 1 commit
-
-
Florian Westphal authored
This exports SO_TIMESTAMP_* function for re-use by MPTCP. Without this there is too much copy & paste needed to support this from mptcp setsockopt path. Acked-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
Florian Westphal <fw@strlen.de> Signed-off-by:
Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
- 03 Jun, 2021 39 commits
-
-
Colin Ian King authored
The comparison of len < 0 is always false because len is a size_t. Fix this by making len a ssize_t instead. Addresses-Coverity: ("Unsigned compared against 0") Fixes: d3953819 ("netdevsim: Add max_vfs to bus_dev") Signed-off-by:Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Andreas Roeseler authored
Including <linux/in.h> and <netinet/in.h> in the dependencies breaks compilation of trinity due to multiple definitions. <linux/in.h> is only used in <linux/icmp.h> to provide the definition of the struct in_addr, but this can be substituted out by using the datatype __be32. Signed-off-by:
Andreas Roeseler <andreas.a.roeseler@gmail.com> Signed-off-by:
-
Nathan Chancellor authored
Clang warns that proto in rmnet_map_v5_checksum_uplink_packet() might be used uninitialized: drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c:283:14: warning: variable 'proto' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized] } else if (skb->protocol == htons(ETH_P_IPV6)) { ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c:295:36: note: uninitialized use occurs here check = rmnet_map_get_csum_field(proto, trans); ^~~~~ drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c:283:10: note: remove the 'if' if its condition is always true } else if (skb->protocol == htons(ETH_P_IPV6)) { ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c:270:11: note: initialize the variable 'proto' to silence this warning u8 proto; ^ = '\0' 1 warning generated. This is technically a false positive because there is an if statement above this one that checks skb->protocol for not being either ETH_P_IP{,V6}. However, it is more obvious to sink that into the if statement as an else branch, which makes the code clearer and fixes the warning. At the same time, move the "IS_ENABLED(CONFIG_IPV6)" into the else if condition so that the else branch of the preprocessor conditional can be shared, since there is no build failure with CONFIG_IPV6 disabled. Fixes: b6e5d27e ("net: ethernet: rmnet: Add support for MAPv5 egress packets") Link: https://github.com/ClangBuiltLinux/linux/issues/1390Signed-off-by:Nathan Chancellor <nathan@kernel.org> Signed-off-by:
-
Nathan Chancellor authored
clang points out that ret in ks8851_read_selftest() is set but unused: drivers/net/ethernet/micrel/ks8851_common.c:1028:6: warning: variable 'ret' set but not used [-Wunused-but-set-variable] int ret = 0; ^ 1 warning generated. The return code of this function has never been checked so just remove ret and make the function return void. Fixes: 3ba81f3e ("net: Micrel KS8851 SPI network driver") Suggested-by:Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Yu Kuai authored
Add description for parameters of htb_add_to_id_tree() to fix gcc W=1 warnings: net/sched/sch_htb.c:282: warning: Function parameter or member 'root' not described in 'htb_add_to_id_tree' net/sched/sch_htb.c:282: warning: Function parameter or member 'cl' not described in 'htb_add_to_id_tree' net/sched/sch_htb.c:282: warning: Function parameter or member 'prio' not described in 'htb_add_to_id_tree' Signed-off-by:
Yu Kuai <yukuai3@huawei.com> Signed-off-by:
-
Colin Ian King authored
The variable ret is being initialized with a value that is never read, it is being updated later on. The assignment is redundant and can be removed. Addresses-Coverity: ("Unused value") Signed-off-by: -
Russell King authored
Rather than open-coding the phy_modify_changed() sequence, use this helper in marvell_set_polarity(). Signed-off-by:
Russell King <rmk+kernel@armlinux.org.uk> Reviewed-by:
Marek Behún <kabel@kernel.org> Signed-off-by:
-
David S. Miller authored
