- 05 Sep, 2018 18 commits
-
-
Jack Morgenstein authored
The upstream kernel commit cited below modified the workqueue in the new CQ API to be bound to a specific CPU (instead of being unbound). This caused ALL users of the new CQ API to use the same bound WQ. Specifically, MAD handling was severely delayed when the CPU bound to the WQ was busy handling (higher priority) interrupts. This caused a delay in the MAD "heartbeat" response handling, which resulted in ports being incorrectly classified as "down". To fix this, add a new "unbound" WQ type to the new CQ API, so that users have the option to choose either a bound WQ or an unbound WQ. For MADs, choose the new "unbound" WQ. Fixes: b7363e67 ("IB/device: Convert ib-comp-wq to be CPU-bound") Signed-off-by:
Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by:
Leon Romanovsky <leonro@mellanox.com> Reviewed-by:
Sagi Grimberg <sagi@grimberg.m> Signed-off-by:
Jason Gunthorpe <jgg@mellanox.com>
-
Joe Perches authored
Consistently use the "QPLIB: " prefix for dev_<level> logging. Miscellanea: o Add missing newlines to avoid possible message interleaving o Coalesce consecutive dev_<level> uses that emit a message header to avoid < 80 column lengths and mistakenly output on multiple lines o Reflow modified lines to use 80 columns where appropriate o Consistently use "%s: " where __func__ is output o QPLIB: is now always output immediately after the dev_<level> header Signed-off-by:
Joe Perches <joe@perches.com> Acked-by:
Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by:
-
Jason Gunthorpe authored
For dependencies, branch based on 'mellanox/mlx5-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux.git Pull Flow actions to mutate packets from Leon Romanovsky: ==================== This series exposes the ability to create flow actions which can mutate packet headers. We do that by exposing two new verbs: * modify header - can change existing packet headers. packet * reformat - can encapsulate or decapsulate a packet. Once created a flow action must be attached to a steering rule for it to take effect. The first 10 patches refactor mlx5_core code, rename internal structures to better reflect their operation and export needed functions so the RDMA side can allocate the action. The last 5 patches expose via the IOCTL infrastructure mlx5_ib methods which do the actual allocation of resources and return an handle to the user. A user of this API is expected to know how to work with the device's spec as the input to those function is HW depended. An example usage of the modify header action is routing, A user can create an action which edits the L2 header and decrease the TTL. An example usage of the packet reformat action is VXLAN encap/decap which is done by the HW. ==================== * branch 'mlx5-flow-mutate': RDMA/mlx5: Extend packet reformat verbs RDMA/mlx5: Add new flow action verb - packet reformat RDMA/uverbs: Add generic function to fill in flow action object RDMA/mlx5: Add a new flow action verb - modify header RDMA/uverbs: Add UVERBS_ATTR_CONST_IN to the specs language net/mlx5: Export packet reformat alloc/dealloc functions net/mlx5: Pass a namespace for packet reformat ID allocation net/mlx5: Expose new packet reformat capabilities {net, RDMA}/mlx5: Rename encap to reformat packet net/mlx5: Move header encap type to IFC header file net/mlx5: Break encap/decap into two separated flow table creation flags net/mlx5: Add support for more namespaces when allocating modify header net/mlx5: Export modify header alloc/dealloc functions net/mlx5: Add proper NIC TX steering flow tables support net/mlx5: Cleanup flow namespace getter switch logic Signed-off-by:
-
Mark Bloch authored
We expose new actions: L2_TO_L2_TUNNEL - A generic encap from L2 to L2, the data passed should be the encapsulating headers. L3_TUNNEL_TO_L2 - Will do decap where the inner packet starts from L3, the data should be mac or mac + vlan (14 or 18 bytes). L2_TO_L3_TUNNEL - Will do encap where is L2 of the original packet will not be included, the data should be the encapsulating header. Signed-off-by:
Mark Bloch <markb@mellanox.com> Signed-off-by:
-
Mark Bloch authored
