From 385156c5f2a61834666f079ee66338f177c65c28 Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 23 Jan 2019 19:29:44 -0800 Subject: IB/hfi: Move RC functions into a header file This patch moves some RC helper functions into a header file so that they can be called from both RC and TID RDMA functions. In addition, a common function for rewinding a request is created in rdmavt so that it can be shared between qib and hfi1 driver. Reviewed-by: Mike Marciniszyn Signed-off-by: Mitko Haralanov Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Doug Ledford --- include/rdma/rdmavt_qp.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'include') diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h index cbafb1878669..56a9221378d9 100644 --- a/include/rdma/rdmavt_qp.h +++ b/include/rdma/rdmavt_qp.h @@ -628,6 +628,16 @@ __be32 rvt_compute_aeth(struct rvt_qp *qp); */ void rvt_get_credit(struct rvt_qp *qp, u32 aeth); +/** + * rvt_restart_sge - rewind the sge state for a wqe + * @ss: the sge state pointer + * @wqe: the wqe to rewind + * @len: the data length from the start of the wqe in bytes + * + * Returns the remaining data length. + */ +u32 rvt_restart_sge(struct rvt_sge_state *ss, struct rvt_swqe *wqe, u32 len); + /** * @qp - the qp pair * @len - the length -- cgit v1.2.3 From 838b6fd2d9ca29998869e4d1ecf4566efe807666 Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 23 Jan 2019 19:30:07 -0800 Subject: IB/hfi1: TID RDMA RcvArray programming and TID allocation TID entries are used by hfi1 hardware to receive data payload from incoming packets directly into a user buffer and thus avoid data copying by software. This patch implements the functions for TID allocation, freeing, and programming TID RcvArray entries in hardware for kernel clients. TID entries are managed via lists of TID groups similar to PSM. Furthermore, to track TID resource allocation for each request, software flows are also allocated and freed as needed. Since software flows consume large amount of memory for tracking TID allocation and freeing, it is generally desirable to allocate them dynamically in the send queue and only for TID RDMA requests, but pre-allocate them for receive queue because the send queue could have thousands of entries while the receive queue has only a limited number of entries. Signed-off-by: Mitko Haralanov Signed-off-by: Ashutosh Dixit Signed-off-by: Mike Marciniszyn Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Doug Ledford --- include/rdma/rdmavt_qp.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include') diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h index 56a9221378d9..9095a0b71250 100644 --- a/include/rdma/rdmavt_qp.h +++ b/include/rdma/rdmavt_qp.h @@ -174,6 +174,7 @@ struct rvt_swqe { u32 lpsn; /* last packet sequence number */ u32 ssn; /* send sequence number */ u32 length; /* total length of data in sg_list */ + void *priv; /* driver dependent field */ struct rvt_sge sg_list[0]; }; @@ -235,6 +236,7 @@ struct rvt_ack_entry { u32 lpsn; u8 opcode; u8 sent; + void *priv; }; #define RC_QP_SCALING_INTERVAL 5 -- cgit v1.2.3 From 742a3826cf82395e304df99f6494d04b0dd03a84 Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 23 Jan 2019 19:30:40 -0800 Subject: IB/hfi1: Add functions to build TID RDMA READ request This patch adds the helper functions to build the TID RDMA READ request on the requester side. The key is to allocate TID resources (TID flow and TID entries) and send the resource information to the responder side along with the read request. Since the TID resources are limited, each TID RDMA READ request has to be split into segments with a default segment size of 256K. A software flow is allocated to track the data transaction for each segment. The work request opcode, packet opcode, and packet formats for TID RDMA READ protocol are also defined in this patch. Reviewed-by: Mike Marciniszyn Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Doug Ledford --- include/rdma/ib_hdrs.h | 9 +++++++- include/rdma/tid_rdma_defs.h | 52 ++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 60 insertions(+), 1 deletion(-) create mode 100644 include/rdma/tid_rdma_defs.h (limited to 'include') diff --git a/include/rdma/ib_hdrs.h b/include/rdma/ib_hdrs.h index 6e35416170a3..58a0a0f99e7f 100644 --- a/include/rdma/ib_hdrs.h +++ b/include/rdma/ib_hdrs.h @@ -1,5 +1,5 @@ /* - * Copyright(c) 2016 Intel Corporation. + * Copyright(c) 2016 - 2018 Intel Corporation. * * This file is provided under a dual BSD/GPLv2 license. When using or * redistributing this file, you may