| Current Path : /usr/src/contrib/ofed/libmlx4/src/ |
FreeBSD hs32.drive.ne.jp 9.1-RELEASE FreeBSD 9.1-RELEASE #1: Wed Jan 14 12:18:08 JST 2015 root@hs32.drive.ne.jp:/sys/amd64/compile/hs32 amd64 |
| Current File : //usr/src/contrib/ofed/libmlx4/src/cq.c |
/*
* Copyright (c) 2005 Topspin Communications. All rights reserved.
* Copyright (c) 2005 Mellanox Technologies Ltd. All rights reserved.
* Copyright (c) 2006, 2007 Cisco Systems. All rights reserved.
*
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above
* copyright notice, this list of conditions and the following
* disclaimer.
*
* - Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials
* provided with the distribution.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#if HAVE_CONFIG_H
# include <config.h>
#endif /* HAVE_CONFIG_H */
#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>
#include <netinet/in.h>
#include <string.h>
#include <infiniband/opcode.h>
#include "mlx4.h"
#include "doorbell.h"
enum {
MLX4_CQ_DOORBELL = 0x20
};
enum {
CQ_OK = 0,
CQ_EMPTY = -1,
CQ_POLL_ERR = -2
};
#define MLX4_CQ_DB_REQ_NOT_SOL (1 << 24)
#define MLX4_CQ_DB_REQ_NOT (2 << 24)
enum {
MLX4_CQE_OWNER_MASK = 0x80,
MLX4_CQE_IS_SEND_MASK = 0x40,
MLX4_CQE_OPCODE_MASK = 0x1f
};
enum {
MLX4_CQE_SYNDROME_LOCAL_LENGTH_ERR = 0x01,
MLX4_CQE_SYNDROME_LOCAL_QP_OP_ERR = 0x02,
MLX4_CQE_SYNDROME_LOCAL_PROT_ERR = 0x04,
MLX4_CQE_SYNDROME_WR_FLUSH_ERR = 0x05,
MLX4_CQE_SYNDROME_MW_BIND_ERR = 0x06,
MLX4_CQE_SYNDROME_BAD_RESP_ERR = 0x10,
MLX4_CQE_SYNDROME_LOCAL_ACCESS_ERR = 0x11,
MLX4_CQE_SYNDROME_REMOTE_INVAL_REQ_ERR = 0x12,
MLX4_CQE_SYNDROME_REMOTE_ACCESS_ERR = 0x13,
MLX4_CQE_SYNDROME_REMOTE_OP_ERR = 0x14,
MLX4_CQE_SYNDROME_TRANSPORT_RETRY_EXC_ERR = 0x15,
MLX4_CQE_SYNDROME_RNR_RETRY_EXC_ERR = 0x16,
MLX4_CQE_SYNDROME_REMOTE_ABORTED_ERR = 0x22,
};
struct mlx4_cqe {
uint32_t my_qpn;
uint32_t immed_rss_invalid;
uint32_t g_mlpath_rqpn;
uint8_t sl;
uint8_t reserved1;
uint16_t rlid;
uint32_t reserved2;
uint32_t byte_cnt;
uint16_t wqe_index;
uint16_t checksum;
uint8_t reserved3[3];
uint8_t owner_sr_opcode;
};
struct mlx4_err_cqe {
uint32_t my_qpn;
uint32_t reserved1[5];
uint16_t wqe_index;
uint8_t vendor_err;
uint8_t syndrome;
uint8_t reserved2[3];
uint8_t owner_sr_opcode;
};
static struct mlx4_cqe *get_cqe(struct mlx4_cq *cq, int entry)
{
return cq->buf.buf + entry * MLX4_CQ_ENTRY_SIZE;
}
static void *get_sw_cqe(struct mlx4_cq *cq, int n)
{
struct mlx4_cqe *cqe = get_cqe(cq, n & cq->ibv_cq.cqe);
return (!!(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK) ^
!!(n & (cq->ibv_cq.cqe + 1))) ? NULL : cqe;
}
static struct mlx4_cqe *next_cqe_sw(struct mlx4_cq *cq)
{
return get_sw_cqe(cq, cq->cons_index);
}
static void update_cons_index(struct mlx4_cq *cq)
{
