iw_cxgbe: iWARP driver for Chelsio T4/T5 chips. This is a straight port

of the iw_cxgb4 found in OFED distributions.

Obtained from:	Chelsio
This commit is contained in:
Navdeep Parhar 2013-10-17 18:37:25 +00:00
parent 1b182736a6
commit fb93f5c47f
Notes: svn2git 2020-12-20 02:59:44 +00:00
svn path=/head/; revision=256694
14 changed files with 9199 additions and 1 deletions

2458
sys/dev/cxgbe/iw_cxgbe/cm.c Normal file

File diff suppressed because it is too large Load Diff

926
sys/dev/cxgbe/iw_cxgbe/cq.c Normal file
View File

@ -0,0 +1,926 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/bus.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/rwlock.h>
#include <sys/socket.h>
#include <sys/sbuf.h>
#include "iw_cxgbe.h"
#include "user.h"
static int destroy_cq(struct c4iw_rdev *rdev, struct t4_cq *cq,
struct c4iw_dev_ucontext *uctx)
{
struct adapter *sc = rdev->adap;
struct fw_ri_res_wr *res_wr;
struct fw_ri_res *res;
int wr_len;
struct c4iw_wr_wait wr_wait;
struct wrqe *wr;
wr_len = sizeof *res_wr + sizeof *res;
wr = alloc_wrqe(wr_len, &sc->sge.mgmtq);
if (wr == NULL)
return (0);
res_wr = wrtod(wr);
memset(res_wr, 0, wr_len);
res_wr->op_nres = cpu_to_be32(
V_FW_WR_OP(FW_RI_RES_WR) |
V_FW_RI_RES_WR_NRES(1) |
F_FW_WR_COMPL);
res_wr->len16_pkd = cpu_to_be32(DIV_ROUND_UP(wr_len, 16));
res_wr->cookie = (unsigned long) &wr_wait;
res = res_wr->res;
res->u.cq.restype = FW_RI_RES_TYPE_CQ;
res->u.cq.op = FW_RI_RES_OP_RESET;
res->u.cq.iqid = cpu_to_be32(cq->cqid);
c4iw_init_wr_wait(&wr_wait);
t4_wrq_tx(sc, wr);
c4iw_wait_for_reply(rdev, &wr_wait, 0, 0, __func__);
kfree(cq->sw_queue);
contigfree(cq->queue, cq->memsize, M_DEVBUF);
c4iw_put_cqid(rdev, cq->cqid, uctx);
return 0;
}
static int
create_cq(struct c4iw_rdev *rdev, struct t4_cq *cq,
struct c4iw_dev_ucontext *uctx)
{
struct adapter *sc = rdev->adap;
struct fw_ri_res_wr *res_wr;
struct fw_ri_res *res;
int wr_len;
int user = (uctx != &rdev->uctx);
struct c4iw_wr_wait wr_wait;
int ret;
struct wrqe *wr;
cq->cqid = c4iw_get_cqid(rdev, uctx);
if (!cq->cqid) {
ret = -ENOMEM;
goto err1;
}
if (!user) {
cq->sw_queue = kzalloc(cq->memsize, GFP_KERNEL);
if (!cq->sw_queue) {
ret = -ENOMEM;
goto err2;
}
}
cq->queue = contigmalloc(cq->memsize, M_DEVBUF, M_NOWAIT, 0ul, ~0ul,
PAGE_SIZE, 0);
if (cq->queue)
cq->dma_addr = vtophys(cq->queue);
else {
ret = -ENOMEM;
goto err3;
}
pci_unmap_addr_set(cq, mapping, cq->dma_addr);
memset(cq->queue, 0, cq->memsize);
/* build fw_ri_res_wr */
wr_len = sizeof *res_wr + sizeof *res;
wr = alloc_wrqe(wr_len, &sc->sge.mgmtq);
if (wr == NULL)
return (0);
res_wr = wrtod(wr);
memset(res_wr, 0, wr_len);
res_wr->op_nres = cpu_to_be32(
V_FW_WR_OP(FW_RI_RES_WR) |
V_FW_RI_RES_WR_NRES(1) |
F_FW_WR_COMPL);
res_wr->len16_pkd = cpu_to_be32(DIV_ROUND_UP(wr_len, 16));
res_wr->cookie = (unsigned long) &wr_wait;
res = res_wr->res;
res->u.cq.restype = FW_RI_RES_TYPE_CQ;
res->u.cq.op = FW_RI_RES_OP_WRITE;
res->u.cq.iqid = cpu_to_be32(cq->cqid);
//Fixme: Always use first queue id for IQANDSTINDEX. Linux does the same.
res->u.cq.iqandst_to_iqandstindex = cpu_to_be32(
V_FW_RI_RES_WR_IQANUS(0) |
V_FW_RI_RES_WR_IQANUD(1) |
F_FW_RI_RES_WR_IQANDST |
V_FW_RI_RES_WR_IQANDSTINDEX(sc->sge.ofld_rxq[0].iq.abs_id));
res->u.cq.iqdroprss_to_iqesize = cpu_to_be16(
F_FW_RI_RES_WR_IQDROPRSS |
V_FW_RI_RES_WR_IQPCIECH(2) |
V_FW_RI_RES_WR_IQINTCNTTHRESH(0) |
F_FW_RI_RES_WR_IQO |
V_FW_RI_RES_WR_IQESIZE(1));
res->u.cq.iqsize = cpu_to_be16(cq->size);
res->u.cq.iqaddr = cpu_to_be64(cq->dma_addr);
c4iw_init_wr_wait(&wr_wait);
t4_wrq_tx(sc, wr);
CTR2(KTR_IW_CXGBE, "%s wait_event wr_wait %p", __func__, &wr_wait);
ret = c4iw_wait_for_reply(rdev, &wr_wait, 0, 0, __func__);
if (ret)
goto err4;
cq->gen = 1;
cq->gts = (void *)((unsigned long)rman_get_virtual(sc->regs_res) +
MYPF_REG(SGE_PF_GTS));
cq->rdev = rdev;
if (user) {
cq->ugts = (u64)((char*)rman_get_virtual(sc->udbs_res) +
(cq->cqid << rdev->cqshift));
cq->ugts &= PAGE_MASK;
CTR5(KTR_IW_CXGBE,
"%s: UGTS %p cqid %x cqshift %d page_mask %x", __func__,
cq->ugts, cq->cqid, rdev->cqshift, PAGE_MASK);
}
return 0;
err4:
contigfree(cq->queue, cq->memsize, M_DEVBUF);
err3:
kfree(cq->sw_queue);
err2:
c4iw_put_cqid(rdev, cq->cqid, uctx);
err1:
return ret;
}
static void insert_recv_cqe(struct t4_wq *wq, struct t4_cq *cq)
{
struct t4_cqe cqe;
CTR5(KTR_IW_CXGBE, "%s wq %p cq %p sw_cidx %u sw_pidx %u", __func__, wq,
cq, cq->sw_cidx, cq->sw_pidx);
memset(&cqe, 0, sizeof(cqe));
cqe.header = cpu_to_be32(V_CQE_STATUS(T4_ERR_SWFLUSH) |
V_CQE_OPCODE(FW_RI_SEND) |
V_CQE_TYPE(0) |
V_CQE_SWCQE(1) |
V_CQE_QPID(wq->sq.qid));
cqe.bits_type_ts = cpu_to_be64(V_CQE_GENBIT((u64)cq->gen));
cq->sw_queue[cq->sw_pidx] = cqe;
t4_swcq_produce(cq);
}
int c4iw_flush_rq(struct t4_wq *wq, struct t4_cq *cq, int count)
{
int flushed = 0;
int in_use = wq->rq.in_use - count;
BUG_ON(in_use < 0);
CTR5(KTR_IW_CXGBE, "%s wq %p cq %p rq.in_use %u skip count %u",
__func__, wq, cq, wq->rq.in_use, count);
while (in_use--) {
insert_recv_cqe(wq, cq);
flushed++;
}
return flushed;
}
static void insert_sq_cqe(struct t4_wq *wq, struct t4_cq *cq,
struct t4_swsqe *swcqe)
{
struct t4_cqe cqe;
CTR5(KTR_IW_CXGBE, "%s wq %p cq %p sw_cidx %u sw_pidx %u", __func__, wq,
cq, cq->sw_cidx, cq->sw_pidx);
memset(&cqe, 0, sizeof(cqe));
cqe.header = cpu_to_be32(V_CQE_STATUS(T4_ERR_SWFLUSH) |
V_CQE_OPCODE(swcqe->opcode) |
V_CQE_TYPE(1) |
V_CQE_SWCQE(1) |
V_CQE_QPID(wq->sq.qid));
CQE_WRID_SQ_IDX(&cqe) = swcqe->idx;
cqe.bits_type_ts = cpu_to_be64(V_CQE_GENBIT((u64)cq->gen));
cq->sw_queue[cq->sw_pidx] = cqe;
t4_swcq_produce(cq);
}
int c4iw_flush_sq(struct t4_wq *wq, struct t4_cq *cq, int count)
{
int flushed = 0;
struct t4_swsqe *swsqe = &wq->sq.sw_sq[wq->sq.cidx + count];
int in_use = wq->sq.in_use - count;
BUG_ON(in_use < 0);
while (in_use--) {
swsqe->signaled = 0;
insert_sq_cqe(wq, cq, swsqe);
swsqe++;
if (swsqe == (wq->sq.sw_sq + wq->sq.size))
swsqe = wq->sq.sw_sq;
flushed++;
}
return flushed;
}
/*
* Move all CQEs from the HWCQ into the SWCQ.
*/
void c4iw_flush_hw_cq(struct t4_cq *cq)
{
struct t4_cqe *cqe = NULL, *swcqe;
int ret;
CTR3(KTR_IW_CXGBE, "%s cq %p cqid 0x%x", __func__, cq, cq->cqid);
ret = t4_next_hw_cqe(cq, &cqe);
while (!ret) {
CTR3(KTR_IW_CXGBE, "%s flushing hwcq cidx 0x%x swcq pidx 0x%x",
__func__, cq->cidx, cq->sw_pidx);
swcqe = &cq->sw_queue[cq->sw_pidx];
*swcqe = *cqe;
swcqe->header |= cpu_to_be32(V_CQE_SWCQE(1));
t4_swcq_produce(cq);
t4_hwcq_consume(cq);
ret = t4_next_hw_cqe(cq, &cqe);
}
}
static int cqe_completes_wr(struct t4_cqe *cqe, struct t4_wq *wq)
{
if (CQE_OPCODE(cqe) == FW_RI_TERMINATE)
return 0;
if ((CQE_OPCODE(cqe) == FW_RI_RDMA_WRITE) && RQ_TYPE(cqe))
return 0;
if ((CQE_OPCODE(cqe) == FW_RI_READ_RESP) && SQ_TYPE(cqe))
return 0;
if (CQE_SEND_OPCODE(cqe) && RQ_TYPE(cqe) && t4_rq_empty(wq))
return 0;
return 1;
}
void c4iw_count_scqes(struct t4_cq *cq, struct t4_wq *wq, int *count)
{