Alex Elder says: ==================== net: ipa: support inline checksum offload Inline offload--required for checksum offload support on IPA version 4.5 and above--is now supported by the RMNet driver: https://lore.kernel.org/netdev/162259440606.2786.10278242816453240434.git-patchwork-notify@kernel.org/ Add support for it in the IPA driver, and revert the commit that disabled it pending acceptance of the RMNet code. ==================== Signed-off-by:
-
Alex Elder authored
This reverts commit c88c34fc. The RMNet driver now supports inline checksum offload. Signed-off-by:
Alex Elder <elder@linaro.org> Signed-off-by:
-
Alex Elder authored
Starting with IPA v4.5, IP payload checksum offload is implemented differently. Prior to v4.5, the IPA hardware appends an rmnet_map_dl_csum_trailer structure to each packet if checksum offload is enabled in the download direction (modem->AP). In the upload direction (AP->modem) a rmnet_map_ul_csum_header structure is prepended before each sent packet. Starting with IPA v4.5, checksum offload is implemented using a single new rmnet_map_v5_csum_header structure which sits between the QMAP header and the packet data. The same header structure is used in both directions. The new header contains a header type (CSUM_OFFLOAD); a checksum flag; and a flag indicating whether any other headers follow this one. The checksum flag indicates whether the hardware should compute (and insert) the checksum on a sent packet. On a received packet the checksum flag indicates whether the hardware confirms the checksum value in the payload is correct. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxDavid S. Miller authored
Saeed Mahameed says: ==================== This series provides misc updates for mlx5 drivers. For more information please see tag log below. Please pull and let me know if there is any problem. mlx5-updates-2021-06-03 This series contains misc updates for mlx5 driver 1) Alaa disables advanced features when kdump mode to save on memory 2) Jakub counts all link flap events 3) Meir adds support for IPoIB NDR speed 4) Various misc cleanup ==================== Signed-off-by: -
Íñigo Huguet authored
Signed-off-by:
Íñigo Huguet <ihuguet@redhat.com> Signed-off-by:
-
Íñigo Huguet authored
OFLD and CTRL TX queues can be stopped if there is no room in their DMA rings. If this happens, they're tried to be restarted later after having made some room in the corresponding ring. The tasks of restarting these queues were triggered using tasklets, but they can be replaced for workqueue works, getting them out of softirq context. This queues stop/restart probably doesn't happen often and they can be quite lengthy because they try to send all pending skbs. Moreover, given that probably the ring is not empty yet, so the DMA still has work to do, we don't need to be so fast to justify using tasklets/softirq instead of running in a thread. Signed-off-by:
-
Yuchung Cheng authored
This patch aims to improve the situation when reordering and loss are ocurring in the same flight of packets. Previously the reordering would first induce a spurious recovery, then the subsequent ACK may undo the cwnd (based on the timestamps e.g.). However the current loss recovery does not proceed to invoke RACK to install a reordering timer. If some packets are also lost, this may lead to a long RTO-based recovery. An example is https://groups.google.com/g/bbr-dev/c/OFHADvJbTEI The solution is to after reverting the recovery, always invoke RACK to either mount the RACK timer to fast retransmit after the reordering window, or restarts the recovery if new loss is identified. Hence it is possible the sender may go from Recovery to Disorder/Open to Recovery again in one ACK. Reported-by:
mingkun bian <bianmingkun@gmail.com> Signed-off-by:
Yuchung Cheng <ycheng@google.com> Signed-off-by:
Neal Cardwell <ncardwell@google.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
-
Kees Cook authored
Silence this warning by using strscpy_pad() directly: drivers/net/bonding/bond_main.c:4877:3: warning: 'strncpy' specified bound 16 equals destination size [-Wstringop-truncation] 4877 | strncpy(params->primary, primary, IFNAMSIZ); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Additionally replace other strncpy() uses, as it is considered deprecated: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-stringsReported-by:kernel test robot <lkp@intel.com> Link: https://lore.kernel.org/lkml/202102150705.fdR6obB0-lkp@intel.comAcked-by:
Jay Vosburgh <jay.vosburgh@canonical.com> Signed-off-by:
Kees Cook <keescook@chromium.org> Signed-off-by:
-
Kees Cook authored
Use strscpy_pad() instead of strncpy() which is considered deprecated: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-stringsSigned-off-by:
-
David S. Miller authored
Shai Malin says: ==================== NVMeTCP Offload ULP With the goal of enabling a generic infrastructure that allows NVMe/TCP offload devices like NICs to seamlessly plug into the NVMe-oF stack, this patch series introduces the nvme-tcp-offload ULP host layer, which will be a new transport type called "tcp-offload" and will serve as an abstraction layer to work with vendor specific nvme-tcp offload drivers. NVMeTCP offload is a full offload of the NVMeTCP protocol, this includes both the TCP level and the NVMeTCP level. The nvme-tcp-offload transport can co-exist with the existing tcp and other transports. The tcp offload was designed so that stack changes are kept to a bare minimum: only registering new transports. All other APIs, ops etc. are identical to the regular tcp transport. Representing the TCP offload as a new transport allows clear and manageable differentiation between the connections which should use the offload path and those that are not offloaded (even on the same device). The nvme-tcp-offload layers and API compared to nvme-tcp and nvme-rdma: * NVMe layer: * [ nvme/nvme-fabrics/blk-mq ] | (nvme API and blk-mq API) | | * Vendor agnostic transport layer: * [ nvme-rdma ] [ nvme-tcp ] [ nvme-tcp-offload ] | | | (Verbs) | | | | (Socket) | | | | | (nvme-tcp-offload API) | | | | | | * Vendor Specific Driver: * | | | [ qedr ] | | [ qede ] | [ qedn ] Performance: ============ With this implementation on top of the Marvell qedn driver (using the Marvell FastLinQ NIC), we were able to demonstrate the following CPU utilization improvement: On AMD EPYC 7402, 2.80GHz, 28 cores: - For 16K queued read IOs, 16jobs, 4qd (50Gbps line rate): Improved the CPU utilization from 15.1% with NVMeTCP SW to 4.7% with NVMeTCP offload. On Intel(R) Xeon(R) Gold 5122 CPU, 3.60GHz, 16 cores: - For 512K queued read IOs, 16jobs, 4qd (25Gbps line rate): Improved the CPU utilization from 16.3% with NVMeTCP SW to 1.1% with NVMeTCP offload. In addition, we were able to demonstrate the following latency improvement: - For 200K read IOPS (16 jobs, 16 qd, with fio rate limiter): Improved the average latency from 105 usec with NVMeTCP SW to 39 usec with NVMeTCP offload. Improved the 99.99 tail latency from 570 usec with NVMeTCP SW to 91 usec with NVMeTCP offload. The end-to-end offload latency was measured from fio while running against back end of null device. Upstream plan: ============== The RFC series "NVMeTCP Offload ULP and QEDN Device Driver" https://lore.kernel.org/netdev/20210531225222.16992-1-smalin@marvell.com/ was designed in a modular way so that part 1 (nvme-tcp-offload) and part 2 (qed) are independent and part 3 (qedn) depends on both parts 1+2. - Part 1 (RFC patch 1-8): NVMeTCP Offload ULP The nvme-tcp-offload patches, will be sent to 'linux-nvme@lists.infradead.org'. - Part 2 (RFC patches 9-15): QED NVMeTCP Offload The qed infrastructure, will be sent to 'netdev@vger.kernel.org'. Once part 1 and 2 are accepted: - Part 3 (RFC patches 16-27): QEDN NVMeTCP Offload The qedn patches, will be sent to 'linux-nvme@lists.infradead.org'. Marvell is fully committed to maintain, test, and address issues with the new nvme-tcp-offload layer. Usage: ====== With the Marvell NVMeTCP offload design, the network-device (qede) and the offload-device (qedn) are paired on each port - Logically similar to the RDMA model. The user will interact with the network-device in order to configure the ip/vlan. The NVMeTCP configuration is populated as part of the nvme connect command. Example: Assign IP to the net-device (from any existing Linux tool): ip addr add 100.100.0.101/24 dev