*cq->set_ci_db = htonl(cq->cons_index & 0xffffff);
}
static void mlx4_handle_error_cqe(struct mlx4_err_cqe *cqe, struct ibv_wc *wc)
{
if (cqe->syndrome == MLX4_CQE_SYNDROME_LOCAL_QP_OP_ERR)
printf(PFX "local QP operation err "
"(QPN %06x, WQE index %x, vendor syndrome %02x, "
"opcode = %02x)\n",
htonl(cqe->my_qpn), htonl(cqe->wqe_index),
cqe->vendor_err,
cqe->owner_sr_opcode & ~MLX4_CQE_OWNER_MASK);
switch (cqe->syndrome) {
case MLX4_CQE_SYNDROME_LOCAL_LENGTH_ERR:
wc->status = IBV_WC_LOC_LEN_ERR;
break;
case MLX4_CQE_SYNDROME_LOCAL_QP_OP_ERR:
wc->status = IBV_WC_LOC_QP_OP_ERR;
break;
case MLX4_CQE_SYNDROME_LOCAL_PROT_ERR:
wc->status = IBV_WC_LOC_PROT_ERR;
break;
case MLX4_CQE_SYNDROME_WR_FLUSH_ERR:
wc->status = IBV_WC_WR_FLUSH_ERR;
break;
case MLX4_CQE_SYNDROME_MW_BIND_ERR:
wc->status = IBV_WC_MW_BIND_ERR;
break;
case MLX4_CQE_SYNDROME_BAD_RESP_ERR:
wc->status = IBV_WC_BAD_RESP_ERR;
break;
case MLX4_CQE_SYNDROME_LOCAL_ACCESS_ERR:
wc->status = IBV_WC_LOC_ACCESS_ERR;
break;
case MLX4_CQE_SYNDROME_REMOTE_INVAL_REQ_ERR:
wc->status = IBV_WC_REM_INV_REQ_ERR;
break;
case MLX4_CQE_SYNDROME_REMOTE_ACCESS_ERR:
wc->status = IBV_WC_REM_ACCESS_ERR;
break;
case MLX4_CQE_SYNDROME_REMOTE_OP_ERR:
wc->status = IBV_WC_REM_OP_ERR;
break;
case MLX4_CQE_SYNDROME_TRANSPORT_RETRY_EXC_ERR:
wc->status = IBV_WC_RETRY_EXC_ERR;
break;
case MLX4_CQE_SYNDROME_RNR_RETRY_EXC_ERR:
wc->status = IBV_WC_RNR_RETRY_EXC_ERR;
break;
case MLX4_CQE_SYNDROME_REMOTE_ABORTED_ERR:
wc->status = IBV_WC_REM_ABORT_ERR;
break;
default:
wc->status = IBV_WC_GENERAL_ERR;
break;
}
wc->vendor_err = cqe->vendor_err;
}
static int mlx4_poll_one(struct mlx4_cq *cq,
struct mlx4_qp **cur_qp,
struct ibv_wc *wc)
{
struct mlx4_wq *wq;
struct mlx4_cqe *cqe;
struct mlx4_srq *srq = NULL;
uint32_t qpn;
uint32_t srqn;
uint32_t g_mlpath_rqpn;
uint16_t wqe_index;
int is_error;
int is_send;
cqe = next_cqe_sw(cq);
if (!cqe)
return CQ_EMPTY;
++cq->cons_index;
VALGRIND_MAKE_MEM_DEFINED(cqe, sizeof *cqe);
/*
* Make sure we read CQ entry contents after we've checked the
* ownership bit.
*/
rmb();
qpn = ntohl(cqe->my_qpn);
is_send = cqe->owner_sr_opcode & MLX4_CQE_IS_SEND_MASK;
is_error = (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
MLX4_CQE_OPCODE_ERROR;
if (qpn & MLX4_XRC_QPN_BIT && !is_send) {
srqn = ntohl(cqe->g_mlpath_rqpn) & 0xffffff;
/*
* We do not have to take the XRC SRQ table lock here,
* because CQs will be locked while XRC SRQs are removed
* from the table.
*/
srq = mlx4_find_xrc_srq(to_mctx(cq->ibv_cq.context), srqn);
if (!srq)
return CQ_POLL_ERR;
} else if (!*cur_qp || (qpn & 0xffffff) != (*cur_qp)->ibv_qp.qp_num) {
/*
* We do not have to take the QP table lock here,
* because CQs will be locked while QPs are removed
* from the table.