struct t4_cqe *cqe;
u32 ptr;
*count = 0;
ptr = cq->sw_cidx;
while (ptr != cq->sw_pidx) {
cqe = &cq->sw_queue[ptr];
if ((SQ_TYPE(cqe) || ((CQE_OPCODE(cqe) == FW_RI_READ_RESP) &&
wq->sq.oldest_read)) &&
(CQE_QPID(cqe) == wq->sq.qid))
(*count)++;
if (++ptr == cq->size)
ptr = 0;
}
CTR3(KTR_IW_CXGBE, "%s cq %p count %d", __func__, cq, *count);
}
void c4iw_count_rcqes(struct t4_cq *cq, struct t4_wq *wq, int *count)
{
struct t4_cqe *cqe;
u32 ptr;
*count = 0;
CTR2(KTR_IW_CXGBE, "%s count zero %d", __func__, *count);
ptr = cq->sw_cidx;
while (ptr != cq->sw_pidx) {
cqe = &cq->sw_queue[ptr];
if (RQ_TYPE(cqe) && (CQE_OPCODE(cqe) != FW_RI_READ_RESP) &&
(CQE_QPID(cqe) == wq->sq.qid) && cqe_completes_wr(cqe, wq))
(*count)++;
if (++ptr == cq->size)
ptr = 0;
}
CTR3(KTR_IW_CXGBE, "%s cq %p count %d", __func__, cq, *count);
}
static void flush_completed_wrs(struct t4_wq *wq, struct t4_cq *cq)
{
struct t4_swsqe *swsqe;
u16 ptr = wq->sq.cidx;
int count = wq->sq.in_use;
int unsignaled = 0;
swsqe = &wq->sq.sw_sq[ptr];
while (count--)
if (!swsqe->signaled) {
if (++ptr == wq->sq.size)
ptr = 0;
swsqe = &wq->sq.sw_sq[ptr];
unsignaled++;
} else if (swsqe->complete) {
/*
* Insert this completed cqe into the swcq.
*/
CTR3(KTR_IW_CXGBE,
"%s moving cqe into swcq sq idx %u cq idx %u",
__func__, ptr, cq->sw_pidx);
swsqe->cqe.header |= htonl(V_CQE_SWCQE(1));
cq->sw_queue[cq->sw_pidx] = swsqe->cqe;
t4_swcq_produce(cq);
swsqe->signaled = 0;
wq->sq.in_use -= unsignaled;
break;
} else
break;
}
static void create_read_req_cqe(struct t4_wq *wq, struct t4_cqe *hw_cqe,
struct t4_cqe *read_cqe)
{
read_cqe->u.scqe.cidx = wq->sq.oldest_read->idx;
read_cqe->len = cpu_to_be32(wq->sq.oldest_read->read_len);
read_cqe->header = htonl(V_CQE_QPID(CQE_QPID(hw_cqe)) |
V_CQE_SWCQE(SW_CQE(hw_cqe)) |
V_CQE_OPCODE(FW_RI_READ_REQ) |
V_CQE_TYPE(1));
read_cqe->bits_type_ts = hw_cqe->bits_type_ts;
}
/*
* Return a ptr to the next read wr in the SWSQ or NULL.
*/
static void advance_oldest_read(struct t4_wq *wq)
{
u32 rptr = wq->sq.oldest_read - wq->sq.sw_sq + 1;
if (rptr == wq->sq.size)
rptr = 0;
while (rptr != wq->sq.pidx) {
wq->sq.oldest_read = &wq->sq.sw_sq[rptr];
if (wq->sq.oldest_read->opcode == FW_RI_READ_REQ)
return;
if (++rptr == wq->sq.size)
rptr = 0;
}
wq->sq.oldest_read = NULL;
}
/*
* poll_cq
*
* Caller must:
* check the validity of the first CQE,
* supply the wq assicated with the qpid.
*
* credit: cq credit to return to sge.
* cqe_flushed: 1 iff the CQE is flushed.
* cqe: copy of the polled CQE.
*
* return value:
* 0 CQE returned ok.
* -EAGAIN CQE skipped, try again.
* -EOVERFLOW CQ overflow detected.
*/
static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
u8 *cqe_flushed, u64 *cookie, u32 *credit)
{
int ret = 0;
struct t4_cqe *hw_cqe, read_cqe;
*cqe_flushed = 0;
*credit = 0;
ret = t4_next_cqe(cq, &hw_cqe);
if (ret)
return ret;
CTR6(KTR_IW_CXGBE,
"%s CQE OVF %u qpid 0x%0x genbit %u type %u status 0x%0x", __func__,
CQE_OVFBIT(hw_cqe), CQE_QPID(hw_cqe), CQE_GENBIT(hw_cqe),
CQE_TYPE(hw_cqe), CQE_STATUS(hw_cqe));
CTR5(KTR_IW_CXGBE,
"%s opcode 0x%0x len 0x%0x wrid_hi_stag 0x%x wrid_low_msn 0x%x",
__func__, CQE_OPCODE(hw_cqe), CQE_LEN(hw_cqe), CQE_WRID_HI(hw_cqe),
CQE_WRID_LOW(hw_cqe));
/*
* skip cqe's not affiliated with a QP.
*/
if (wq == NULL) {
ret = -EAGAIN;
goto skip_cqe;
}
/*
* Gotta tweak READ completions:
* 1) the cqe doesn't contain the sq_wptr from the wr.
* 2) opcode not reflected from the wr.
* 3) read_len not reflected from the wr.
* 4) cq_type is RQ_TYPE not SQ_TYPE.
*/
if (RQ_TYPE(hw_cqe) && (CQE_OPCODE(hw_cqe) == FW_RI_READ_RESP)) {
/*
* If this is an unsolicited read response, then the read
* was generated by the kernel driver as part of peer-2-peer
* connection setup. So ignore the completion.
*/
if (!wq->sq.oldest_read) {
if (CQE_STATUS(hw_cqe))
t4_set_wq_in_error(wq);
ret = -EAGAIN;
goto skip_cqe;
}
/*
* Don't write to the HWCQ, so create a new read req CQE
* in local memory.
*/
create_read_req_cqe(wq, hw_cqe, &read_cqe);
hw_cqe = &read_cqe;
advance_oldest_read(wq);
}
if (CQE_STATUS(hw_cqe) || t4_wq_in_error(wq)) {
*cqe_flushed = t4_wq_in_error(wq);
t4_set_wq_in_error(wq);
goto proc_cqe;
}
if (CQE_OPCODE(hw_cqe) == FW_RI_TERMINATE) {
ret = -EAGAIN;
goto skip_cqe;
}
/*
* RECV completion.
*/
if (RQ_TYPE(hw_cqe)) {
/*
* HW only validates 4 bits of MSN. So we must validate that
* the MSN in the SEND is the next expected MSN. If its not,
* then we complete this with T4_ERR_MSN and mark the wq in
* error.
*/
if (t4_rq_empty(wq)) {
t4_set_wq_in_error(wq);
ret = -EAGAIN;
goto skip_cqe;
}
if (unlikely((CQE_WRID_MSN(hw_cqe) != (wq->rq.msn)))) {
t4_set_wq_in_error(wq);
hw_cqe->header |= htonl(V_CQE_STATUS(T4_ERR_MSN));
goto proc_cqe;
}
goto proc_cqe;
}
/*
* If we get here its a send completion.
*
* Handle out of order completion. These get stuffed
* in the SW SQ. Then the SW SQ is walked to move any
* now in-order completions into the SW CQ. This handles
* 2 cases:
* 1) reaping unsignaled WRs when the first subsequent
* signaled WR is completed.
* 2) out of order read completions.
*/
if (!SW_CQE(hw_cqe) && (CQE_WRID_SQ_IDX(hw_cqe) != wq->sq.cidx)) {
struct t4_swsqe *swsqe;
CTR2(KTR_IW_CXGBE,
"%s out of order completion going in sw_sq at idx %u",
__func__, CQE_WRID_SQ_IDX(hw_cqe));
swsqe = &wq->sq.sw_sq[CQE_WRID_SQ_IDX(hw_cqe)];
swsqe->cqe = *hw_cqe;
swsqe->complete = 1;
ret = -EAGAIN;
goto flush_wq;
}
proc_cqe:
*cqe = *hw_cqe;
/*
* Reap the associated WR(s) that are freed up with this
* completion.
*/
if (SQ_TYPE(hw_cqe)) {
wq->sq.cidx = CQE_WRID_SQ_IDX(hw_cqe);
CTR2(KTR_IW_CXGBE, "%s completing sq idx %u",
__func__, wq->sq.cidx);
*cookie = wq->sq.sw_sq[wq->sq.cidx].wr_id;
t4_sq_consume(wq);
} else {
CTR2(KTR_IW_CXGBE, "%s completing rq idx %u",
__func__, wq->rq.cidx);
*cookie = wq->rq.sw_rq[wq->rq.cidx].wr_id;
BUG_ON(t4_rq_empty(wq));
t4_rq_consume(wq);
}
flush_wq:
/*
* Flush any completed cqes that are now in-order.
*/
flush_completed_wrs(wq, cq);
skip_cqe:
if (SW_CQE(hw_cqe)) {
CTR4(KTR_IW_CXGBE, "%s cq %p cqid 0x%x skip sw cqe cidx %u",
__func__, cq, cq->cqid, cq->sw_cidx);
t4_swcq_consume(cq);
} else {
CTR4(KTR_IW_CXGBE, "%s cq %p cqid 0x%x skip hw cqe cidx %u",
__func__, cq, cq->cqid, cq->cidx);
t4_hwcq_consume(cq);
}
return ret;
}
/*
* Get one cq entry from c4iw and map it to openib.
*
* Returns:
* 0 cqe returned
* -ENODATA EMPTY;
* -EAGAIN caller must try again
* any other -errno fatal error
*/
static int c4iw_poll_cq_one(struct c4iw_cq *chp, struct ib_wc *wc)
{
struct c4iw_qp *qhp = NULL;
struct t4_cqe cqe = {0, 0}, *rd_cqe;
struct t4_wq *wq;
u32 credit = 0;
u8 cqe_flushed;
u64 cookie = 0;
int ret;
ret = t4_next_cqe(&chp->cq, &rd_cqe);
if (ret)
return ret;
qhp = get_qhp(chp->rhp, CQE_QPID(rd_cqe));
if (!qhp)
wq = NULL;
else {
spin_lock(&qhp->lock);
wq = &(qhp->wq);
}
ret = poll_cq(wq, &(chp->cq), &cqe, &cqe_flushed, &cookie, &credit);
if (ret)
goto out;