p1p1 This IP will be used by both net-device (qede) and offload-device (qedn). In order to connect from "sw" nvme-tcp through the net-device (qede): nvme connect -t tcp -s 4420 -a 100.100.0.100 -n testnqn In order to connect from "offload" nvme-tcp through the offload-device (qedn): nvme connect -t tcp_offload -s 4420 -a 100.100.0.100 -n testnqn An alternative approach, and as a future enhancement that will not impact this series will be to modify nvme-cli with a new flag that will determine if "-t tcp" should be the regular nvme-tcp (which will be the default) or nvme-tcp-offload. Exmaple: nvme connect -t tcp -s 4420 -a 100.100.0.100 -n testnqn -[new flag] Queue Initialization Design: ============================ The nvme-tcp-offload ULP module shall register with the existing nvmf_transport_ops (.name = "tcp_offload"), nvme_ctrl_ops and blk_mq_ops. The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following ops: - claim_dev() - in order to resolve the route to the target according to the paired net_dev. - create_queue() - in order to create offloaded nvme-tcp queue. The nvme-tcp-offload ULP module shall manage all the controller level functionalities, call claim_dev and based on the return values shall call the relevant module create_queue in order to create the admin queue and the IO queues. IO-path Design: =============== The nvme-tcp-offload shall work at the IO-level - the nvme-tcp-offload ULP module shall pass the request (the IO) to the nvme-tcp-offload vendor driver and later, the nvme-tcp-offload vendor driver returns the request completion (the IO completion). No additional handling is needed in between; this design will reduce the CPU utilization as we will describe below. The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following IO-path ops: - send_req() - in order to pass the request to the handling of the offload driver that shall pass it to the vendor specific device. - poll_queue() Once the IO completes, the nvme-tcp-offload vendor driver shall call command.done() that will invoke the nvme-tcp-offload ULP layer to complete the request. TCP events: =========== The Marvell FastLinQ NIC HW engine handle all the TCP re-transmissions and OOO events. Teardown and errors: ==================== In case of NVMeTCP queue error the nvme-tcp-offload vendor driver shall call the nvme_tcp_ofld_report_queue_err. The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following teardown ops: - drain_queue() - destroy_queue() The Marvell FastLinQ NIC HW engine: ==================================== The Marvell NIC HW engine is capable of offloading the entire TCP/IP stack and managing up to 64K connections per PF, already implemented and upstream use cases for this include iWARP (by the Marvell qedr driver) and iSCSI (by the Marvell qedi driver). In addition, the Marvell NIC HW engine offloads the NVMeTCP queue layer and is able to manage the IO level also in case of TCP re-transmissions and OOO events. The HW engine enables direct data placement (including the data digest CRC calculation and validation) and direct data transmission (including data digest CRC calculation). The Marvell qedn driver: ======================== The new driver will be added under "drivers/nvme/hw" and will be enabled by the Kconfig "Marvell NVM Express over Fabrics TCP offload". As part of the qedn init, the driver will register as a pci device driver and will work with the Marvell fastlinQ NIC. As part of the probe, the driver will register to the nvme_tcp_offload (ULP) and to the qed module (qed_nvmetcp_ops) - similar to other "qed_*_ops" which are used by the qede, qedr, qedf and qedi device drivers. nvme-tcp-offload Future work: ============================= - NVMF_OPT_HOST_IFACE Support. Changes since RFC v1: ===================== - nvme-tcp-offload: Fix nvme_tcp_ofld_ops return values. - nvme-tcp-offload: Remove NVMF_TRTYPE_TCP_OFFLOAD. - nvme-tcp-offload: Add nvme_tcp_ofld_poll() implementation. - nvme-tcp-offload: Fix nvme_tcp_ofld_queue_rq() to check map_sg() and send_req() return values. Changes since RFC v2: ===================== - nvme-tcp-offload: Fixes in controller and