The HIP09 introduces the DCA(Dynamic Context Attachment) feature which supports many RC QPs to share the WQE buffer in a memory pool. If a QP enables DCA feature, the WQE's buffer will not be allocated when creating but when the users start to post WRs. This will reduce the memory consumption when there are too many QPs are inactive. The related userspace series is named "libhns: Add support for Dynamic Context Attachment". Changes since v3: * alloc_dca_states() replaces GFP_NOWAIT with the GFP_KERNEL in #1. * Remove the calculation of DCA percentage in #12. * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1627356452-30564-1-git-send-email-liangwenpeng@xxxxxxxxxx/ Changes since v2: * Refactor flow of modify_qp and then add a shared memory mechanism to synchronize status of DCA, which can save more memory than previous scheme. * Remove some inline functions from #1. * Add support for dumping DCA mem status in restrack. * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1620732161-27180-1-git-send-email-liweihang@xxxxxxxxxx/ Changes since v1: * Modify return type of hns_roce_enable_dca() to void. * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1620650889-61650-1-git-send-email-liweihang@xxxxxxxxxx/ Two RFC versions of this series has been sent before, and it's associated with the userspace one "libhns: Add support for Dynamic Context Attachment". Changes since RFC v2: * Just fix a typo in commit message of #6. * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1611394994-50363-1-git-send-email-liweihang@xxxxxxxxxx/ Changes since RFC v1: * Replace all GFP_ATOMIC with GFP_NOWAIT, because the former may use emergency pool if no regular memory can be found. * Change size of cap_flags of alloc_ucontext_resp from 32 to 64 to avoid a potential problem when pass it back to the userspace. * Move definition of HNS_ROCE_CAP_FLAG_DCA_MODE to hns-abi.h. * Rename free_mem_states() to free_dca_states() in #1. * Link: https://patchwork.kernel.org/project/linux-rdma/cover/1610706138-4219-1-git-send-email-liweihang@xxxxxxxxxx/ Xi Wang (12): RDMA/hns: Introduce DCA for RC QP RDMA/hns: Add method for shrinking DCA memory pool RDMA/hns: Configure DCA mode for the userspace QP RDMA/hns: Refactor QP modify flow RDMA/hns: Add method for attaching WQE buffer RDMA/hns: Setup the configuration of WQE addressing to QPC RDMA/hns: Add method to detach WQE buffer RDMA/hns: Add method to query WQE buffer's address RDMA/hns: Add a shared memory to sync DCA status RDMA/hns: Sync DCA status by the shared memory RDMA/nldev: Add detailed CTX information support RDMA/hns: Dump detailed driver-specific UCTX drivers/infiniband/core/device.c | 1 + drivers/infiniband/core/nldev.c | 8 +- drivers/infiniband/hw/hns/Makefile | 2 +- drivers/infiniband/hw/hns/hns_roce_dca.c | 1438 +++++++++++++++++++++++++ drivers/infiniband/hw/hns/hns_roce_dca.h | 73 ++ drivers/infiniband/hw/hns/hns_roce_device.h | 47 +- drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 12 +- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 173 ++- drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 1 + drivers/infiniband/hw/hns/hns_roce_main.c | 128 ++- drivers/infiniband/hw/hns/hns_roce_qp.c | 169 ++- drivers/infiniband/hw/hns/hns_roce_restrack.c | 50 + include/rdma/ib_verbs.h | 1 + include/uapi/rdma/hns-abi.h | 86 ++ 14 files changed, 2113 insertions(+), 76 deletions(-) create mode 100644 drivers/infiniband/hw/hns/hns_roce_dca.c create mode 100644 drivers/infiniband/hw/hns/hns_roce_dca.h -- 2.8.1