wc->wr_id = cookie;
wc->qp = &qhp->ibqp;
wc->vendor_err = CQE_STATUS(&cqe);
wc->wc_flags = 0;
CTR5(KTR_IW_CXGBE, "%s qpid 0x%x type %d opcode %d status 0x%x",
__func__, CQE_QPID(&cqe), CQE_TYPE(&cqe), CQE_OPCODE(&cqe),
CQE_STATUS(&cqe));
CTR5(KTR_IW_CXGBE, "%s len %u wrid hi 0x%x lo 0x%x cookie 0x%llx",
__func__, CQE_LEN(&cqe), CQE_WRID_HI(&cqe), CQE_WRID_LOW(&cqe),
(unsigned long long)cookie);
if (CQE_TYPE(&cqe) == 0) {
if (!CQE_STATUS(&cqe))
wc->byte_len = CQE_LEN(&cqe);
else
wc->byte_len = 0;
wc->opcode = IB_WC_RECV;
if (CQE_OPCODE(&cqe) == FW_RI_SEND_WITH_INV ||
CQE_OPCODE(&cqe) == FW_RI_SEND_WITH_SE_INV) {
wc->ex.invalidate_rkey = CQE_WRID_STAG(&cqe);
wc->wc_flags |= IB_WC_WITH_INVALIDATE;
}
} else {
switch (CQE_OPCODE(&cqe)) {
case FW_RI_RDMA_WRITE:
wc->opcode = IB_WC_RDMA_WRITE;
break;
case FW_RI_READ_REQ:
wc->opcode = IB_WC_RDMA_READ;
wc->byte_len = CQE_LEN(&cqe);
break;
case FW_RI_SEND_WITH_INV:
case FW_RI_SEND_WITH_SE_INV:
wc->opcode = IB_WC_SEND;
wc->wc_flags |= IB_WC_WITH_INVALIDATE;
break;
case FW_RI_SEND:
case FW_RI_SEND_WITH_SE:
wc->opcode = IB_WC_SEND;
break;
case FW_RI_BIND_MW:
wc->opcode = IB_WC_BIND_MW;
break;
case FW_RI_LOCAL_INV:
wc->opcode = IB_WC_LOCAL_INV;
break;
case FW_RI_FAST_REGISTER:
wc->opcode = IB_WC_FAST_REG_MR;
break;
default:
printf("Unexpected opcode %d "
"in the CQE received for QPID = 0x%0x\n",
CQE_OPCODE(&cqe), CQE_QPID(&cqe));
ret = -EINVAL;
goto out;
}
}
if (cqe_flushed)
wc->status = IB_WC_WR_FLUSH_ERR;
else {
switch (CQE_STATUS(&cqe)) {
case T4_ERR_SUCCESS:
wc->status = IB_WC_SUCCESS;
break;
case T4_ERR_STAG:
wc->status = IB_WC_LOC_ACCESS_ERR;
break;
case T4_ERR_PDID:
wc->status = IB_WC_LOC_PROT_ERR;
break;
case T4_ERR_QPID:
case T4_ERR_ACCESS:
wc->status = IB_WC_LOC_ACCESS_ERR;
break;
case T4_ERR_WRAP:
wc->status = IB_WC_GENERAL_ERR;
break;
case T4_ERR_BOUND:
wc->status = IB_WC_LOC_LEN_ERR;
break;
case T4_ERR_INVALIDATE_SHARED_MR:
case T4_ERR_INVALIDATE_MR_WITH_MW_BOUND:
wc->status = IB_WC_MW_BIND_ERR;
break;
case T4_ERR_CRC:
case T4_ERR_MARKER:
case T4_ERR_PDU_LEN_ERR:
case T4_ERR_OUT_OF_RQE:
case T4_ERR_DDP_VERSION:
case T4_ERR_RDMA_VERSION:
case T4_ERR_DDP_QUEUE_NUM:
case T4_ERR_MSN:
case T4_ERR_TBIT:
case T4_ERR_MO:
case T4_ERR_MSN_RANGE:
case T4_ERR_IRD_OVERFLOW:
case T4_ERR_OPCODE:
case T4_ERR_INTERNAL_ERR:
wc->status = IB_WC_FATAL_ERR;
break;
case T4_ERR_SWFLUSH:
wc->status = IB_WC_WR_FLUSH_ERR;
break;
default:
printf("Unexpected cqe_status 0x%x for QPID = 0x%0x\n",
CQE_STATUS(&cqe), CQE_QPID(&cqe));
ret = -EINVAL;
}
}
out:
if (wq)
spin_unlock(&qhp->lock);
return ret;
}
int c4iw_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
{
struct c4iw_cq *chp;
unsigned long flags;
int npolled;
int err = 0;
chp = to_c4iw_cq(ibcq);
spin_lock_irqsave(&chp->lock, flags);
for (npolled = 0; npolled < num_entries; ++npolled) {
do {
err = c4iw_poll_cq_one(chp, wc + npolled);
} while (err == -EAGAIN);
if (err)
break;
}
spin_unlock_irqrestore(&chp->lock, flags);
return !err || err == -ENODATA ? npolled : err;
}
int c4iw_destroy_cq(struct ib_cq *ib_cq)
{
struct c4iw_cq *chp;
struct c4iw_ucontext *ucontext;
CTR2(KTR_IW_CXGBE, "%s ib_cq %p", __func__, ib_cq);
chp = to_c4iw_cq(ib_cq);
remove_handle(chp->rhp, &chp->rhp->cqidr, chp->cq.cqid);
atomic_dec(&chp->refcnt);
wait_event(chp->wait, !atomic_read(&chp->refcnt));
ucontext = ib_cq->uobject ? to_c4iw_ucontext(ib_cq->uobject->context)
: NULL;
destroy_cq(&chp->rhp->rdev, &chp->cq,
ucontext ? &ucontext->uctx : &chp->cq.rdev->uctx);
kfree(chp);
return 0;
}
struct ib_cq *
c4iw_create_cq(struct ib_device *ibdev, int entries, int vector,
struct ib_ucontext *ib_context, struct ib_udata *udata)
{
struct c4iw_dev *rhp;
struct c4iw_cq *chp;
struct c4iw_create_cq_resp uresp;
struct c4iw_ucontext *ucontext = NULL;
int ret;
size_t memsize, hwentries;
struct c4iw_mm_entry *mm, *mm2;
CTR3(KTR_IW_CXGBE, "%s ib_dev %p entries %d", __func__, ibdev, entries);
rhp = to_c4iw_dev(ibdev);
chp = kzalloc(sizeof(*chp), GFP_KERNEL);
if (!chp)
return ERR_PTR(-ENOMEM);
if (ib_context)
ucontext = to_c4iw_ucontext(ib_context);
/* account for the status page. */
entries++;
/* IQ needs one extra entry to differentiate full vs empty. */
entries++;
/*
* entries must be multiple of 16 for HW.
*/
entries = roundup(entries, 16);
/*
* Make actual HW queue 2x to avoid cidx_inc overflows.
*/
hwentries = entries * 2;
/*
* Make HW queue at least 64 entries so GTS updates aren't too
* frequent.
*/
if (hwentries < 64)
hwentries = 64;
memsize = hwentries * sizeof *chp->cq.queue;
/*
* memsize must be a multiple of the page size if its a user cq.
*/
if (ucontext) {
memsize = roundup(memsize, PAGE_SIZE);
hwentries = memsize / sizeof *chp->cq.queue;
while (hwentries > T4_MAX_IQ_SIZE) {
memsize -= PAGE_SIZE;
hwentries = memsize / sizeof *chp->cq.queue;
}
}
chp->cq.size = hwentries;
chp->cq.memsize = memsize;
ret = create_cq(&rhp->rdev, &chp->cq,
ucontext ? &ucontext->uctx : &rhp->rdev.uctx);
if (ret)
goto err1;
chp->rhp = rhp;
chp->cq.size--; /* status page */
chp->ibcq.cqe = entries - 2;
spin_lock_init(&chp->lock);
spin_lock_init(&chp->comp_handler_lock);
atomic_set(&chp->refcnt, 1);
init_waitqueue_head(&chp->wait);
ret = insert_handle(rhp, &rhp->cqidr, chp, chp->cq.cqid);
if (ret)
goto err2;
if (ucontext) {
mm = kmalloc(sizeof *mm, GFP_KERNEL);
if (!mm)
goto err3;
mm2 = kmalloc(sizeof *mm2, GFP_KERNEL);
if (!mm2)
goto err4;
uresp.qid_mask = rhp->rdev.cqmask;
uresp.cqid = chp->cq.cqid;
uresp.size = chp->cq.size;
uresp.memsize = chp->cq.memsize;
spin_lock(&ucontext->mmap_lock);
uresp.key = ucontext->key;
ucontext->key += PAGE_SIZE;
uresp.gts_key = ucontext->key;
ucontext->key += PAGE_SIZE;
spin_unlock(&ucontext->mmap_lock);
ret = ib_copy_to_udata(udata, &uresp, sizeof uresp);
if (ret)
goto err5;
mm->key = uresp.key;
mm->addr = vtophys(chp->cq.queue);
mm->len = chp->cq.memsize;
insert_mmap(ucontext, mm);
mm2->key = uresp.gts_key;
mm2->addr = chp->cq.ugts;
mm2->len = PAGE_SIZE;
insert_mmap(ucontext, mm2);
}
CTR6(KTR_IW_CXGBE,
"%s cqid 0x%0x chp %p size %u memsize %zu, dma_addr 0x%0llx",
__func__, chp->cq.cqid, chp, chp->cq.size, chp->cq.memsize,
(unsigned long long) chp->cq.dma_addr);
return &chp->ibcq;
err5:
kfree(mm2);
err4:
kfree(mm);
err3:
remove_handle(rhp, &rhp->cqidr, chp->cq.cqid);
err2:
destroy_cq(&chp->rhp->rdev, &chp->cq,
ucontext ? &ucontext->uctx : &rhp->rdev.uctx);
err1:
kfree(chp);
return ERR_PTR(ret);
}
int c4iw_resize_cq(struct ib_cq *cq, int cqe, struct ib_udata *udata)
{
return -ENOSYS;
}
int c4iw_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
{
struct c4iw_cq *chp;
int ret;
unsigned long flag;
chp = to_c4iw_cq(ibcq);
spin_lock_irqsave(&chp->lock, flag);
ret = t4_arm_cq(&chp->cq,
(flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED);
spin_unlock_irqrestore(&chp->lock, flag);
if (ret && !(flags & IB_CQ_REPORT_MISSED_EVENTS))
ret = 0;
return ret;
}
#endif