do so under either license. @@ -100,6 +100,8 @@ struct ib_atomic_eth { __be64 compare_data; /* potentially unaligned */ } __packed; +#include + union ib_ehdrs { struct { __be32 deth[2]; @@ -117,6 +119,11 @@ union ib_ehdrs { __be32 aeth; __be32 ieth; struct ib_atomic_eth atomic_eth; + /* TID RDMA headers */ + union { + struct tid_rdma_read_req r_req; + struct tid_rdma_read_resp r_rsp; + } tid_rdma; } __packed; struct ib_other_headers { diff --git a/include/rdma/tid_rdma_defs.h b/include/rdma/tid_rdma_defs.h new file mode 100644 index 000000000000..1c431ea32b52 --- /dev/null +++ b/include/rdma/tid_rdma_defs.h @@ -0,0 +1,52 @@ +/* SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause) */ +/* + * Copyright(c) 2018 Intel Corporation. + * + */ + +#ifndef TID_RDMA_DEFS_H +#define TID_RDMA_DEFS_H + +#include + +struct tid_rdma_read_req { + __le32 kdeth0; + __le32 kdeth1; + struct ib_reth reth; + __be32 tid_flow_psn; + __be32 tid_flow_qp; + __be32 verbs_qp; +}; + +struct tid_rdma_read_resp { + __le32 kdeth0; + __le32 kdeth1; + __be32 aeth; + __be32 reserved[4]; + __be32 verbs_psn; + __be32 verbs_qp; +}; + +/* + * TID RDMA Opcodes + */ +#define IB_OPCODE_TID_RDMA 0xe0 +enum { + IB_OPCODE_READ_REQ = 0x4, + IB_OPCODE_READ_RESP = 0x5, + + IB_OPCODE(TID_RDMA, READ_REQ), + IB_OPCODE(TID_RDMA, READ_RESP), +}; + +#define TID_OP(x) IB_OPCODE_TID_RDMA_##x + +/* + * Define TID RDMA specific WR opcodes. The ib_wr_opcode + * enum already provides some reserved values for use by + * low level drivers. Two of those are used but renamed + * to be more descriptive. + */ +#define IB_WR_TID_RDMA_READ IB_WR_RESERVED2 + +#endif /* TID_RDMA_DEFS_H */ -- cgit v1.2.3 From 039cd3daf19b9acbf080054d765cbceac842b6a0 Mon Sep 17 00:00:00 2001 From: Kaike Wan Date: Wed, 23 Jan 2019 19:31:57 -0800 Subject: IB/hfi1: Increment the retry timeout value for TID RDMA READ request The RC retry timeout value is based on the estimated time for the response packet to come back. However, for TID RDMA READ request, due to the use of header suppression, the driver is normally not notified for each incoming response packet until the last TID RDMA READ response packet. Consequently, the retry timeout value should be extended to cover the transaction time for the entire length of a segment (default 256K) instead of that for a single packet. This patch addresses the issue by introducing new retry timer functions to account for multiple packets and wrapper functions for backward compatibility. Reviewed-by: Mike Marciniszyn Signed-off-by: Kaike Wan Signed-off-by: Dennis Dalessandro Signed-off-by: Doug Ledford --- include/rdma/rdma_vt.h | 12 +++++++++--- include/rdma/rdmavt_qp.h | 6 +++++- 2 files changed, 14 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h index 168e40be183c..87d66c9630d7 100644 --- a/include/rdma/rdma_vt.h +++ b/include/rdma/rdma_vt.h @@ -574,9 +574,10 @@ static inline struct rvt_qp *rvt_lookup_qpn(struct rvt_dev_info *rdi, /** * rvt_mod_retry_timer - mod a retry timer * @qp - the QP + * @shift - timeout shift to wait for multiple packets * Modify a potentially already running retry timer */ -static inline void rvt_mod_retry_timer(struct rvt_qp *qp) +static inline void rvt_mod_retry_timer_ext(struct rvt_qp *qp, u8 shift) { struct ib_qp *ibqp = &qp->ibqp; struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device); @@ -584,8 +585,13 @@ static inline void rvt_mod_retry_timer(struct rvt_qp *qp) lockdep_assert_held(&qp->s_lock); qp->s_flags |= RVT_S_TIMER; /* 4.096 usec. * (1 << qp->timeout) */ - mod_timer(&qp->s_timer, jiffies + qp->timeout_jiffies + - rdi->busy_jiffies); + mod_timer(&qp->s_timer, jiffies + rdi->busy_jiffies + + (qp->timeout_jiffies << shift)); +} + +static inline void rvt_mod_retry_timer(struct rvt_qp *qp) +{ + return rvt_mod_retry_timer_ext(qp, 0); } struct rvt_dev_info *rvt_alloc_device(size_t size, int nports); diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h index 9095a0b71250..d8d88d023092 100644 --- a/include/rdma/rdmavt_qp.h +++ b/include/rdma/rdmavt_qp.h @@ -688,7 +688,11 @@ enum hrtimer_restart rvt_rc_rnr_retry(struct hrtimer *t); void rvt_add_rnr_timer(struct rvt_qp *qp, u32 aeth); void rvt_del_timers_sync(struct rvt_qp *qp); void rvt_stop_rc_timers(struct rvt_qp *qp); -void rvt_add_retry_timer(struct rvt_qp *qp); +void rvt_add_retry_timer_ext(struct rvt_qp *qp, u8 shift); +static inline void rvt_add_retry_timer(struct rvt_qp *qp) +{ + rvt_add_retry_timer_ext(qp, 0); +} void rvt_copy_sge(struct rvt_qp *qp, struct rvt_sge_state *ss, void *data, u32 length, -- cgit v1.2.3