queue level (patches 3-6). - qedn: Add the Marvell's NVMeTCP HW offload vendor driver init and probe (patches 8-11). Changes since RFC v3: ===================== - nvme-tcp-offload: Add the full implementation of the nvme-tcp-offload layer including the new ops: setup_ctrl(), release_ctrl(), commit_rqs() and new flows (ASYNC and timeout). - nvme-tcp-offload: Add device maximums: max_hw_sectors, max_segments. - nvme-tcp-offload: layer design and optimization changes. Changes since RFC v4: ===================== (Many thanks to Hannes Reinecke for his feedback) - nvme_tcp_offload: Add num_hw_vectors in order to limit the number of queues. - nvme_tcp_offload: Add per device private_data. - nvme_tcp_offload: Fix header digest, data digest and tos initialization. Changes since RFC v5: ===================== (Many thanks to Sagi Grimberg for his feedback) - nvme-fabrics: Expose nvmf_check_required_opts() globally (as a new patch). - nvme_tcp_offload: Remove io-queues BLK_MQ_F_BLOCKING. - nvme_tcp_offload: Fix the nvme_tcp_ofld_stop_queue (drain_queue) flow. - nvme_tcp_offload: Fix the nvme_tcp_ofld_free_queue (destroy_queue) flow. - nvme_tcp_offload: Change rwsem to mutex. - nvme_tcp_offload: remove redundant fields. - nvme_tcp_offload: Remove the "new" from setup_ctrl(). - nvme_tcp_offload: Remove the init_req() and commit_rqs() ops. - nvme_tcp_offload: Minor fixes in nvme_tcp_ofld_create_ctrl() ansd nvme_tcp_ofld_free_queue(). - nvme_tcp_offload: Patch 8 (timeout and async) was squeashed into patch 7 (io level). Changes since RFC v6: ===================== - No changes in nvme_tcp_offload (only in qedn). ==================== Signed-off-by: -
Dean Balandin authored
In this patch, we present the IO level functionality. The nvme-tcp-offload shall work on the IO-level, meaning the nvme-tcp-offload ULP module shall pass the request to the nvme-tcp-offload vendor driver and shall expect for the request completion. No additional handling is needed in between, this design will reduce the CPU utilization as we will describe below. The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following IO-path ops: - send_req - in order to pass the request to the handling of the offload driver that shall pass it to the vendor specific device - poll_queue The vendor driver will manage the context from which the request will be executed and the request aggregations. Once the IO completed, the nvme-tcp-offload vendor driver shall call command.done() that shall invoke the nvme-tcp-offload ULP layer for completing the request. This patch also add support for the nvme-tcp-offload timeout and nvme-tcp-offload ASYNC flow. Acked-by:
Igor Russkikh <irusskikh@marvell.com> Signed-off-by:
Dean Balandin <dbalandin@marvell.com> Signed-off-by:
Prabhakar Kushwaha <pkushwaha@marvell.com> Signed-off-by:
Omkar Kulkarni <okulkarni@marvell.com> Signed-off-by:
Michal Kalderon <mkalderon@marvell.com> Signed-off-by:
Ariel Elior <aelior@marvell.com> Signed-off-by:
Shai Malin <smalin@marvell.com> Reviewed-by:
Hannes Reinecke <hare@suse.de> Reviewed-by:
Himanshu Madhani <himanshu.madhani@oracle.com> Signed-off-by:
-
Dean Balandin authored
In this patch we implement queue level functionality. The implementation is similar to the nvme-tcp module, the main difference being that we call the vendor specific create_queue op which creates the TCP connection, and NVMeTPC connection including icreq+icresp negotiation. Once create_queue returns successfully, we can move on to the fabrics connect. Acked-by:
-
Arie Gershberg authored
In this patch, we implement controller level error handling and recovery. Upon an error discovered by the ULP or reset controller initiated by the nvme-core (using reset_ctrl workqueue), the ULP will initiate a controller recovery which includes teardown and re-connect of all queues. Acked-by:
Arie Gershberg <agershberg@marvell.com> Signed-off-by:
-
Arie Gershberg authored