View File

@ -0,0 +1,368 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#include <sys/ktr.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <rdma/ib_verbs.h>
#include <linux/idr.h>
#ifdef TCP_OFFLOAD
#include "iw_cxgbe.h"
int spg_creds = 2; /* Default status page size is 2 credits = 128B */
void
c4iw_release_dev_ucontext(struct c4iw_rdev *rdev,
struct c4iw_dev_ucontext *uctx)
{
struct list_head *pos, *nxt;
struct c4iw_qid_list *entry;
mutex_lock(&uctx->lock);
list_for_each_safe(pos, nxt, &uctx->qpids) {
entry = list_entry(pos, struct c4iw_qid_list, entry);
list_del_init(&entry->entry);
if (!(entry->qid & rdev->qpmask)) {
c4iw_put_resource(&rdev->resource.qid_table,
entry->qid);
mutex_lock(&rdev->stats.lock);
rdev->stats.qid.cur -= rdev->qpmask + 1;
mutex_unlock(&rdev->stats.lock);
}
kfree(entry);
}
list_for_each_safe(pos, nxt, &uctx->qpids) {
entry = list_entry(pos, struct c4iw_qid_list, entry);
list_del_init(&entry->entry);
kfree(entry);
}
mutex_unlock(&uctx->lock);
}
void
c4iw_init_dev_ucontext(struct c4iw_rdev *rdev, struct c4iw_dev_ucontext *uctx)
{
INIT_LIST_HEAD(&uctx->qpids);
INIT_LIST_HEAD(&uctx->cqids);
mutex_init(&uctx->lock);
}
static int
c4iw_rdev_open(struct c4iw_rdev *rdev)
{
struct adapter *sc = rdev->adap;
int rc;
c4iw_init_dev_ucontext(rdev, &rdev->uctx);
/* Save the status page size set by if_cxgbe */
spg_creds = (t4_read_reg(sc, A_SGE_CONTROL) & F_EGRSTATUSPAGESIZE) ?
2 : 1;
/* XXX: we can probably make this work */
if (sc->sge.eq_s_qpp > PAGE_SHIFT || sc->sge.iq_s_qpp > PAGE_SHIFT) {
device_printf(sc->dev,
"doorbell density too high (eq %d, iq %d, pg %d).\n",
sc->sge.eq_s_qpp, sc->sge.eq_s_qpp, PAGE_SHIFT);
rc = -EINVAL;
goto err1;
}
rdev->qpshift = PAGE_SHIFT - sc->sge.eq_s_qpp;
rdev->qpmask = (1 << sc->sge.eq_s_qpp) - 1;
rdev->cqshift = PAGE_SHIFT - sc->sge.iq_s_qpp;
rdev->cqmask = (1 << sc->sge.iq_s_qpp) - 1;
if (c4iw_num_stags(rdev) == 0) {
rc = -EINVAL;
goto err1;
}
rdev->stats.pd.total = T4_MAX_NUM_PD;
rdev->stats.stag.total = sc->vres.stag.size;
rdev->stats.pbl.total = sc->vres.pbl.size;
rdev->stats.rqt.total = sc->vres.rq.size;
rdev->stats.qid.total = sc->vres.qp.size;
rc = c4iw_init_resource(rdev, c4iw_num_stags(rdev), T4_MAX_NUM_PD);
if (rc) {
device_printf(sc->dev, "error %d initializing resources\n", rc);
goto err1;
}
rc = c4iw_pblpool_create(rdev);
if (rc) {
device_printf(sc->dev, "error %d initializing pbl pool\n", rc);
goto err2;
}
rc = c4iw_rqtpool_create(rdev);
if (rc) {
device_printf(sc->dev, "error %d initializing rqt pool\n", rc);
goto err3;
}
return (0);
err3:
c4iw_pblpool_destroy(rdev);
err2:
c4iw_destroy_resource(&rdev->resource);
err1:
return (rc);
}
static void c4iw_rdev_close(struct c4iw_rdev *rdev)
{
c4iw_pblpool_destroy(rdev);
c4iw_rqtpool_destroy(rdev);
c4iw_destroy_resource(&rdev->resource);
}
static void
c4iw_dealloc(struct c4iw_dev *iwsc)
{
c4iw_rdev_close(&iwsc->rdev);
idr_destroy(&iwsc->cqidr);
idr_destroy(&iwsc->qpidr);
idr_destroy(&iwsc->mmidr);
ib_dealloc_device(&iwsc->ibdev);
}
static struct c4iw_dev *
c4iw_alloc(struct adapter *sc)
{
struct c4iw_dev *iwsc;
int rc;
iwsc = (struct c4iw_dev *)ib_alloc_device(sizeof(*iwsc));
if (iwsc == NULL) {
device_printf(sc->dev, "Cannot allocate ib device.\n");
return (ERR_PTR(-ENOMEM));
}
iwsc->rdev.adap = sc;
rc = c4iw_rdev_open(&iwsc->rdev);
if (rc != 0) {
device_printf(sc->dev, "Unable to open CXIO rdev (%d)\n", rc);
ib_dealloc_device(&iwsc->ibdev);
return (ERR_PTR(rc));
}
idr_init(&iwsc->cqidr);
idr_init(&iwsc->qpidr);
idr_init(&iwsc->mmidr);
spin_lock_init(&iwsc->lock);
mutex_init(&iwsc->rdev.stats.lock);
return (iwsc);
}
static int c4iw_mod_load(void);
static int c4iw_mod_unload(void);
static int c4iw_activate(struct adapter *);
static int c4iw_deactivate(struct adapter *);
static struct uld_info c4iw_uld_info = {
.uld_id = ULD_IWARP,
.activate = c4iw_activate,
.deactivate = c4iw_deactivate,
};
static int
c4iw_activate(struct adapter *sc)
{
struct c4iw_dev *iwsc;
int rc;
ASSERT_SYNCHRONIZED_OP(sc);
if (isset(&sc->offload_map, MAX_NPORTS)) {
KASSERT(0, ("%s: RDMA already eanbled on sc %p", __func__, sc));
return (0);
}
if (sc->rdmacaps == 0) {
device_printf(sc->dev,
"RDMA not supported or RDMA cap is not enabled.\n");
return (ENOSYS);
}
iwsc = c4iw_alloc(sc);
if (IS_ERR(iwsc)) {
rc = -PTR_ERR(iwsc);
device_printf(sc->dev, "initialization failed: %d\n", rc);
return (rc);
}
sc->iwarp_softc = iwsc;
c4iw_cm_init_cpl(sc);
rc = -c4iw_register_device(iwsc);
if (rc) {
device_printf(sc->dev, "RDMA registration failed: %d\n", rc);
c4iw_dealloc(iwsc);
sc->iwarp_softc = NULL;
}
return (rc);
}
static int
c4iw_deactivate(struct adapter *sc)
{
struct c4iw_dev *iwsc = sc->iwarp_softc;
ASSERT_SYNCHRONIZED_OP(sc);
c4iw_unregister_device(iwsc);
c4iw_dealloc(iwsc);
sc->iwarp_softc = NULL;
return (0);
}
static void
c4iw_activate_all(struct adapter *sc, void *arg __unused)
{
if (begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4iwact") != 0)
return;
if (!isset(&sc->offload_map, MAX_NPORTS) &&
t4_activate_uld(sc, ULD_IWARP) == 0)
setbit(&sc->offload_map, MAX_NPORTS);
end_synchronized_op(sc, 0);
}
static void
c4iw_deactivate_all(struct adapter *sc, void *arg __unused)
{
if (begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4iwdea") != 0)
return;
if (isset(&sc->offload_map, MAX_NPORTS) &&
t4_deactivate_uld(sc, ULD_IWARP) == 0)
clrbit(&sc->offload_map, MAX_NPORTS);
end_synchronized_op(sc, 0);
}
static int
c4iw_mod_load(void)
{
int rc;
rc = -c4iw_cm_init();
if (rc != 0)
return (rc);
rc = t4_register_uld(&c4iw_uld_info);
if (rc != 0) {
c4iw_cm_term();
return (rc);
}
t4_iterate(c4iw_activate_all, NULL);
return (rc);
}
static int
c4iw_mod_unload(void)
{
t4_iterate(c4iw_deactivate_all, NULL);
c4iw_cm_term();
if (t4_unregister_uld(&c4iw_uld_info) == EBUSY)
return (EBUSY);
return (0);
}
#endif
#undef MODULE_VERSION
#include <sys/module.h>
/*
* t4_tom won't load on kernels without TCP_OFFLOAD and this module's dependency
* on t4_tom ensures that it won't either. So we don't directly check for
* TCP_OFFLOAD here.
*/
static int
c4iw_modevent(module_t mod, int cmd, void *arg)
{
int rc = 0;
#ifdef TCP_OFFLOAD
switch (cmd) {
case MOD_LOAD:
rc = c4iw_mod_load();
if (rc == 0)
printf("iw_cxgbe: Chelsio T4/T5 RDMA driver loaded.\n");
break;
case MOD_UNLOAD:
rc = c4iw_mod_unload();
break;
default:
rc = EINVAL;
}
#else
printf("t4_tom: compiled without TCP_OFFLOAD support.\n");
rc = EOPNOTSUPP;
#endif
return (rc);
}
static moduledata_t c4iw_mod_data = {
"iw_cxgbe",
c4iw_modevent,
0
};
MODULE_VERSION(iw_cxgbe, 1);
MODULE_DEPEND(iw_cxgbe, t4_tom, 1, 1, 1);
MODULE_DEPEND(iw_cxgbe, ibcore, 1, 1, 1);
DECLARE_MODULE(iw_cxgbe, c4iw_mod_data, SI_SUB_EXEC, SI_ORDER_ANY);

206
sys/dev/cxgbe/iw_cxgbe/ev.c Normal file
View File

@ -0,0 +1,206 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <linux/slab.h>
#include "iw_cxgbe.h"
static void post_qp_event(struct c4iw_dev *dev, struct c4iw_cq *chp,
struct c4iw_qp *qhp,
struct t4_cqe *err_cqe,
enum ib_event_type ib_event)
{
struct ib_event event;
struct c4iw_qp_attributes attrs;
unsigned long flag;
if ((qhp->attr.state == C4IW_QP_STATE_ERROR) ||
(qhp->attr.state == C4IW_QP_STATE_TERMINATE)) {
CTR4(KTR_IW_CXGBE, "%s AE received after RTS - "
"qp state %d qpid 0x%x status 0x%x", __func__,
qhp->attr.state, qhp->wq.sq.qid, CQE_STATUS(err_cqe));
return;
}
printf("AE qpid 0x%x opcode %d status 0x%x "
"type %d wrid.hi 0x%x wrid.lo 0x%x\n",
CQE_QPID(err_cqe), CQE_OPCODE(err_cqe),
CQE_STATUS(err_cqe), CQE_TYPE(err_cqe),
CQE_WRID_HI(err_cqe), CQE_WRID_LOW(err_cqe));
if (qhp->attr.state == C4IW_QP_STATE_RTS) {
attrs.next_state = C4IW_QP_STATE_TERMINATE;
c4iw_modify_qp(qhp->rhp, qhp, C4IW_QP_ATTR_NEXT_STATE,
&attrs, 0);
}
event.event = ib_event;
event.device = chp->ibcq.device;
if (ib_event == IB_EVENT_CQ_ERR)
event.element.cq = &chp->ibcq;
else
event.element.qp = &qhp->ibqp;
if (qhp->ibqp.event_handler)
(*qhp->ibqp.event_handler)(&event, qhp->ibqp.qp_context);
spin_lock_irqsave(&chp->comp_handler_lock, flag);
(*chp->ibcq.comp_handler)(&chp->ibcq, chp->ibcq.cq_context);
spin_unlock_irqrestore(&chp->comp_handler_lock, flag);
}
void c4iw_ev_dispatch(struct c4iw_dev *dev, struct t4_cqe *err_cqe)
{
struct c4iw_cq *chp;
struct c4iw_qp *qhp;
u32 cqid;
spin_lock_irq(&dev->lock);
qhp = get_qhp(dev, CQE_QPID(err_cqe));
if (!qhp) {
printf("BAD AE qpid 0x%x opcode %d "
"status 0x%x type %d wrid.hi 0x%x wrid.lo 0x%x\n",
CQE_QPID(err_cqe),
CQE_OPCODE(err_cqe), CQE_STATUS(err_cqe),
CQE_TYPE(err_cqe), CQE_WRID_HI(err_cqe),
CQE_WRID_LOW(err_cqe));
spin_unlock_irq(&dev->lock);
goto out;
}
if (SQ_TYPE(err_cqe))
cqid = qhp->attr.scq;
else
cqid = qhp->attr.rcq;
chp = get_chp(dev, cqid);
if (!chp) {
printf("BAD AE cqid 0x%x qpid 0x%x opcode %d "
"status 0x%x type %d wrid.hi 0x%x wrid.lo 0x%x\n",
cqid, CQE_QPID(err_cqe),
CQE_OPCODE(err_cqe), CQE_STATUS(err_cqe),
CQE_TYPE(err_cqe), CQE_WRID_HI(err_cqe),
CQE_WRID_LOW(err_cqe));
spin_unlock_irq(&dev->lock);
goto out;
}
c4iw_qp_add_ref(&qhp->ibqp);
atomic_inc(&chp->refcnt);
spin_unlock_irq(&dev->lock);
/* Bad incoming write */
if (RQ_TYPE(err_cqe) &&
(CQE_OPCODE(err_cqe) == FW_RI_RDMA_WRITE)) {
post_qp_event(dev, chp, qhp, err_cqe, IB_EVENT_QP_REQ_ERR);
goto done;
}
switch (CQE_STATUS(err_cqe)) {
/* Completion Events */
case T4_ERR_SUCCESS:
printf(KERN_ERR MOD "AE with status 0!\n");
break;
case T4_ERR_STAG:
case T4_ERR_PDID:
case T4_ERR_QPID:
case T4_ERR_ACCESS:
case T4_ERR_WRAP:
case T4_ERR_BOUND:
case T4_ERR_INVALIDATE_SHARED_MR:
case T4_ERR_INVALIDATE_MR_WITH_MW_BOUND:
post_qp_event(dev, chp, qhp, err_cqe, IB_EVENT_QP_ACCESS_ERR);
break;
/* Device Fatal Errors */
case T4_ERR_ECC:
case T4_ERR_ECC_PSTAG:
case T4_ERR_INTERNAL_ERR:
post_qp_event(dev, chp, qhp, err_cqe, IB_EVENT_DEVICE_FATAL);
break;
/* QP Fatal Errors */
case T4_ERR_OUT_OF_RQE:
case T4_ERR_PBL_ADDR_BOUND:
case T4_ERR_CRC:
case T4_ERR_MARKER:
case T4_ERR_PDU_LEN_ERR:
case T4_ERR_DDP_VERSION:
case T4_ERR_RDMA_VERSION:
case T4_ERR_OPCODE:
case T4_ERR_DDP_QUEUE_NUM:
case T4_ERR_MSN:
case T4_ERR_TBIT:
case T4_ERR_MO:
case T4_ERR_MSN_GAP:
case T4_ERR_MSN_RANGE:
case T4_ERR_RQE_ADDR_BOUND:
case T4_ERR_IRD_OVERFLOW:
post_qp_event(dev, chp, qhp, err_cqe, IB_EVENT_QP_FATAL);
break;
default:
printf("Unknown T4 status 0x%x QPID 0x%x\n",
CQE_STATUS(err_cqe), qhp->wq.sq.qid);
post_qp_event(dev, chp, qhp, err_cqe, IB_EVENT_QP_FATAL);
break;
}
done:
if (atomic_dec_and_test(&chp->refcnt))
wake_up(&chp->wait);
c4iw_qp_rem_ref(&qhp->ibqp);
out:
return;
}
int c4iw_ev_handler(struct sge_iq *iq, const struct rsp_ctrl *rc)
{
struct c4iw_dev *dev = iq->adapter->iwarp_softc;
u32 qid = be32_to_cpu(rc->pldbuflen_qid);
struct c4iw_cq *chp;
unsigned long flag;
chp = get_chp(dev, qid);
if (chp) {
spin_lock_irqsave(&chp->comp_handler_lock, flag);
(*chp->ibcq.comp_handler)(&chp->ibcq, chp->ibcq.cq_context);
spin_unlock_irqrestore(&chp->comp_handler_lock, flag);
} else
CTR2(KTR_IW_CXGBE, "%s unknown cqid 0x%x", __func__, qid);
return 0;
}
#endif