For now, only add L2_TUNNEL_TO_L2 option. This will allow to perform generic decap operation if the encapsulating protocol is L2 based, and the inner packet is also L2 based. For example this can be used to decap VXLAN packets. Signed-off-by:
-
Mark Bloch authored
Refactor the initialization of a flow action object to a common function. Signed-off-by:
-
Mark Bloch authored
Expose the ability to create a flow action which changes packet headers. The data passed from userspace should be modify header actions as defined by HW specification. Signed-off-by:
-
Mark Bloch authored
This makes it clear and safe to access constants passed in from user space. We define a consistent ABI of u64 for all constants, and verify that the data passed in can be represented by the type the user supplies. The expectation is this will always be used with an enum declaring the constant values, and the user will use the enum type as input to the accessor. To retrieve the attribute value we introduce two helper calls - one standard which may fail if attribute is not valid and one where caller can provide a default value which will be used in case the attribute is not valid (useful when attribute is optional). Signed-off-by:
Ariel Levkovich <lariel@mellanox.com> Signed-off-by:
-
Mark Bloch authored
This will allow for the RDMA side to allocate packet reformat context. Signed-off-by:
Saeed Mahameed <saeedm@mellanox.com> Signed-off-by:
-
Mark Bloch authored
Currently we attach packet reformat actions only to the FDB namespace. In preparation to be able to use that for NIC steering, pass the actual namespace as a parameter. Signed-off-by:
Or Gerlitz <ogerlitz@mellanox.com> Reviewed-by:
-
Mark Bloch authored
Expose new abilities when creating a packet reformat context. The new types which can be created are: MLX5_REFORMAT_TYPE_L2_TO_L2_TUNNEL: Ability to create generic encap operation to be done by the HW. MLX5_REFORMAT_TYPE_L3_TUNNEL_TO_L2: Ability to create generic decap operation where the inner packet doesn't contain L2. MLX5_REFORMAT_TYPE_L2_TO_L3_TUNNEL: Ability to create generic encap operation to be done by the HW. The L2 of the original packet is dropped. Signed-off-by:
-
Mark Bloch authored
Renames all encap mlx5_{core,ib} code to use the new naming of packet reformat. This change doesn't introduce any function change and is needed to properly reflect the operation being done by this action. For example not only can we encapsulate a packet, but also decapsulate it. Signed-off-by: -
Mark Bloch authored
Those bits are hardware specification and should be defined in the IFC header file. Signed-off-by:
-
Mark Bloch authored
Today we are able to attach encap and decap actions only to the FDB. In preparation to enable those actions on the NIC flow tables, break the single flag into two. Those flags control whatever a decap or encap operations can be attached to the flow table created. For FDB, if encapsulation is required, we set both of them. Signed-off-by:
-
Mark Bloch authored
There are RX and TX flow steering namespaces with different number of actions. Initialize them accordingly. Signed-off-by:
-
Mark Bloch authored
Those functions will be used by the RDMA side to create modify header actions to be attached to flow steering rules via verbs. Signed-off-by:
-
Mark Bloch authored
Extend the ability to add steering rules to NIC TX flow tables. For now, we are only adding TX bypass (egress) which is used by the RDMA side. This will allow to shape outgoing traffic and tweak it if needed, for example performing encapsulation or rewriting headers. Signed-off-by:
-
Mark Bloch authored
Refactor the switch logic so it's simpler to follow and understand. Signed-off-by:
-
- 04 Sep, 2018 3 commits
-
-
Majd Dibbiny authored
In the current code, the TX affinity is per RoCE device, which can cause unfairness between different contexts. e.g. if we open two contexts, and each open 10 QPs concurrently, all of the QPs of the first context might end up on the first port instead of distributed on the two ports as expected To overcome this unfairness between processes, we maintain per device TX affinity, and per process TX affinity. The allocation algorithm is as follow: 1. Hold two tx_port_affinity atomic variables, one per RoCE device and one per ucontext. Both initialized to 0. 2. In mlx5_ib_alloc_ucontext do: 2.1. ucontext.tx_port_affinity = device.tx_port_affinity 2.2. device.tx_port_affinity += 1 3. In modify QP INIT2RST: 3.1. qp.tx_port_affinity = ucontext.tx_port_affinity % MLX5_PORT_NUM 3.2. ucontext.tx_port_affinity += 1 Signed-off-by:
Majd Dibbiny <majd@mellanox.com> Reviewed-by:
Moni Shoua <monis@mellanox.com> Signed-off-by:
-
Ariel Levkovich authored
Adding the alloc/dealloc memic FW command opcodes to avoid "unknown command" prints in the command string converter and internal error status handler. Signed-off-by:
-
Moni Shoua authored
The field atomic_mode is 4 bits wide and therefore can hold values from 0x0 to 0xf. Remove the unnecessary 20 bit shift that made the values be incorrect. While that, remove unused enum values. Fixes: 57cda166 ("net/mlx5: Add DCT command interface") Signed-off-by:
Artemy Kovalyov <artemyko@mellanox.com> Signed-off-by:
-
- 01 Sep, 2018 9 commits
-
-
Michael J. Ruhl authored
User contexts use the receive URGENT interrupt. However, enabling the IRQ SRC in the file_ops module is not as clean as it could be. Augment the _rcvctl() function to be able to enable/disable the IRQ source. Use the new interface from file_ops to enable/disable the IRQ. Reviewed-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by:
Sadanand Warrier <sadanand.warrier@intel.com> Signed-off-by:
Michael J. Ruhl <michael.j.ruhl@intel.com> Signed-off-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by:
Doug Ledford <dledford@redhat.com>
-
Michael J. Ruhl authored
The current IRQ API is an all or nothing interface. This has two problems: 1. All IRQs are enabled regardless of use 2. Moving from general interrupt to MSIx handling is difficult Introduce a new API to enable/disable specific IRQs or a range of IRQs. Do not enable and disable all IRQs in one step. Rework various modules to enable/disable IRQs when needed. Reviewed-by:
-
Kamenee Arumugam authored
Retry the PCIe link training up to 'pcie_retry' times if the PCIe link width is narrower than the previous width. Reviewed-by:
Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by:
Kamenee Arumugam <kamenee.arumugam@intel.com> Signed-off-by:
-
Michael J. Ruhl authored
The current method of allocating MSIx resources is a bit cumbersome, and not very easily added to. Refactor and re-order the code paths into a more consistent interface. Update the interface so that allocations are not order dependent. Reviewed-by:
-
Michael J. Ruhl authored
The current HFI1 MSIx API is difficult to follow, change, or add to. In anticipation of moving to an more flexible API, move the current MSIx functionality to the new msix.c module. Reviewed-by:
-
Michael J. Ruhl authored
Currently several things occur before the hfi1_devdata structure is allocated. This leads to an inconsistent logging ability and makes it more difficult to restructure some code paths. Allocate (and do a minimal init) the structure as soon as possible. Reviewed-by:
-
Michael J. Ruhl authored
The tune_pcie_caps needs to occur sometime after PCI is enabled, but before the HFI is enabled. Currently it is placed in the MSIx allocation code which doesn't really fit. Moving it to just after the gen3 bump. Clean up the associated code (modules, etc.). Reviewed-by:
-
Michael J. Ruhl authored
TXREQ defines are duplicated, incompletely, in the sdma header file. Remove duplicate defines. Reviewed-by:
-
Dennis Dalessandro authored
We want to keep files in alphabetical order in our makefile, however this just makes for messy diffs when adding (or removing) files. Let's just clean this up and make it line by line. Reviewed-by:
-
- 30 Aug, 2018 10 commits
-
-
Vijay Immanuel authored