In this patch we implement controller level functionality including: - create_ctrl. - delete_ctrl. - free_ctrl. The implementation is similar to other nvme fabrics modules, the main difference being that the nvme-tcp-offload ULP calls the vendor specific claim_dev() op with the given TCP/IP parameters to determine which device will be used for this controller. Once found, the vendor specific device and controller will be paired and kept in a controller list managed by the ULP. Acked-by:
-
Dean Balandin authored
As part of create_ctrl(), it scans the registered devices and calls the claim_dev op on each of them, to find the first devices that matches the connection params. Once the correct devices is found (claim_dev returns true), we raise the refcnt of that device and return that device as the device to be used for ctrl currently being created. Acked-by:
-
Prabhakar Kushwaha authored
nvmf_check_required_opts() is used to check if user provided opts has the required_opts or not. if not, it will log which options are not provided. It can be leveraged by nvme-tcp-offload to check if provided opts are supported by this specific vendor driver or not. So expose nvmf_check_required_opts() globally. Acked-by:
-
Prabhakar Kushwaha authored
Move NVMF_ALLOWED_OPTS and NVMF_REQUIRED_OPTS definitions to header file, so it can be used by the different HW devices. NVMeTCP offload devices might have different limitations of the allowed options, for example, a device that does not support all the queue types. With tcp and rdma, only the nvme-tcp and nvme-rdma layers handle those attributes and the HW devices do not create any limitations for the allowed options. An alternative design could be to add separate fields in nvme_tcp_ofld_ops such as max_hw_sectors and max_segments that we already have in this series. Acked-by:
Sagi Grimberg <sagi@grimberg.me> Signed-off-by:
-
Shai Malin authored
This patch will present the structure for the NVMeTCP offload common layer driver. This module is added under "drivers/nvme/host/" and future offload drivers which will register to it will be placed under "drivers/nvme/hw". This new driver will be enabled by the Kconfig "NVM Express over Fabrics TCP offload commmon layer". In order to support the new transport type, for host mode, no change is needed. Each new vendor-specific offload driver will register to this ULP during its probe function, by filling out the nvme_tcp_ofld_dev->ops and nvme_tcp_ofld_dev->private_data and calling nvme_tcp_ofld_register_dev with the initialized struct. The internal implementation: - tcp-offload.h: Includes all common structs and ops to be used and shared by offload drivers. - tcp-offload.c: Includes the init function which registers as a NVMf transport just like any other transport. Acked-by:
-
David S. Miller authored
Jon Maloy says: ==================== tipc: some small cleanups We make some minor code cleanups and improvements. v2: Changed value of TIPC_ANY_SCOPE macro in patch #3 to avoid compiler warning ==================== Signed-off-by: -
Jon Maloy authored
We introduce a new macro TIPC_ANY_SCOPE to make the handling of the lookup scope value more comprehensible during multicast reception. The (unchanged) rules go as follows: 1) Multicast messages sent from own node are delivered to all matching sockets on the own node, irrespective of their binding scope. 2) Multicast messages sent from other nodes arrive here because they have found TIPC_CLUSTER_SCOPE bindings emanating from this node. Those messages should be delivered to exactly those sockets, but not to local sockets bound with TIPC_NODE_SCOPE, since the latter obviously were not meant to be visible for those senders. 3) Group multicast/broadcast messages are delivered to the sockets with a binding scope matching exactly the lookup scope indicated in the message header, and nobody else. Reviewed-by:
Xin Long <lucien.xin@gmail.com> Tested-by:
Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by:
Jon Maloy <jmaloy@redhat.com> Signed-off-by:
-
Jon Maloy authored
We refactor tipc_sk_anc_data_recv() to make it slightly more comprehensible, but also to facilitate application of some additions to the code in a future commit. Reviewed-by:
-
Jon Maloy authored