View File

@ -0,0 +1,118 @@
/*
* Copyright (c) 2011-2013 Chelsio Communications. 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <sys/libkern.h>
#include "iw_cxgbe.h"
#define RANDOM_SKIP 16
/*
* Trivial bitmap-based allocator. If the random flag is set, the
* allocator is designed to:
* - pseudo-randomize the id returned such that it is not trivially predictable.
* - avoid reuse of recently used id (at the expense of predictability)
*/
u32 c4iw_id_alloc(struct c4iw_id_table *alloc)
{
unsigned long flags;
u32 obj;
spin_lock_irqsave(&alloc->lock, flags);
obj = find_next_zero_bit(alloc->table, alloc->max, alloc->last);
if (obj >= alloc->max)
obj = find_first_zero_bit(alloc->table, alloc->max);
if (obj < alloc->max) {
if (alloc->flags & C4IW_ID_TABLE_F_RANDOM)
alloc->last += arc4random() % RANDOM_SKIP;
else
alloc->last = obj + 1;
if (alloc->last >= alloc->max)
alloc->last = 0;
set_bit(obj, alloc->table);
obj += alloc->start;
} else
obj = -1;
spin_unlock_irqrestore(&alloc->lock, flags);
return obj;
}
void c4iw_id_free(struct c4iw_id_table *alloc, u32 obj)
{
unsigned long flags;
obj -= alloc->start;
BUG_ON((int)obj < 0);
spin_lock_irqsave(&alloc->lock, flags);
clear_bit(obj, alloc->table);
spin_unlock_irqrestore(&alloc->lock, flags);
}
int c4iw_id_table_alloc(struct c4iw_id_table *alloc, u32 start, u32 num,
u32 reserved, u32 flags)
{
int i;
alloc->start = start;
alloc->flags = flags;
if (flags & C4IW_ID_TABLE_F_RANDOM)
alloc->last = arc4random() % RANDOM_SKIP;
else
alloc->last = 0;
alloc->max = num;
spin_lock_init(&alloc->lock);
alloc->table = kmalloc(BITS_TO_LONGS(num) * sizeof(long),
GFP_KERNEL);
if (!alloc->table)
return -ENOMEM;
bitmap_zero(alloc->table, num);
if (!(alloc->flags & C4IW_ID_TABLE_F_EMPTY))
for (i = 0; i < reserved; ++i)
set_bit(i, alloc->table);
return 0;
}
void c4iw_id_table_free(struct c4iw_id_table *alloc)
{
kfree(alloc->table);
}
#endif