When a read request is retried for the remaining partial data, the response may restart from read response first or read response only. So support those cases. Do not advance the comp psn beyond the current wqe's last_psn as that could skip over an entire read wqe and will cause the req_retry() logic to set an incorrect req psn. An example sequence is as follows: Write PSN 40 -- this is the current WQE. Read request PSN 41 Write PSN 42 Receive ACK PSN 42 -- this will complete the current WQE for PSN 40, and set the comp psn to 42 which is a problem because the read request at PSN 41 has been skipped over. So when req_retry() tries to retransmit the read request, it sets the req psn to 42 which is incorrect. When retrying a read request, calculate the number of psns completed based on the dma resid instead of the wqe first_psn. The wqe first_psn could have moved if the read request was retried multiple times. Set the reth length to the dma resid to handle read retries for the remaining partial data. Signed-off-by:
Vijay Immanuel <vijayi@attalasystems.com> Signed-off-by:
-
Vijay Immanuel authored
Error retries can occur due to timeouts, NAKs or receiving packets beyond the current read request. Avoid back-to-back retries due to packet processing, by only retrying the initial attempt immediately. Subsequent retries must be due to timeouts. Continue to process completion packets after scheduling a retry. Signed-off-by:
-
Vijay Immanuel authored
Don't reset the resp opcode for a replayed read response. The resp opcode could be in the middle of a write or send sequence, when the duplicate read request was received. An example sequence is as follows: - Receive read request for 12KB PSN 20. Transmit read response first, middle and last with PSNs 20,21,22. - Receive write first PSN 23. At this point the resp psn is 24 and resp opcode is write first. - The sender notices that PSN 20 is dropped and retransmits. Receive read request for 12KB PSN 20. Transmit read response first, middle and last with PSNs 20,21,22. The resp opcode is set to -1, the resp psn remains 24. - Receive write first PSN 23. This is processed by duplicate_request(). The resp opcode remains -1 and resp psn remains 24. - Receive write middle PSN 24. check_op_seq() reports a missing first error since the resp opcode is -1. When sending an ack for a duplicate send or write request, use the psn of the previous ack sent. Do not use the psn of a read response for the ack. An example sequence is as follows: - Receive write PSN 30. Transmit ACK for PSN 30. - Receive read request 4KB PSN 31. Transmit read response with PSN 31. The resp psn is now 32. - The sender notices that PSN 30 is dropped and retransmits. Receive write PSN 30. duplicate_request() sends an ACK with PSN 31. That is incorrect since PSN 31 was a read request. Signed-off-by:
-
Parav Pandit authored
Consolidate all error checks under single if() condition and use helper unlikely() macro for them, in addition drop unneeded goto labels. rxe_pool_get_index() already provides RB tree based efficient lookup. Avoid doing extra checks for error cases which are rare and already covered by rxe_pool_get_index(). Signed-off-by:
Parav Pandit <parav@mellanox.com> Reviewed-by:
Daniel Jurgens <danielj@mellanox.com> Signed-off-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
Parav Pandit authored
While performing lookup in a pool, if entry is found, take the reference right there, instead of checking again outside the loop and save one branch. Signed-off-by:
-
Parav Pandit authored
Normal practice is to have enum defines in capital letters. Signed-off-by:
-
Parav Pandit authored
Concurrent readers which read rb tree are protected using read lock. Concurrent writers which add element to pool are protected using write lock. Signed-off-by:
-
Parav Pandit authored
rxe_prepare() is called on an skb which has ndev already initialized by rxe_init_packet(). Therefore avoid querying the GID attribute again and use the available netdevice from the skb->dev. Signed-off-by:
Zhu Yanjun <yanjun.zhu@oracle.com> Signed-off-by:
-
Zhu Yanjun authored
In the commit 536ca245 ("IB/rxe: Drop QP0 silently"), if qpn is zero, the function directly returns. So in the following function, it is not necessary to check qpn. The qpn check in the function check_keys is removed. Fixes: 536ca245 ("IB/rxe: Drop QP0 silently") CC: Srinivas Eeda <srinivas.eeda@oracle.com> CC: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by:
-
Vijay Immanuel authored
Select the source udp port number for a QP based on the source QPN. This provides a better spread of traffic across NIC RX queues for RC/UC QPs. Signed-off-by:
-