We eliminate the redundant fields conn_type and conn_instance in struct tipc_sock. On the connecting side, this information is already present in the unused (after the connection is established) part of the pre-allocated header, and on the accepting side, we put it there when the new socket is created. Reviewed-by:
-
David S. Miller authored
Shai Malin says: ==================== QED NVMeTCP Offload Intro: ====== This is the qed part of Marvell’s NVMeTCP offload series, shared as RFC series "NVMeTCP Offload ULP and QEDN Device Drive". This part is a standalone series, and is not dependent on other parts of the RFC. The overall goal is to add qedn as the offload driver for NVMeTCP, alongside the existing offload drivers (qedr, qedi and qedf for rdma, iscsi and fcoe respectively). In this series we are making the necessary changes to qed to enable this by exposing APIs for FW/HW initializations. The qedn series (and required changes to NVMe stack) will be sent to the linux-nvme mailing list. I have included more details on the upstream plan under section with the same name below. The Series Patches: =================== 1. qed: Add TCP_ULP FW resource layout – replacing iSCSI when common with NVMeTCP. 2. qed: Add NVMeTCP Offload PF Level FW and HW HSI. 3. qed: Add NVMeTCP Offload Connection Level FW and HW HSI. 4. qed: Add support of HW filter block – enables redirecting NVMeTCP traffic to the dedicated PF. 5. qed: Add NVMeTCP Offload IO Level FW and HW HSI. 6. qed: Add NVMeTCP Offload IO Level FW Initializations. 7. qed: Add IP services APIs support –VLAN, IP routing and reserving TCP ports for the offload device. The NVMeTCP Offload: ==================== With the goal of enabling a generic infrastructure that allows NVMe/TCP offload devices like NICs to seamlessly plug into the NVMe-oF stack, this patch series introduces the nvme-tcp-offload ULP host layer, which will be a new transport type called "tcp-offload" and will serve as an abstraction layer to work with vendor specific nvme-tcp offload drivers. NVMeTCP offload is a full offload of the NVMeTCP protocol, this includes both the TCP level and the NVMeTCP level. The nvme-tcp-offload transport can co-exist with the existing tcp and other transports. The tcp offload was designed so that stack changes are kept to a bare minimum: only registering new transports. All other APIs, ops etc. are identical to the regular tcp transport. Representing the TCP offload as a new transport allows clear and manageable differentiation between the connections which should use the offload path and those that are not offloaded (even on the same device). The nvme-tcp-offload layers and API compared to nvme-tcp and nvme-rdma: * NVMe layer: * [ nvme/nvme-fabrics/blk-mq ] | (nvme API and blk-mq API) | | * Vendor agnostic transport layer: * [ nvme-rdma ] [ nvme-tcp ] [ nvme-tcp-offload ] | | | (Verbs) | | | | (Socket) | | | | | (nvme-tcp-offload API) | | | | | | * Vendor Specific Driver: * | | | [ qedr ] | | [ qede ] | [ qedn ] Performance: ============ With this implementation on top of the Marvell qedn driver (using the Marvell FastLinQ NIC), we were able to demonstrate the following CPU utilization improvement: On AMD EPYC 7402, 2.80GHz, 28 cores: - For 16K queued read IOs, 16jobs, 4qd (50Gbps line rate): Improved the CPU utilization from 15.1% with NVMeTCP SW to 4.7% with NVMeTCP offload. On Intel(R) Xeon(R) Gold 5122 CPU, 3.60GHz, 16 cores: - For 512K queued read IOs, 16jobs, 4qd (25Gbps line rate): Improved the CPU utilization from 16.3% with NVMeTCP SW to 1.1% with NVMeTCP offload. In addition, we were able to demonstrate the following latency improvement: - For 200K read IOPS (16 jobs, 16 qd, with fio rate limiter): Improved the average latency from 105 usec with NVMeTCP SW to 39 usec with NVMeTCP offload. Improved the 99.99 tail latency from 570 usec with NVMeTCP SW to 91 usec with NVMeTCP offload. The end-to-end offload latency was measured from fio while running against back end of null device. The Marvell FastLinQ NIC HW engine: ==================================== The Marvell NIC HW engine is capable of offloading the entire TCP/IP stack and managing up to 64K connections per PF, already implemented and upstream use cases for this include iWARP (by the Marvell qedr