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,828 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <linux/types.h>
#include <linux/kref.h>
#include <rdma/ib_umem.h>
#include <asm/atomic.h>
#include <common/t4_msg.h>
#include "iw_cxgbe.h"
#define T4_ULPTX_MIN_IO 32
#define C4IW_MAX_INLINE_SIZE 96
static int
write_adapter_mem(struct c4iw_rdev *rdev, u32 addr, u32 len, void *data)
{
struct adapter *sc = rdev->adap;
struct ulp_mem_io *ulpmc;
struct ulptx_idata *ulpsc;
u8 wr_len, *to_dp, *from_dp;
int copy_len, num_wqe, i, ret = 0;
struct c4iw_wr_wait wr_wait;
struct wrqe *wr;
u32 cmd;
cmd = cpu_to_be32(V_ULPTX_CMD(ULP_TX_MEM_WRITE));
if (is_t4(sc))
cmd |= cpu_to_be32(F_ULP_MEMIO_ORDER);
else
cmd |= cpu_to_be32(F_T5_ULP_MEMIO_IMM);
addr &= 0x7FFFFFF;
CTR3(KTR_IW_CXGBE, "%s addr 0x%x len %u", __func__, addr, len);
num_wqe = DIV_ROUND_UP(len, C4IW_MAX_INLINE_SIZE);
c4iw_init_wr_wait(&wr_wait);
for (i = 0; i < num_wqe; i++) {
copy_len = min(len, C4IW_MAX_INLINE_SIZE);
wr_len = roundup(sizeof *ulpmc + sizeof *ulpsc +
roundup(copy_len, T4_ULPTX_MIN_IO), 16);
wr = alloc_wrqe(wr_len, &sc->sge.mgmtq);
if (wr == NULL)
return (0);
ulpmc = wrtod(wr);
memset(ulpmc, 0, wr_len);
INIT_ULPTX_WR(ulpmc, wr_len, 0, 0);
if (i == (num_wqe-1)) {
ulpmc->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_ULPTX_WR) |
F_FW_WR_COMPL);
ulpmc->wr.wr_lo = (__force __be64)(unsigned long) &wr_wait;
} else
ulpmc->wr.wr_hi = cpu_to_be32(V_FW_WR_OP(FW_ULPTX_WR));
ulpmc->wr.wr_mid = cpu_to_be32(
V_FW_WR_LEN16(DIV_ROUND_UP(wr_len, 16)));
ulpmc->cmd = cmd;
ulpmc->dlen = cpu_to_be32(V_ULP_MEMIO_DATA_LEN(
DIV_ROUND_UP(copy_len, T4_ULPTX_MIN_IO)));
ulpmc->len16 = cpu_to_be32(DIV_ROUND_UP(wr_len-sizeof(ulpmc->wr),
16));
ulpmc->lock_addr = cpu_to_be32(V_ULP_MEMIO_ADDR(addr + i * 3));
ulpsc = (struct ulptx_idata *)(ulpmc + 1);
ulpsc->cmd_more = cpu_to_be32(V_ULPTX_CMD(ULP_TX_SC_IMM));
ulpsc->len = cpu_to_be32(roundup(copy_len, T4_ULPTX_MIN_IO));
to_dp = (u8 *)(ulpsc + 1);
from_dp = (u8 *)data + i * C4IW_MAX_INLINE_SIZE;
if (data)
memcpy(to_dp, from_dp, copy_len);
else
memset(to_dp, 0, copy_len);
if (copy_len % T4_ULPTX_MIN_IO)
memset(to_dp + copy_len, 0, T4_ULPTX_MIN_IO -
(copy_len % T4_ULPTX_MIN_IO));
t4_wrq_tx(sc, wr);
len -= C4IW_MAX_INLINE_SIZE;
}
ret = c4iw_wait_for_reply(rdev, &wr_wait, 0, 0, __func__);
return ret;
}
/*
* Build and write a TPT entry.
* IN: stag key, pdid, perm, bind_enabled, zbva, to, len, page_size,
* pbl_size and pbl_addr
* OUT: stag index
*/
static int write_tpt_entry(struct c4iw_rdev *rdev, u32 reset_tpt_entry,
u32 *stag, u8 stag_state, u32 pdid,
enum fw_ri_stag_type type, enum fw_ri_mem_perms perm,
int bind_enabled, u32 zbva, u64 to,
u64 len, u8 page_size, u32 pbl_size, u32 pbl_addr)
{
int err;
struct fw_ri_tpte tpt;
u32 stag_idx;
static atomic_t key;
if (c4iw_fatal_error(rdev))
return -EIO;
stag_state = stag_state > 0;
stag_idx = (*stag) >> 8;
if ((!reset_tpt_entry) && (*stag == T4_STAG_UNSET)) {
stag_idx = c4iw_get_resource(&rdev->resource.tpt_table);
if (!stag_idx)
return -ENOMEM;
mutex_lock(&rdev->stats.lock);
rdev->stats.stag.cur += 32;
if (rdev->stats.stag.cur > rdev->stats.stag.max)
rdev->stats.stag.max = rdev->stats.stag.cur;
mutex_unlock(&rdev->stats.lock);
*stag = (stag_idx << 8) | (atomic_inc_return(&key) & 0xff);
}
CTR5(KTR_IW_CXGBE,
"%s stag_state 0x%0x type 0x%0x pdid 0x%0x, stag_idx 0x%x",
__func__, stag_state, type, pdid, stag_idx);
/* write TPT entry */
if (reset_tpt_entry)
memset(&tpt, 0, sizeof(tpt));
else {
tpt.valid_to_pdid = cpu_to_be32(F_FW_RI_TPTE_VALID |
V_FW_RI_TPTE_STAGKEY((*stag & M_FW_RI_TPTE_STAGKEY)) |
V_FW_RI_TPTE_STAGSTATE(stag_state) |
V_FW_RI_TPTE_STAGTYPE(type) | V_FW_RI_TPTE_PDID(pdid));
tpt.locread_to_qpid = cpu_to_be32(V_FW_RI_TPTE_PERM(perm) |
(bind_enabled ? F_FW_RI_TPTE_MWBINDEN : 0) |
V_FW_RI_TPTE_ADDRTYPE((zbva ? FW_RI_ZERO_BASED_TO :
FW_RI_VA_BASED_TO))|
V_FW_RI_TPTE_PS(page_size));
tpt.nosnoop_pbladdr = !pbl_size ? 0 : cpu_to_be32(
V_FW_RI_TPTE_PBLADDR(PBL_OFF(rdev, pbl_addr)>>3));
tpt.len_lo = cpu_to_be32((u32)(len & 0xffffffffUL));
tpt.va_hi = cpu_to_be32((u32)(to >> 32));
tpt.va_lo_fbo = cpu_to_be32((u32)(to & 0xffffffffUL));
tpt.dca_mwbcnt_pstag = cpu_to_be32(0);
tpt.len_hi = cpu_to_be32((u32)(len >> 32));
}
err = write_adapter_mem(rdev, stag_idx +
(rdev->adap->vres.stag.start >> 5),
sizeof(tpt), &tpt);
if (reset_tpt_entry) {
c4iw_put_resource(&rdev->resource.tpt_table, stag_idx);
mutex_lock(&rdev->stats.lock);
rdev->stats.stag.cur -= 32;
mutex_unlock(&rdev->stats.lock);
}
return err;
}
static int write_pbl(struct c4iw_rdev *rdev, __be64 *pbl,
u32 pbl_addr, u32 pbl_size)
{
int err;
CTR4(KTR_IW_CXGBE, "%s *pdb_addr 0x%x, pbl_base 0x%x, pbl_size %d",
__func__, pbl_addr, rdev->adap->vres.pbl.start, pbl_size);
err = write_adapter_mem(rdev, pbl_addr >> 5, pbl_size << 3, pbl);
return err;
}
static int dereg_mem(struct c4iw_rdev *rdev, u32 stag, u32 pbl_size,
u32 pbl_addr)
{
return write_tpt_entry(rdev, 1, &stag, 0, 0, 0, 0, 0, 0, 0UL, 0, 0,
pbl_size, pbl_addr);
}
static int allocate_window(struct c4iw_rdev *rdev, u32 * stag, u32 pdid)
{
*stag = T4_STAG_UNSET;
return write_tpt_entry(rdev, 0, stag, 0, pdid, FW_RI_STAG_MW, 0, 0, 0,
0UL, 0, 0, 0, 0);
}
static int deallocate_window(struct c4iw_rdev *rdev, u32 stag)
{
return write_tpt_entry(rdev, 1, &stag, 0, 0, 0, 0, 0, 0, 0UL, 0, 0, 0,
0);
}
static int allocate_stag(struct c4iw_rdev *rdev, u32 *stag, u32 pdid,
u32 pbl_size, u32 pbl_addr)
{
*stag = T4_STAG_UNSET;
return write_tpt_entry(rdev, 0, stag, 0, pdid, FW_RI_STAG_NSMR, 0, 0, 0,
0UL, 0, 0, pbl_size, pbl_addr);
}
static int finish_mem_reg(struct c4iw_mr *mhp, u32 stag)
{
u32 mmid;
mhp->attr.state = 1;
mhp->attr.stag = stag;
mmid = stag >> 8;
mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
CTR3(KTR_IW_CXGBE, "%s mmid 0x%x mhp %p", __func__, mmid, mhp);
return insert_handle(mhp->rhp, &mhp->rhp->mmidr, mhp, mmid);
}
static int register_mem(struct c4iw_dev *rhp, struct c4iw_pd *php,
struct c4iw_mr *mhp, int shift)
{
u32 stag = T4_STAG_UNSET;
int ret;
ret = write_tpt_entry(&rhp->rdev, 0, &stag, 1, mhp->attr.pdid,
FW_RI_STAG_NSMR, mhp->attr.perms,
mhp->attr.mw_bind_enable, mhp->attr.zbva,
mhp->attr.va_fbo, mhp->attr.len, shift - 12,
mhp->attr.pbl_size, mhp->attr.pbl_addr);
if (ret)
return ret;
ret = finish_mem_reg(mhp, stag);
if (ret)
dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size,
mhp->attr.pbl_addr);
return ret;
}
static int reregister_mem(struct c4iw_dev *rhp, struct c4iw_pd *php,
struct c4iw_mr *mhp, int shift, int npages)
{
u32 stag;
int ret;
if (npages > mhp->attr.pbl_size)
return -ENOMEM;
stag = mhp->attr.stag;
ret = write_tpt_entry(&rhp->rdev, 0, &stag, 1, mhp->attr.pdid,
FW_RI_STAG_NSMR, mhp->attr.perms,
mhp->attr.mw_bind_enable, mhp->attr.zbva,
mhp->attr.va_fbo, mhp->attr.len, shift - 12,
mhp->attr.pbl_size, mhp->attr.pbl_addr);
if (ret)
return ret;
ret = finish_mem_reg(mhp, stag);
if (ret)
dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size,
mhp->attr.pbl_addr);
return ret;
}
static int alloc_pbl(struct c4iw_mr *mhp, int npages)
{
mhp->attr.pbl_addr = c4iw_pblpool_alloc(&mhp->rhp->rdev,
npages << 3);
if (!mhp->attr.pbl_addr)
return -ENOMEM;
mhp->attr.pbl_size = npages;
return 0;
}
static int build_phys_page_list(struct ib_phys_buf *buffer_list,
int num_phys_buf, u64 *iova_start,
u64 *total_size, int *npages,
int *shift, __be64 **page_list)
{
u64 mask;
int i, j, n;
mask = 0;
*total_size = 0;
for (i = 0; i < num_phys_buf; ++i) {
if (i != 0 && buffer_list[i].addr & ~PAGE_MASK)
return -EINVAL;
if (i != 0 && i != num_phys_buf - 1 &&
(buffer_list[i].size & ~PAGE_MASK))
return -EINVAL;
*total_size += buffer_list[i].size;
if (i > 0)
mask |= buffer_list[i].addr;
else
mask |= buffer_list[i].addr & PAGE_MASK;
if (i != num_phys_buf - 1)
mask |= buffer_list[i].addr + buffer_list[i].size;
else
mask |= (buffer_list[i].addr + buffer_list[i].size +
PAGE_SIZE - 1) & PAGE_MASK;
}
if (*total_size > 0xFFFFFFFFULL)
return -ENOMEM;
/* Find largest page shift we can use to cover buffers */
for (*shift = PAGE_SHIFT; *shift < 27; ++(*shift))
if ((1ULL << *shift) & mask)
break;
buffer_list[0].size += buffer_list[0].addr & ((1ULL << *shift) - 1);
buffer_list[0].addr &= ~0ull << *shift;
*npages = 0;
for (i = 0; i < num_phys_buf; ++i)
*npages += (buffer_list[i].size +
(1ULL << *shift) - 1) >> *shift;
if (!*npages)
return -EINVAL;
*page_list = kmalloc(sizeof(u64) * *npages, GFP_KERNEL);
if (!*page_list)
return -ENOMEM;
n = 0;
for (i = 0; i < num_phys_buf; ++i)
for (j = 0;
j < (buffer_list[i].size + (1ULL << *shift) - 1) >> *shift;
++j)
(*page_list)[n++] = cpu_to_be64(buffer_list[i].addr +
((u64) j << *shift));
CTR6(KTR_IW_CXGBE,
"%s va 0x%llx mask 0x%llx shift %d len %lld pbl_size %d", __func__,
(unsigned long long)*iova_start, (unsigned long long)mask, *shift,
(unsigned long long)*total_size, *npages);
return 0;
}
int c4iw_reregister_phys_mem(struct ib_mr *mr, int mr_rereg_mask,
struct ib_pd *pd, struct ib_phys_buf *buffer_list,
int num_phys_buf, int acc, u64 *iova_start)
{
struct c4iw_mr mh, *mhp;
struct c4iw_pd *php;
struct c4iw_dev *rhp;
__be64 *page_list = NULL;
int shift = 0;
u64 total_size;
int npages = 0;
int ret;
CTR3(KTR_IW_CXGBE, "%s ib_mr %p ib_pd %p", __func__, mr, pd);
