- 12 May, 2020 3 commits
-
-
Lang Cheng authored
It's easier to understand and maintain enable flags of qp using a single field in type of unsigned long than defining a field for every flags in the structure hns_roce_qp, and we can add new flags for features more conveniently in the future. Link: https://lore.kernel.org/r/1588674607-25337-4-git-send-email-liweihang@huawei.comSigned-off-by:
Lang Cheng <chenglang@huawei.com> Signed-off-by:
Weihang Li <liweihang@huawei.com> Signed-off-by:
Jason Gunthorpe <jgg@mellanox.com>
-
Weihang Li authored
12 bits is not enough for HIP08_C, so extend a new field in length of 16 bits for it. Link: https://lore.kernel.org/r/1588674607-25337-3-git-send-email-liweihang@huawei.comSigned-off-by:
-
Leon Romanovsky authored
The librdmacm uses node_guid as identifier to correlate between IB devices and CMA devices. However FW resets cause to such "connection" to be lost and require from the user to restart its application. Extend UCMA to return IB device index, which is stable identifier. Link: https://lore.kernel.org/r/20200504132541.355710-1-leon@kernel.orgSigned-off-by:
Leon Romanovsky <leonro@mellanox.com> Signed-off-by:
-
- 08 May, 2020 1 commit
-
-
Colin Ian King authored
The variable rcqe_sz is being unnecessarily assigned twice, fix this by removing one of the duplicates. Fixes: 8bde2c50 ("RDMA/mlx5: Update all DRIVER QP places to use QP subtype") Link: https://lore.kernel.org/r/20200507151610.52636-1-colin.king@canonical.com Addresses-Coverity: ("Evaluation order violation") Signed-off-by:
Colin Ian King <colin.king@canonical.com> Acked-by:
-
- 06 May, 2020 20 commits
-
-
Mark Bloch authored
When operating in switchdev mode or using devlink to disable RoCE only raw Ethernet QPs are allowed to be created. When in switchdev mode this can lead to passing an invalid port number as part of the modify qp firmware cmd and will lead to a syndrome reported back to the user, such as: * mlx5_cmd_check:803:(pid 50148): RST2INIT_QP(0x502) op_mod(0x0) failed, status bad parameter(0x3), syndrome (0x177405). Internal UD QP might be used to test for write combining support (even if externally we report RoCE as disabled) check for that specific flag and allow is specifically. Fixes: b5ca15ad ("IB/mlx5: Add proper representors support") Link: https://lore.kernel.org/r/20200506071602.7177-3-leon@kernel.orgSigned-off-by:
Mark Bloch <markb@mellanox.com> Signed-off-by:
-
Mark Bloch authored
Assign the profile to the IB device before executing stages. This will allow to check which profile is being used from within a stage. Link: https://lore.kernel.org/r/20200506071602.7177-2-leon@kernel.orgSigned-off-by:
-
Leon Romanovsky authored
Split qp.c by removing all WR logic to separate file. Link: https://lore.kernel.org/r/20200506065513.4668-4-leon@kernel.orgSigned-off-by:
-
Max Gurtovoy authored
Add small helpers in order to avoid code duplication and improve code readability. Decrease the amount of code in the gigantic post_send function and divide it to readable methods that will help in code maintenance in the future. Link: https://lore.kernel.org/r/20200506065513.4668-3-leon@kernel.orgSigned-off-by:
Max Gurtovoy <maxg@mellanox.com> Signed-off-by:
-
Leon Romanovsky authored
Reuse newly introduced mlx5_cmd_exec_in() and mlx5_cmd_exec_inout() to reduce code duplication in mlx5_ib module. Link: https://lore.kernel.org/r/20200506065513.4668-2-leon@kernel.orgSigned-off-by:
-
Wenpeng Liang authored
These caps are assigned in query_pf_caps() or set_default_caps(), and should not be assigned out of these two functions. Link: https://lore.kernel.org/r/1588242691-12913-4-git-send-email-liweihang@huawei.comSigned-off-by:
Wenpeng Liang <liangwenpeng@huawei.com> Signed-off-by:
-
Weihang Li authored
lp_pktn_ini means the number of loopback slice packets for long messages, it should depend on MTU(fixed to 4096B currently) and max size of SQ inline. Link: https://lore.kernel.org/r/1588242691-12913-3-git-send-email-liweihang@huawei.comSigned-off-by:
-
Weihang Li authored
There is a comments with some chinese semicolons that cause encoding issues each time hns_roc_hw_v2.h was modified from a IDE. So fix this by using correct symbols. Link: https://lore.kernel.org/r/1588242691-12913-2-git-send-email-liweihang@huawei.comSigned-off-by:
-
Xi Wang authored
Optimize the SRQ's WQE buffer parameters calculating process to make the codes more readable by using new functions about multi-hop addressing to calculating capabilities of SRQ. Link: https://lore.kernel.org/r/1588071823-40200-6-git-send-email-liweihang@huawei.comSigned-off-by:
Xi Wang <wangxi11@huawei.com> Signed-off-by:
-
Yixian Liu authored
Just move the SRQ related code to more reasonable place, and unify format of some prints. Link: https://lore.kernel.org/r/1588071823-40200-5-git-send-email-liweihang@huawei.comSigned-off-by:
Yixian Liu <liuyixian@huawei.com> Signed-off-by:
-
Xi Wang authored
Optimize the QP's WQE buffer parameters calculating process to make the codes more readable mainly by merging calculation of extended sge space of kernel and userspace. In addition, add some inline functions to simply codes about multi-hop addressing. Link: https://lore.kernel.org/r/1588071823-40200-4-git-send-email-liweihang@huawei.comSigned-off-by:
-
Xi Wang authored
The MTT (Memory Translate Table) interface is no longer used to configure the buffer address to BT (Base Address Table) that requires driver mapping. Because the MTT is not compatible with multi-hop addressing of the hip08, it is replaced by MTR (Memory Translate Region) interface, and all the MTT functions should be removed. Link: https://lore.kernel.org/r/1588071823-40200-3-git-send-email-liweihang@huawei.comSigned-off-by:
-
Xi Wang authored
PBL table has its own implementation for multi-hop addressing currently, but for the hardware, all table's addressing use the same logic, there is no need to implement repeatedly. So optimize the PBL buffer allocation process by using the mtr's interfaces. Link: https://lore.kernel.org/r/1588071823-40200-2-git-send-email-liweihang@huawei.comSigned-off-by:
kbuild test robot <lkp@intel.com> Signed-off-by:
-
Mark Zhang authored
Calculate UDP source port based on the grh.flow_label. If grh.flow_label is not valid, we will use minimal supported UDP source port. Link: https://lore.kernel.org/r/20200504051935.269708-6-leon@kernel.orgSigned-off-by:
Mark Zhang <markz@mellanox.com> Reviewed-by:
Maor Gottlieb <maorg@mellanox.com> Signed-off-by:
-
Mark Zhang authored
If flow label is not set by the user or it's not IPv4, initialize it with the cma src/dst based on the "Kernighan and Ritchie's hash function". Link: https://lore.kernel.org/r/20200504051935.269708-5-leon@kernel.orgSigned-off-by:
-
Mark Zhang authored
Calculate and set UDP source port based on the flow label. If flow label is not defined in GRH then calculate it based on lqpn/rqpn. Link: https://lore.kernel.org/r/20200504051935.269708-4-leon@kernel.orgSigned-off-by:
-
Maor Gottlieb authored
Use rdma_flow_label_to_udp_sport to calculate the UDP source port of the RoCEV2 packet. Link: https://lore.kernel.org/r/20200504051935.269708-3-leon@kernel.orgSigned-off-by:
-
Mark Zhang authored
Add two hash functions to distribute RoCE v2 UDP source and Flowlabel symmetrically. These are user visible API and any change in the implementation needs to be tested for inter-operability between old and new variant. Link: https://lore.kernel.org/r/20200504051935.269708-2-leon@kernel.orgSigned-off-by:
-
Jason Gunthorpe authored
When a client is added it isn't allowed to fail, but all the client's have various failure paths within their add routines. This creates the very fringe condition where the client was added, failed during add and didn't set the client_data. The core code will then still call other client_data centric ops like remove(), rename(), get_nl_info(), and get_net_dev_by_params() with NULL client_data - which is confusing and unexpected. If the add() callback fails, then do not call any more client ops for the device, even remove. Remove all the now redundant checks for NULL client_data in ops callbacks. Update all the add() callbacks to return error codes appropriately. EOPNOTSUPP is used for cases where the ULP does not support the ib_device - eg because it only works with IB. Link: https://lore.kernel.org/r/20200421172440.387069-1-leon@kernel.orgSigned-off-by:
Ursula Braun <ubraun@linux.ibm.com> Signed-off-by:
-
Maor Gottlieb authored
Snoop interface is not used. Remove it. Link: https://lore.kernel.org/r/20200413132408.931084-1-leon@kernel.orgSigned-off-by:
-
- 02 May, 2020 9 commits
-
-
Gal Pressman authored
Add a new stat that counts admin commands failures, which might help when debugging different issues. Link: https://lore.kernel.org/r/20200420062213.44577-4-galpress@amazon.comReviewed-by:
Daniel Kranzdorf <dkkranzd@amazon.com> Reviewed-by:
Yossi Leybovich <sleybo@amazon.com> Signed-off-by:
Gal Pressman <galpress@amazon.com> Signed-off-by:
-
Gal Pressman authored
Add a new stat that counts mmap failures, which might help when debugging different issues. Link: https://lore.kernel.org/r/20200420062213.44577-3-galpress@amazon.comReviewed-by:
Firas JahJah <firasj@amazon.com> Reviewed-by:
-
Gal Pressman authored