*/
*cur_qp = mlx4_find_qp(to_mctx(cq->ibv_cq.context),
qpn & 0xffffff);
if (!*cur_qp)
return CQ_POLL_ERR;
}
wc->qp_num = qpn & 0xffffff;
if (is_send) {
wq = &(*cur_qp)->sq;
wqe_index = ntohs(cqe->wqe_index);
wq->tail += (uint16_t) (wqe_index - (uint16_t) wq->tail);
wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
++wq->tail;
} else if (srq) {
wqe_index = htons(cqe->wqe_index);
wc->wr_id = srq->wrid[wqe_index];
mlx4_free_srq_wqe(srq, wqe_index);
} else if ((*cur_qp)->ibv_qp.srq) {
srq = to_msrq((*cur_qp)->ibv_qp.srq);
wqe_index = htons(cqe->wqe_index);
wc->wr_id = srq->wrid[wqe_index];
mlx4_free_srq_wqe(srq, wqe_index);
} else {
wq = &(*cur_qp)->rq;
wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
++wq->tail;
}
if (is_error) {
mlx4_handle_error_cqe((struct mlx4_err_cqe *) cqe, wc);
return CQ_OK;
}
wc->status = IBV_WC_SUCCESS;
if (is_send) {
wc->wc_flags = 0;
switch (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) {
case MLX4_OPCODE_RDMA_WRITE_IMM:
wc->wc_flags |= IBV_WC_WITH_IMM;
case MLX4_OPCODE_RDMA_WRITE:
wc->opcode = IBV_WC_RDMA_WRITE;
break;
case MLX4_OPCODE_SEND_IMM:
wc->wc_flags |= IBV_WC_WITH_IMM;
case MLX4_OPCODE_SEND:
wc->opcode = IBV_WC_SEND;
break;
case MLX4_OPCODE_RDMA_READ:
wc->opcode = IBV_WC_RDMA_READ;
wc->byte_len = ntohl(cqe->byte_cnt);
break;
case MLX4_OPCODE_ATOMIC_CS:
wc->opcode = IBV_WC_COMP_SWAP;
wc->byte_len = 8;
break;
case MLX4_OPCODE_ATOMIC_FA:
wc->opcode = IBV_WC_FETCH_ADD;
wc->byte_len = 8;
break;
case MLX4_OPCODE_BIND_MW:
wc->opcode = IBV_WC_BIND_MW;
break;
default:
/* assume it's a send completion */
wc->opcode = IBV_WC_SEND;
break;
}
} else {
wc->byte_len = ntohl(cqe->byte_cnt);
switch (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) {
case MLX4_RECV_OPCODE_RDMA_WRITE_IMM:
wc->opcode = IBV_WC_RECV_RDMA_WITH_IMM;
wc->wc_flags = IBV_WC_WITH_IMM;
wc->imm_data = cqe->immed_rss_invalid;
break;
case MLX4_RECV_OPCODE_SEND:
wc->opcode = IBV_WC_RECV;
wc->wc_flags = 0;
break;
case MLX4_RECV_OPCODE_SEND_IMM:
wc->opcode = IBV_WC_RECV;
wc->wc_flags = IBV_WC_WITH_IMM;
wc->imm_data = cqe->immed_rss_invalid;
break;
}
wc->slid = ntohs(cqe->rlid);
wc->sl = cqe->sl >> 4;
g_mlpath_rqpn = ntohl(cqe->g_mlpath_rqpn);
wc->src_qp = g_mlpath_rqpn & 0xffffff;
wc->dlid_path_bits = (g_mlpath_rqpn >> 24) & 0x7f;
wc->wc_flags |= g_mlpath_rqpn & 0x80000000 ? IBV_WC_GRH : 0;
wc->pkey_index = ntohl(cqe->immed_rss_invalid) & 0x7f;
}
return CQ_OK;
}
int mlx4_poll_cq(struct ibv_cq *ibcq, int ne, struct ibv_wc *wc)
{
struct mlx4_cq *cq = to_mcq(ibcq);
struct mlx4_qp *qp = NULL;
int npolled;
int err = CQ_OK;
pthread_spin_lock(&cq->lock);
for (npolled = 0; npolled < ne; ++npolled) {
err = mlx4_poll_one(cq, &qp, wc + npolled);
if (err != CQ_OK)
break;
}
if (npolled)
update_cons_index(cq);
pthread_spin_unlock(&cq->lock);
return err == CQ_POLL_ERR ? err : npolled;
}
int mlx4_arm_cq(struct ibv_cq *ibvcq, int solicited)
{
struct mlx4_cq *cq = to_mcq(ibvcq);
uint32_t doorbell[2];
uint32_t sn;
uint32_t ci;
uint32_t cmd;
sn = cq->arm_sn & 3;
ci = cq->cons_index & 0xffffff;
cmd = solicited ? MLX4_CQ_DB_REQ_NOT_SOL : MLX4_CQ_DB_REQ_NOT;
*cq->arm_db = htonl(sn << 28 | cmd | ci);
/*
* Make sure that the doorbell record in host memory is
* written before ringing the doorbell via PCI MMIO.