/* There can be no memory windows */
if (atomic_read(&mr->usecnt))
return -EINVAL;
mhp = to_c4iw_mr(mr);
rhp = mhp->rhp;
php = to_c4iw_pd(mr->pd);
/* make sure we are on the same adapter */
if (rhp != php->rhp)
return -EINVAL;
memcpy(&mh, mhp, sizeof *mhp);
if (mr_rereg_mask & IB_MR_REREG_PD)
php = to_c4iw_pd(pd);
if (mr_rereg_mask & IB_MR_REREG_ACCESS) {
mh.attr.perms = c4iw_ib_to_tpt_access(acc);
mh.attr.mw_bind_enable = (acc & IB_ACCESS_MW_BIND) ==
IB_ACCESS_MW_BIND;
}
if (mr_rereg_mask & IB_MR_REREG_TRANS) {
ret = build_phys_page_list(buffer_list, num_phys_buf,
iova_start,
&total_size, &npages,
&shift, &page_list);
if (ret)
return ret;
}
ret = reregister_mem(rhp, php, &mh, shift, npages);
kfree(page_list);
if (ret)
return ret;
if (mr_rereg_mask & IB_MR_REREG_PD)
mhp->attr.pdid = php->pdid;
if (mr_rereg_mask & IB_MR_REREG_ACCESS)
mhp->attr.perms = c4iw_ib_to_tpt_access(acc);
if (mr_rereg_mask & IB_MR_REREG_TRANS) {
mhp->attr.zbva = 0;
mhp->attr.va_fbo = *iova_start;
mhp->attr.page_size = shift - 12;
mhp->attr.len = (u32) total_size;
mhp->attr.pbl_size = npages;
}
return 0;
}
struct ib_mr *c4iw_register_phys_mem(struct ib_pd *pd,
struct ib_phys_buf *buffer_list,
int num_phys_buf, int acc, u64 *iova_start)
{
__be64 *page_list;
int shift;
u64 total_size;
int npages;
struct c4iw_dev *rhp;
struct c4iw_pd *php;
struct c4iw_mr *mhp;
int ret;
CTR2(KTR_IW_CXGBE, "%s ib_pd %p", __func__, pd);
php = to_c4iw_pd(pd);
rhp = php->rhp;
mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
if (!mhp)
return ERR_PTR(-ENOMEM);
mhp->rhp = rhp;
/* First check that we have enough alignment */
if ((*iova_start & ~PAGE_MASK) != (buffer_list[0].addr & ~PAGE_MASK)) {
ret = -EINVAL;
goto err;
}
if (num_phys_buf > 1 &&
((buffer_list[0].addr + buffer_list[0].size) & ~PAGE_MASK)) {
ret = -EINVAL;
goto err;
}
ret = build_phys_page_list(buffer_list, num_phys_buf, iova_start,
&total_size, &npages, &shift,
&page_list);
if (ret)
goto err;
ret = alloc_pbl(mhp, npages);
if (ret) {
kfree(page_list);
goto err_pbl;
}
ret = write_pbl(&mhp->rhp->rdev, page_list, mhp->attr.pbl_addr,
npages);
kfree(page_list);
if (ret)
goto err_pbl;
mhp->attr.pdid = php->pdid;
mhp->attr.zbva = 0;
mhp->attr.perms = c4iw_ib_to_tpt_access(acc);
mhp->attr.va_fbo = *iova_start;
mhp->attr.page_size = shift - 12;
mhp->attr.len = (u32) total_size;
mhp->attr.pbl_size = npages;
ret = register_mem(rhp, php, mhp, shift);
if (ret)
goto err_pbl;
return &mhp->ibmr;
err_pbl:
c4iw_pblpool_free(&mhp->rhp->rdev, mhp->attr.pbl_addr,
mhp->attr.pbl_size << 3);
err:
kfree(mhp);
return ERR_PTR(ret);
}
struct ib_mr *c4iw_get_dma_mr(struct ib_pd *pd, int acc)
{
struct c4iw_dev *rhp;
struct c4iw_pd *php;
struct c4iw_mr *mhp;
int ret;
u32 stag = T4_STAG_UNSET;
CTR2(KTR_IW_CXGBE, "%s ib_pd %p", __func__, pd);
php = to_c4iw_pd(pd);
rhp = php->rhp;
mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
if (!mhp)
return ERR_PTR(-ENOMEM);
mhp->rhp = rhp;
mhp->attr.pdid = php->pdid;
mhp->attr.perms = c4iw_ib_to_tpt_access(acc);
mhp->attr.mw_bind_enable = (acc&IB_ACCESS_MW_BIND) == IB_ACCESS_MW_BIND;
mhp->attr.zbva = 0;
mhp->attr.va_fbo = 0;
mhp->attr.page_size = 0;
mhp->attr.len = ~0UL;
mhp->attr.pbl_size = 0;
ret = write_tpt_entry(&rhp->rdev, 0, &stag, 1, php->pdid,
FW_RI_STAG_NSMR, mhp->attr.perms,
mhp->attr.mw_bind_enable, 0, 0, ~0UL, 0, 0, 0);
if (ret)
goto err1;
ret = finish_mem_reg(mhp, stag);
if (ret)
goto err2;
return &mhp->ibmr;
err2:
dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size,
mhp->attr.pbl_addr);
err1:
kfree(mhp);
return ERR_PTR(ret);
}
struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
u64 virt, int acc, struct ib_udata *udata, int mr_id)
{
__be64 *pages;
int shift, n, len;
int i, j, k;
int err = 0;
struct ib_umem_chunk *chunk;
struct c4iw_dev *rhp;
struct c4iw_pd *php;
struct c4iw_mr *mhp;
CTR2(KTR_IW_CXGBE, "%s ib_pd %p", __func__, pd);
if (length == ~0ULL)
return ERR_PTR(-EINVAL);
if ((length + start) < start)
return ERR_PTR(-EINVAL);
php = to_c4iw_pd(pd);
rhp = php->rhp;
mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
if (!mhp)
return ERR_PTR(-ENOMEM);
mhp->rhp = rhp;
mhp->umem = ib_umem_get(pd->uobject->context, start, length, acc, 0);
if (IS_ERR(mhp->umem)) {
err = PTR_ERR(mhp->umem);
kfree(mhp);
return ERR_PTR(err);
}
shift = ffs(mhp->umem->page_size) - 1;
n = 0;
list_for_each_entry(chunk, &mhp->umem->chunk_list, list)
n += chunk->nents;
err = alloc_pbl(mhp, n);
if (err)
goto err;
pages = (__be64 *) __get_free_page(GFP_KERNEL);
if (!pages) {
err = -ENOMEM;
goto err_pbl;
}
i = n = 0;
list_for_each_entry(chunk, &mhp->umem->chunk_list, list)
for (j = 0; j < chunk->nmap; ++j) {
len = sg_dma_len(&chunk->page_list[j]) >> shift;
for (k = 0; k < len; ++k) {
pages[i++] = cpu_to_be64(sg_dma_address(
&chunk->page_list[j]) +
mhp->umem->page_size * k);
if (i == PAGE_SIZE / sizeof *pages) {
err = write_pbl(&mhp->rhp->rdev,
pages,
mhp->attr.pbl_addr + (n << 3), i);
if (err)
goto pbl_done;
n += i;
i = 0;
}
}
}
if (i)
err = write_pbl(&mhp->rhp->rdev, pages,
mhp->attr.pbl_addr + (n << 3), i);
pbl_done:
free_page((unsigned long) pages);
if (err)
goto err_pbl;
mhp->attr.pdid = php->pdid;
mhp->attr.zbva = 0;
mhp->attr.perms = c4iw_ib_to_tpt_access(acc);
mhp->attr.va_fbo = virt;
mhp->attr.page_size = shift - 12;
mhp->attr.len = length;
err = register_mem(rhp, php, mhp, shift);
if (err)
goto err_pbl;
return &mhp->ibmr;
err_pbl:
c4iw_pblpool_free(&mhp->rhp->rdev, mhp->attr.pbl_addr,
mhp->attr.pbl_size << 3);
err:
ib_umem_release(mhp->umem);
kfree(mhp);
return ERR_PTR(err);
}
struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd)
{
struct c4iw_dev *rhp;
struct c4iw_pd *php;
struct c4iw_mw *mhp;
u32 mmid;
u32 stag = 0;
int ret;
php = to_c4iw_pd(pd);
rhp = php->rhp;
mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
if (!mhp)
return ERR_PTR(-ENOMEM);
ret = allocate_window(&rhp->rdev, &stag, php->pdid);
if (ret) {
kfree(mhp);
return ERR_PTR(ret);
}
mhp->rhp = rhp;
mhp->attr.pdid = php->pdid;
mhp->attr.type = FW_RI_STAG_MW;
mhp->attr.stag = stag;
mmid = (stag) >> 8;
mhp->ibmw.rkey = stag;
if (insert_handle(rhp, &rhp->mmidr, mhp, mmid)) {
deallocate_window(&rhp->rdev, mhp->attr.stag);
kfree(mhp);
return ERR_PTR(-ENOMEM);
}
CTR4(KTR_IW_CXGBE, "%s mmid 0x%x mhp %p stag 0x%x", __func__, mmid, mhp,
stag);
return &(mhp->ibmw);
}
int c4iw_dealloc_mw(struct ib_mw *mw)
{
struct c4iw_dev *rhp;
struct c4iw_mw *mhp;
u32 mmid;
mhp = to_c4iw_mw(mw);
rhp = mhp->rhp;
mmid = (mw->rkey) >> 8;
remove_handle(rhp, &rhp->mmidr, mmid);
deallocate_window(&rhp->rdev, mhp->attr.stag);
kfree(mhp);
CTR4(KTR_IW_CXGBE, "%s ib_mw %p mmid 0x%x ptr %p", __func__, mw, mmid,
mhp);
return 0;
}
struct ib_mr *c4iw_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth)
{
struct c4iw_dev *rhp;
struct c4iw_pd *php;
struct c4iw_mr *mhp;
u32 mmid;
u32 stag = 0;
int ret = 0;
php = to_c4iw_pd(pd);
rhp = php->rhp;
mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
if (!mhp) {
ret = -ENOMEM;
goto err;
}
mhp->rhp = rhp;
ret = alloc_pbl(mhp, pbl_depth);
if (ret)
goto err1;
mhp->attr.pbl_size = pbl_depth;
ret = allocate_stag(&rhp->rdev, &stag, php->pdid,
mhp->attr.pbl_size, mhp->attr.pbl_addr);
if (ret)
goto err2;
mhp->attr.pdid = php->pdid;
mhp->attr.type = FW_RI_STAG_NSMR;
mhp->attr.stag = stag;
mhp->attr.state = 1;
mmid = (stag) >> 8;
mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
if (insert_handle(rhp, &rhp->mmidr, mhp, mmid)) {
ret = -ENOMEM;
goto err3;
}
CTR4(KTR_IW_CXGBE, "%s mmid 0x%x mhp %p stag 0x%x", __func__, mmid, mhp,
stag);
return &(mhp->ibmr);
err3:
dereg_mem(&rhp->rdev, stag, mhp->attr.pbl_size,
mhp->attr.pbl_addr);
err2:
c4iw_pblpool_free(&mhp->rhp->rdev, mhp->attr.pbl_addr,
mhp->attr.pbl_size << 3);
err1:
kfree(mhp);
err:
return ERR_PTR(ret);
}
struct ib_fast_reg_page_list *c4iw_alloc_fastreg_pbl(struct ib_device *device,
int page_list_len)
{
struct c4iw_fr_page_list *c4pl;
struct c4iw_dev *dev = to_c4iw_dev(device);
bus_addr_t dma_addr;
int size = sizeof *c4pl + page_list_len * sizeof(u64);
c4pl = contigmalloc(size,
M_DEVBUF, M_NOWAIT, 0ul, ~0ul, 4096, 0);
if (c4pl)
dma_addr = vtophys(c4pl);
else
return ERR_PTR(-ENOMEM);;
pci_unmap_addr_set(c4pl, mapping, dma_addr);
c4pl->dma_addr = dma_addr;
c4pl->dev = dev;
c4pl->size = size;
c4pl->ibpl.page_list = (u64 *)(c4pl + 1);
c4pl->ibpl.max_page_list_len = page_list_len;
return &c4pl->ibpl;
}
void c4iw_free_fastreg_pbl(struct ib_fast_reg_page_list *ibpl)
{
struct c4iw_fr_page_list *c4pl = to_c4iw_fr_page_list(ibpl);
contigfree(c4pl, c4pl->size, M_DEVBUF);
}
int c4iw_dereg_mr(struct ib_mr *ib_mr)
{
struct c4iw_dev *rhp;
struct c4iw_mr *mhp;
u32 mmid;
CTR2(KTR_IW_CXGBE, "%s ib_mr %p", __func__, ib_mr);
/* There can be no memory windows */
if (atomic_read(&ib_mr->usecnt))
return -EINVAL;
mhp = to_c4iw_mr(ib_mr);
rhp = mhp->rhp;
mmid = mhp->attr.stag >> 8;
remove_handle(rhp, &rhp->mmidr, mmid);
dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size,
mhp->attr.pbl_addr);
if (mhp->attr.pbl_size)
c4iw_pblpool_free(&mhp->rhp->rdev, mhp->attr.pbl_addr,
mhp->attr.pbl_size << 3);
if (mhp->kva)
kfree((void *) (unsigned long) mhp->kva);
if (mhp->umem)
ib_umem_release(mhp->umem);
CTR3(KTR_IW_CXGBE, "%s mmid 0x%x ptr %p", __func__, mmid, mhp);
kfree(mhp);
return 0;
}
#endif