Create CQ errors are already being counted, report them along all other counters. Link: https://lore.kernel.org/r/20200420062213.44577-2-galpress@amazon.comReviewed-by:
-
Maor Gottlieb authored
The patch sets the lag tx affinity of the data QPs and the GSI QPs according to the LAG xmit slave. For GSI QPs, in case the link layer is Ethenet (RoCE) we create two GSI QPs, one for each physical port. When the driver selects the GSI QP, it will consider the port affinity result. For connected QPs, the driver sets the affinity of the xmit slave. The above, ensures that RC QP and it's corresponding GSI QP will transmit from the same physical port. Link: https://lore.kernel.org/r/20200430192146.12863-17-maorg@mellanox.comSigned-off-by:
-
Maor Gottlieb authored
Move affinity related code in modify qp to function. It's a preparation for next patch the extend the affinity calculation to consider the xmit slave. Link: https://lore.kernel.org/r/20200430192146.12863-16-maorg@mellanox.comSigned-off-by:
-
Maor Gottlieb authored
Add a call to rdma_lag_get_ah_roce_slave() when the address handle is created. Lower driver can use it to select the QP's affinity port. Link: https://lore.kernel.org/r/20200430192146.12863-15-maorg@mellanox.comSigned-off-by:
-
Maor Gottlieb authored
Add support to get the RoCE LAG xmit slave by building skb of the RoCE packet and call to master_get_xmit_slave. If driver wants to get the slave assume all slaves are available, then need to set RDMA_LAG_FLAGS_HASH_ALL_SLAVES in flags. Link: https://lore.kernel.org/r/20200430192146.12863-14-maorg@mellanox.comSigned-off-by:
-
Maor Gottlieb authored
Following patch adds additional argument to the create AH function, so it make sense to group ah_attr and flags arguments in struct. Link: https://lore.kernel.org/r/20200430192146.12863-13-maorg@mellanox.comSigned-off-by:
Devesh Sharma <devesh.sharma@broadcom.com> Acked-by:
-
Jason Gunthorpe authored
From the mlx5-next branch at git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux Required for dependencies in following patches * mellanox/mlx5-next: net/mlx5: Add support to get lag physical port net/mlx5: Change lag mutex lock to spin lock bonding: Implement ndo_get_xmit_slave bonding: Add array of all slaves bonding: Add function to get the xmit slave in active-backup mode bonding: Add helper function to get the xmit slave in rr mode bonding: Add helper function to get the xmit slave based on hash bonding/alb: Add helper functions to get the xmit slave bonding: Rename slave_arr to usable_slaves bonding: Export skip slave logic to function net/core: Introduce netdev_get_xmit_slave
-
- 01 May, 2020 7 commits
-
-
Maor Gottlieb authored
Add function to get the device physical port of the lag slave. Signed-off-by:
David S. Miller <davem@davemloft.net> Signed-off-by:
Saeed Mahameed <saeedm@mellanox.com>
-
Maor Gottlieb authored
The lag lock could be a spin lock, the critical section is short and there is no need that the thread will sleep. Change the lock that protects the LAG structure from mutex to spin lock. It is required for next patch that need to access this structure from context that we can't sleep. In addition there is no need to hold this lock when query the congestion counters. Signed-off-by:
-
Maor Gottlieb authored
Add implementation of ndo_get_xmit_slave. Find the slave by using the helper function according to the bond mode. If the caller set all_slaves to true, then it assumes that all slaves are available to transmit. Signed-off-by:
Jay Vosburgh <jay.vosburgh@canonical.com> Reviewed-by:
Jiri Pirko <jiri@mellanox.com> Acked-by:
-
Maor Gottlieb authored
Keep all slaves in array so it could be used to get the xmit slave assume all the slaves are active. The logic to add slave to the array is like the usable slaves, except that we also add slaves that currently can't transmit - not up or active. Signed-off-by:
-
Maor Gottlieb authored
Add helper function to get the xmit slave in active-backup mode. It's only one line function that return the curr_active_slave, but it will used both in the xmit flow and by the new .ndo to get the xmit slave. Signed-off-by:
-
Maor Gottlieb authored
Add helper function to get the xmit slave when bond is in round robin mode. Change bond_xmit_slave_id to bond_get_slave_by_id, then the logic for find the next slave for transmit could be used both by the xmit flow and the .ndo to get the xmit slave. Signed-off-by:
-
Maor Gottlieb authored
Both xor and 802.3ad modes use bond_xmit_hash to get the xmit slave. Export the logic to helper function so it could be used in the following patches by the .ndo to get the xmit slave. Signed-off-by:
-