*/
wmb();
doorbell[0] = htonl(sn << 28 | cmd | cq->cqn);
doorbell[1] = htonl(ci);
mlx4_write64(doorbell, to_mctx(ibvcq->context), MLX4_CQ_DOORBELL);
return 0;
}
void mlx4_cq_event(struct ibv_cq *cq)
{
to_mcq(cq)->arm_sn++;
}
void __mlx4_cq_clean(struct mlx4_cq *cq, uint32_t qpn, struct mlx4_srq *srq)
{
struct mlx4_cqe *cqe, *dest;
uint32_t prod_index;
uint8_t owner_bit;
int nfreed = 0;
int is_xrc_srq = 0;
if (srq && srq->ibv_srq.xrc_cq)
is_xrc_srq = 1;
/*
* First we need to find the current producer index, so we
* know where to start cleaning from. It doesn't matter if HW
* adds new entries after this loop -- the QP we're worried
* about is already in RESET, so the new entries won't come
* from our QP and therefore don't need to be checked.
*/
for (prod_index = cq->cons_index; get_sw_cqe(cq, prod_index); ++prod_index)
if (prod_index == cq->cons_index + cq->ibv_cq.cqe)
break;
/*
* Now sweep backwards through the CQ, removing CQ entries
* that match our QP by copying older entries on top of them.
*/
while ((int) --prod_index - (int) cq->cons_index >= 0) {
cqe = get_cqe(cq, prod_index & cq->ibv_cq.cqe);
if (is_xrc_srq &&
(ntohl(cqe->g_mlpath_rqpn & 0xffffff) == srq->srqn) &&
!(cqe->owner_sr_opcode & MLX4_CQE_IS_SEND_MASK)) {
mlx4_free_srq_wqe(srq, ntohs(cqe->wqe_index));
++nfreed;
} else if ((ntohl(cqe->my_qpn) & 0xffffff) == qpn) {
if (srq && !(cqe->owner_sr_opcode & MLX4_CQE_IS_SEND_MASK))
mlx4_free_srq_wqe(srq, ntohs(cqe->wqe_index));
++nfreed;
} else if (nfreed) {
dest = get_cqe(cq, (prod_index + nfreed) & cq->ibv_cq.cqe);
owner_bit = dest->owner_sr_opcode & MLX4_CQE_OWNER_MASK;
memcpy(dest, cqe, sizeof *cqe);
dest->owner_sr_opcode = owner_bit |
(dest->owner_sr_opcode & ~MLX4_CQE_OWNER_MASK);
}
}
if (nfreed) {
cq->cons_index += nfreed;
/*
* Make sure update of buffer contents is done before
* updating consumer index.
*/
wmb();
update_cons_index(cq);
}
}
void mlx4_cq_clean(struct mlx4_cq *cq, uint32_t qpn, struct mlx4_srq *srq)
{
pthread_spin_lock(&cq->lock);
__mlx4_cq_clean(cq, qpn, srq);
pthread_spin_unlock(&cq->lock);
}
int mlx4_get_outstanding_cqes(struct mlx4_cq *cq)
{
uint32_t i;
for (i = cq->cons_index; get_sw_cqe(cq, (i & cq->ibv_cq.cqe)); ++i)
;
return i - cq->cons_index;
}
void mlx4_cq_resize_copy_cqes(struct mlx4_cq *cq, void *buf, int old_cqe)
{
struct mlx4_cqe *cqe;
int i;
i = cq->cons_index;
cqe = get_cqe(cq, (i & old_cqe));
while ((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) != MLX4_CQE_OPCODE_RESIZE) {
cqe->owner_sr_opcode = (cqe->owner_sr_opcode & ~MLX4_CQE_OWNER_MASK) |
(((i + 1) & (cq->ibv_cq.cqe + 1)) ? MLX4_CQE_OWNER_MASK : 0);
memcpy(buf + ((i + 1) & cq->ibv_cq.cqe) * MLX4_CQ_ENTRY_SIZE,
cqe, MLX4_CQ_ENTRY_SIZE);
++i;
cqe = get_cqe(cq, (i & old_cqe));
}
++cq->cons_index;
}
int mlx4_alloc_cq_buf(struct mlx4_device *dev, struct mlx4_buf *buf, int nent)
{
if (mlx4_alloc_buf(buf, align(nent * MLX4_CQ_ENTRY_SIZE, dev->page_size),
dev->page_size))
return -1;
memset(buf->buf, 0, nent * MLX4_CQ_ENTRY_SIZE);
return 0;
}