driver) and iSCSI (by the Marvell qedi driver). In addition, the Marvell NIC HW engine offloads the NVMeTCP queue layer and is able to manage the IO level also in case of TCP re-transmissions and OOO events. The HW engine enables direct data placement (including the data digest CRC calculation and validation) and direct data transmission (including data digest CRC calculation). The Marvell qedn driver: ======================== The new driver will be added under "drivers/nvme/hw" and will be enabled by the Kconfig "Marvell NVM Express over Fabrics TCP offload". As part of the qedn init, the driver will register as a pci device driver and will work with the Marvell fastlinQ NIC. As part of the probe, the driver will register to the nvme_tcp_offload (ULP) and to the qed module (qed_nvmetcp_ops) - similar to other "qed_*_ops" which are used by the qede, qedr, qedf and qedi device drivers. Upstream Plan: ============= The RFC series "NVMeTCP Offload ULP and QEDN Device Driver" https://lore.kernel.org/netdev/20210531225222.16992-1-smalin@marvell.com/ was designed in a modular way so that part 1 (nvme-tcp-offload) and part 2 (qed) are independent and part 3 (qedn) depends on both parts 1+2. - Part 1 (RFC patch 1-8): NVMeTCP Offload ULP The nvme-tcp-offload patches, will be sent to 'linux-nvme@lists.infradead.org'. - Part 2 (RFC patches 9-15): QED NVMeTCP Offload The qed infrastructure, will be sent to 'netdev@vger.kernel.org'. Once part 1 and 2 are accepted: - Part 3 (RFC patches 16-27): QEDN NVMeTCP Offload The qedn patches, will be sent to 'linux-nvme@lists.infradead.org'. ==================== Signed-off-by: -
Nikolay Assa authored
This patch introduces APIs which the NVMeTCP Offload device (qedn) will use through the paired net-device (qede). It includes APIs for: - ipv4/ipv6 routing - get VLAN from net-device - TCP ports reservation Acked-by:
Nikolay Assa <nassa@marvell.com> Signed-off-by:
-
Shai Malin authored
This patch introduces the NVMeTCP FW initializations which is used to initialize the IO level configuration into a per IO HW resource ("task") as part of the IO path flow. This includes: - Write IO FW initialization - Read IO FW initialization. - IC-Req and IC-Resp FW exchange. - FW Cleanup flow (Flush IO). Acked-by: -
Shai Malin authored
This patch introduces the NVMeTCP Offload FW and HW HSI in order to initialize the IO level configuration into a per IO HW resource ("task") as part of the IO path flow. Acked-by: -
Prabhakar Kushwaha authored
This patch introduces the functionality of HW filter block. It adds and removes filters based on source and target TCP port. It also add functionality to clear all filters at once. Acked-by:
-
Shai Malin authored
This patch introduces the NVMeTCP HSI and HSI functionality in order to initialize and interact with the HW device as part of the connection level HSI. This includes: - Connection offload: offload a TCP connection to the FW. - Connection update: update the ICReq-ICResp params - Connection clear SQ: outstanding IOs FW flush. - Connection termination: terminate the TCP connection and flush the FW. Acked-by:
-
Shai Malin authored
This patch introduces the NVMeTCP device and PF level HSI and HSI functionality in order to initialize and interact with the HW device. The patch also adds qed NVMeTCP personality. This patch is based on the qede, qedr, qedi, qedf drivers HSI. Acked-by:
-
Omkar Kulkarni authored
Add TCP_ULP as a storage common TCP offload FW resource layout. This will be used by the core driver (QED) for both the NVMeTCP and iSCSI. Acked-by:
-
Krzysztof Kozlowski authored
In two places the 'ep_desc' and 'skb' local variables are used only within if() or for() block, so they scope can be reduced which makes the entire code slightly easier to follow. No functional change. Suggested-by:
Joe Perches <joe@perches.com> Signed-off-by:
Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com> Signed-off-by:
-
Krzysztof Kozlowski authored
The "continue" statement at the end of a for loop does not have an effect. Entire loop contents can be slightly simplified to increase code readability. No functional change. Suggested-by:
-