View File

@ -0,0 +1,498 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <asm/pgtable.h>
#include <linux/page.h>
#include <rdma/ib_verbs.h>
#include <rdma/ib_user_verbs.h>
#include "iw_cxgbe.h"
#include "user.h"
static int fastreg_support = 1;
module_param(fastreg_support, int, 0644);
MODULE_PARM_DESC(fastreg_support, "Advertise fastreg support (default = 1)");
static int c4iw_modify_port(struct ib_device *ibdev,
u8 port, int port_modify_mask,
struct ib_port_modify *props)
{
return -ENOSYS;
}
static struct ib_ah *c4iw_ah_create(struct ib_pd *pd,
struct ib_ah_attr *ah_attr)
{
return ERR_PTR(-ENOSYS);
}
static int c4iw_ah_destroy(struct ib_ah *ah)
{
return -ENOSYS;
}
static int c4iw_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
return -ENOSYS;
}
static int c4iw_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
return -ENOSYS;
}
static int c4iw_process_mad(struct ib_device *ibdev, int mad_flags,
u8 port_num, struct ib_wc *in_wc,
struct ib_grh *in_grh, struct ib_mad *in_mad,
struct ib_mad *out_mad)
{
return -ENOSYS;
}
static int c4iw_dealloc_ucontext(struct ib_ucontext *context)
{
struct c4iw_dev *rhp = to_c4iw_dev(context->device);
struct c4iw_ucontext *ucontext = to_c4iw_ucontext(context);
struct c4iw_mm_entry *mm, *tmp;
CTR2(KTR_IW_CXGBE, "%s context %p", __func__, context);
list_for_each_entry_safe(mm, tmp, &ucontext->mmaps, entry)
kfree(mm);
c4iw_release_dev_ucontext(&rhp->rdev, &ucontext->uctx);
kfree(ucontext);
return 0;
}
static struct ib_ucontext *c4iw_alloc_ucontext(struct ib_device *ibdev,
struct ib_udata *udata)
{
struct c4iw_ucontext *context;
struct c4iw_dev *rhp = to_c4iw_dev(ibdev);
CTR2(KTR_IW_CXGBE, "%s ibdev %p", __func__, ibdev);
context = kzalloc(sizeof(*context), GFP_KERNEL);
if (!context)
return ERR_PTR(-ENOMEM);
c4iw_init_dev_ucontext(&rhp->rdev, &context->uctx);
INIT_LIST_HEAD(&context->mmaps);
spin_lock_init(&context->mmap_lock);
return &context->ibucontext;
}
static inline pgprot_t t4_pgprot_wc(pgprot_t prot)
{
return pgprot_writecombine(prot);
}
static int c4iw_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
{
int len = vma->vm_end - vma->vm_start;
u32 key = vma->vm_pgoff << PAGE_SHIFT;
struct c4iw_rdev *rdev;
int ret = 0;
struct c4iw_mm_entry *mm;
struct c4iw_ucontext *ucontext;
u64 addr, paddr;
u64 va_regs_res = 0, va_udbs_res = 0;
u64 len_regs_res = 0, len_udbs_res = 0;
CTR3(KTR_IW_CXGBE, "%s:1 ctx %p vma %p", __func__, context, vma);
CTR4(KTR_IW_CXGBE, "%s:1a pgoff 0x%lx key 0x%x len %d", __func__,
vma->vm_pgoff, key, len);
if (vma->vm_start & (PAGE_SIZE-1)) {
CTR3(KTR_IW_CXGBE, "%s:2 unaligned vm_start %u vma %p",
__func__, vma->vm_start, vma);
return -EINVAL;
}
rdev = &(to_c4iw_dev(context->device)->rdev);
ucontext = to_c4iw_ucontext(context);
mm = remove_mmap(ucontext, key, len);
if (!mm) {
CTR4(KTR_IW_CXGBE, "%s:3 ucontext %p key %u len %u", __func__,
ucontext, key, len);
return -EINVAL;
}
addr = mm->addr;
kfree(mm);
va_regs_res = (u64)rman_get_virtual(rdev->adap->regs_res);
len_regs_res = (u64)rman_get_size(rdev->adap->regs_res);
va_udbs_res = (u64)rman_get_virtual(rdev->adap->udbs_res);
len_udbs_res = (u64)rman_get_size(rdev->adap->udbs_res);
CTR6(KTR_IW_CXGBE,
"%s:4 addr %p, masync region %p:%p, udb region %p:%p", __func__,
addr, va_regs_res, va_regs_res+len_regs_res, va_udbs_res,
va_udbs_res+len_udbs_res);
if (addr >= va_regs_res && addr < va_regs_res + len_regs_res) {
CTR4(KTR_IW_CXGBE, "%s:5 MA_SYNC addr %p region %p, reglen %u",
__func__, addr, va_regs_res, len_regs_res);
/*
* MA_SYNC register...
*/
paddr = vtophys(addr);
vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
ret = io_remap_pfn_range(vma, vma->vm_start,
paddr >> PAGE_SHIFT,
len, vma->vm_page_prot);
} else {
if (addr >= va_udbs_res && addr < va_udbs_res + len_udbs_res) {
/*
* Map user DB or OCQP memory...
*/
paddr = vtophys(addr);
CTR4(KTR_IW_CXGBE,
"%s:6 USER DB-GTS addr %p region %p, reglen %u",
__func__, addr, va_udbs_res, len_udbs_res);
#ifdef DOT5
if (is_t5(rdev->lldi.adapter_type) && map_udb_as_wc)
vma->vm_page_prot = t4_pgprot_wc(vma->vm_page_prot);
else
#endif
vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
ret = io_remap_pfn_range(vma, vma->vm_start,
paddr >> PAGE_SHIFT,
len, vma->vm_page_prot);
} else {
/*
* Map WQ or CQ contig dma memory...
*/
CTR4(KTR_IW_CXGBE,
"%s:7 WQ/CQ addr %p vm_start %u vma %p", __func__,
addr, vma->vm_start, vma);
ret = io_remap_pfn_range(vma, vma->vm_start,
addr >> PAGE_SHIFT,
len, vma->vm_page_prot);
}
}
CTR4(KTR_IW_CXGBE, "%s:8 ctx %p vma %p ret %u", __func__, context, vma,
ret);
return ret;
}
static int
c4iw_deallocate_pd(struct ib_pd *pd)
{
struct c4iw_pd *php = to_c4iw_pd(pd);
struct c4iw_dev *rhp = php->rhp;
CTR3(KTR_IW_CXGBE, "%s: pd %p, pdid 0x%x", __func__, pd, php->pdid);
c4iw_put_resource(&rhp->rdev.resource.pdid_table, php->pdid);
mutex_lock(&rhp->rdev.stats.lock);
rhp->rdev.stats.pd.cur--;
mutex_unlock(&rhp->rdev.stats.lock);
kfree(php);
return (0);
}
static struct ib_pd *
c4iw_allocate_pd(struct ib_device *ibdev, struct ib_ucontext *context,
struct ib_udata *udata)
{
struct c4iw_pd *php;
u32 pdid;
struct c4iw_dev *rhp;
CTR4(KTR_IW_CXGBE, "%s: ibdev %p, context %p, data %p", __func__, ibdev,
context, udata);
rhp = (struct c4iw_dev *) ibdev;
pdid = c4iw_get_resource(&rhp->rdev.resource.pdid_table);
if (!pdid)
return ERR_PTR(-EINVAL);
php = kzalloc(sizeof(*php), GFP_KERNEL);
if (!php) {
c4iw_put_resource(&rhp->rdev.resource.pdid_table, pdid);
return ERR_PTR(-ENOMEM);
}
php->pdid = pdid;
php->rhp = rhp;
if (context) {
if (ib_copy_to_udata(udata, &php->pdid, sizeof(u32))) {
c4iw_deallocate_pd(&php->ibpd);
return ERR_PTR(-EFAULT);
}
}
mutex_lock(&rhp->rdev.stats.lock);
rhp->rdev.stats.pd.cur++;
if (rhp->rdev.stats.pd.cur > rhp->rdev.stats.pd.max)
rhp->rdev.stats.pd.max = rhp->rdev.stats.pd.cur;
mutex_unlock(&rhp->rdev.stats.lock);
CTR6(KTR_IW_CXGBE,
"%s: ibdev %p, context %p, data %p, pddid 0x%x, pd %p", __func__,
ibdev, context, udata, pdid, php);
return (&php->ibpd);
}
static int
c4iw_query_pkey(struct ib_device *ibdev, u8 port, u16 index, u16 *pkey)
{
CTR5(KTR_IW_CXGBE, "%s ibdev %p, port %d, index %d, pkey %p", __func__,
ibdev, port, index, pkey);
*pkey = 0;
return (0);
}
static int
c4iw_query_gid(struct ib_device *ibdev, u8 port, int index, union ib_gid *gid)
{
struct c4iw_dev *dev;
struct port_info *pi;
struct adapter *sc;
CTR5(KTR_IW_CXGBE, "%s ibdev %p, port %d, index %d, gid %p", __func__,
ibdev, port, index, gid);
memset(&gid->raw[0], 0, sizeof(gid->raw));
dev = to_c4iw_dev(ibdev);
sc = dev->rdev.adap;
if (port == 0 || port > sc->params.nports)
return (-EINVAL);
pi = sc->port[port - 1];
memcpy(&gid->raw[0], pi->hw_addr, sizeof(pi->hw_addr));
return (0);
}
static int
c4iw_query_device(struct ib_device *ibdev, struct ib_device_attr *props)
{
struct c4iw_dev *dev = to_c4iw_dev(ibdev);
struct adapter *sc = dev->rdev.adap;
CTR3(KTR_IW_CXGBE, "%s ibdev %p, props %p", __func__, ibdev, props);
memset(props, 0, sizeof *props);
memcpy(&props->sys_image_guid, sc->port[0]->hw_addr, 6);
props->hw_ver = sc->params.chipid;
props->fw_ver = sc->params.fw_vers;
props->device_cap_flags = dev->device_cap_flags;
props->page_size_cap = T4_PAGESIZE_MASK;
props->vendor_id = pci_get_vendor(sc->dev);
props->vendor_part_id = pci_get_device(sc->dev);
props->max_mr_size = T4_MAX_MR_SIZE;
props->max_qp = T4_MAX_NUM_QP;
props->max_qp_wr = T4_MAX_QP_DEPTH;
props->max_sge = T4_MAX_RECV_SGE;
props->max_sge_rd = 1;
props->max_qp_rd_atom = c4iw_max_read_depth;
props->max_qp_init_rd_atom = c4iw_max_read_depth;
props->max_cq = T4_MAX_NUM_CQ;
props->max_cqe = T4_MAX_CQ_DEPTH;
props->max_mr = c4iw_num_stags(&dev->rdev);
props->max_pd = T4_MAX_NUM_PD;
props->local_ca_ack_delay = 0;
props->max_fast_reg_page_list_len = T4_MAX_FR_DEPTH;
return (0);
}
/*
* Returns -errno on failure.
*/
static int
c4iw_query_port(struct ib_device *ibdev, u8 port, struct ib_port_attr *props)
{
struct c4iw_dev *dev;
struct adapter *sc;
struct port_info *pi;
struct ifnet *ifp;
CTR4(KTR_IW_CXGBE, "%s ibdev %p, port %d, props %p", __func__, ibdev,
port, props);
dev = to_c4iw_dev(ibdev);
sc = dev->rdev.adap;
if (port > sc->params.nports)
return (-EINVAL);
pi = sc->port[port - 1];
ifp = pi->ifp;
memset(props, 0, sizeof(struct ib_port_attr));
props->max_mtu = IB_MTU_4096;
if (ifp->if_mtu >= 4096)
props->active_mtu = IB_MTU_4096;
else if (ifp->if_mtu >= 2048)
props->active_mtu = IB_MTU_2048;
else if (ifp->if_mtu >= 1024)
props->active_mtu = IB_MTU_1024;
else if (ifp->if_mtu >= 512)
props->active_mtu = IB_MTU_512;
else
props->active_mtu = IB_MTU_256;
props->state = pi->link_cfg.link_ok ? IB_PORT_ACTIVE : IB_PORT_DOWN;
props->port_cap_flags =
IB_PORT_CM_SUP |
IB_PORT_SNMP_TUNNEL_SUP |
IB_PORT_REINIT_SUP |
IB_PORT_DEVICE_MGMT_SUP |
IB_PORT_VENDOR_CLASS_SUP | IB_PORT_BOOT_MGMT_SUP;
props->gid_tbl_len = 1;
props->pkey_tbl_len = 1;
props->active_width = 2;
props->active_speed = 2;
props->max_msg_sz = -1;
return 0;
}
/*
* Returns -errno on error.
*/
int
c4iw_register_device(struct c4iw_dev *dev)
{
struct adapter *sc = dev->rdev.adap;
struct ib_device *ibdev = &dev->ibdev;
struct iw_cm_verbs *iwcm;
int ret;
CTR3(KTR_IW_CXGBE, "%s c4iw_dev %p, adapter %p", __func__, dev, sc);
BUG_ON(!sc->port[0]);
strlcpy(ibdev->name, device_get_nameunit(sc->dev), sizeof(ibdev->name));
memset(&ibdev->node_guid, 0, sizeof(ibdev->node_guid));
memcpy(&ibdev->node_guid, sc->port[0]->hw_addr, 6);
ibdev->owner = THIS_MODULE;
dev->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW;
if (fastreg_support)
dev->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
ibdev->local_dma_lkey = 0;
ibdev->uverbs_cmd_mask =
(1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE) |
(1ull << IB_USER_VERBS_CMD_QUERY_PORT) |
(1ull << IB_USER_VERBS_CMD_ALLOC_PD) |
(1ull << IB_USER_VERBS_CMD_DEALLOC_PD) |
(1ull << IB_USER_VERBS_CMD_REG_MR) |
(1ull << IB_USER_VERBS_CMD_DEREG_MR) |
(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
(1ull << IB_USER_VERBS_CMD_CREATE_CQ) |
(1ull << IB_USER_VERBS_CMD_DESTROY_CQ) |
(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ) |
(1ull << IB_USER_VERBS_CMD_CREATE_QP) |
(1ull << IB_USER_VERBS_CMD_MODIFY_QP) |
(1ull << IB_USER_VERBS_CMD_QUERY_QP) |
(1ull << IB_USER_VERBS_CMD_POLL_CQ) |
(1ull << IB_USER_VERBS_CMD_DESTROY_QP) |
(1ull << IB_USER_VERBS_CMD_POST_SEND) |
(1ull << IB_USER_VERBS_CMD_POST_RECV);
ibdev->node_type = RDMA_NODE_RNIC;
strlcpy(ibdev->node_desc, C4IW_NODE_DESC, sizeof(ibdev->node_desc));
ibdev->phys_port_cnt = sc->params.nports;
ibdev->num_comp_vectors = 1;
ibdev->dma_device = sc->dev;
ibdev->query_device = c4iw_query_device;
ibdev->query_port = c4iw_query_port;
ibdev->modify_port = c4iw_modify_port;
ibdev->query_pkey = c4iw_query_pkey;
ibdev->query_gid = c4iw_query_gid;
ibdev->alloc_ucontext = c4iw_alloc_ucontext;
ibdev->dealloc_ucontext = c4iw_dealloc_ucontext;
ibdev->mmap = c4iw_mmap;
ibdev->alloc_pd = c4iw_allocate_pd;
ibdev->dealloc_pd = c4iw_deallocate_pd;
ibdev->create_ah = c4iw_ah_create;
ibdev->destroy_ah = c4iw_ah_destroy;
ibdev->create_qp = c4iw_create_qp;
ibdev->modify_qp = c4iw_ib_modify_qp;
ibdev->query_qp = c4iw_ib_query_qp;
ibdev->destroy_qp = c4iw_destroy_qp;
ibdev->create_cq = c4iw_create_cq;
ibdev->destroy_cq = c4iw_destroy_cq;
ibdev->resize_cq = c4iw_resize_cq;
ibdev->poll_cq = c4iw_poll_cq;
ibdev->get_dma_mr = c4iw_get_dma_mr;
ibdev->reg_phys_mr = c4iw_register_phys_mem;
ibdev->rereg_phys_mr = c4iw_reregister_phys_mem;
ibdev->reg_user_mr = c4iw_reg_user_mr;
ibdev->dereg_mr = c4iw_dereg_mr;
ibdev->alloc_mw = c4iw_alloc_mw;
ibdev->bind_mw = c4iw_bind_mw;
ibdev->dealloc_mw = c4iw_dealloc_mw;
ibdev->alloc_fast_reg_mr = c4iw_alloc_fast_reg_mr;
ibdev->alloc_fast_reg_page_list = c4iw_alloc_fastreg_pbl;
ibdev->free_fast_reg_page_list = c4iw_free_fastreg_pbl;
ibdev->attach_mcast = c4iw_multicast_attach;
ibdev->detach_mcast = c4iw_multicast_detach;
ibdev->process_mad = c4iw_process_mad;
ibdev->req_notify_cq = c4iw_arm_cq;
ibdev->post_send = c4iw_post_send;
ibdev->post_recv = c4iw_post_receive;
ibdev->uverbs_abi_ver = C4IW_UVERBS_ABI_VERSION;
iwcm = kmalloc(sizeof(*iwcm), GFP_KERNEL);
if (iwcm == NULL)
return (-ENOMEM);
iwcm->connect = c4iw_connect;
iwcm->accept = c4iw_accept_cr;
iwcm->reject = c4iw_reject_cr;
iwcm->create_listen = c4iw_create_listen;
iwcm->destroy_listen = c4iw_destroy_listen;
iwcm->add_ref = c4iw_qp_add_ref;
iwcm->rem_ref = c4iw_qp_rem_ref;
iwcm->get_qp = c4iw_get_qp;
ibdev->iwcm = iwcm;
ret = ib_register_device(&dev->ibdev, NULL);
if (ret)
kfree(iwcm);
return (ret);
}
void
c4iw_unregister_device(struct c4iw_dev *dev)
{
CTR3(KTR_IW_CXGBE, "%s c4iw_dev %p, adapter %p", __func__, dev,
dev->rdev.adap);
ib_unregister_device(&dev->ibdev);
kfree(dev->ibdev.iwcm);
return;
}
#endif

1707
sys/dev/cxgbe/iw_cxgbe/qp.c Normal file

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,342 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*/
/* Crude resource management */
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <linux/spinlock.h>
#include "iw_cxgbe.h"
static int c4iw_init_qid_table(struct c4iw_rdev *rdev)
{
u32 i;
if (c4iw_id_table_alloc(&rdev->resource.qid_table,
rdev->adap->vres.qp.start,
rdev->adap->vres.qp.size,
rdev->adap->vres.qp.size, 0)) {
printf("%s: return ENOMEM\n", __func__);
return -ENOMEM;
}
for (i = rdev->adap->vres.qp.start;
i < rdev->adap->vres.qp.start + rdev->adap->vres.qp.size; i++)
if (!(i & rdev->qpmask))
c4iw_id_free(&rdev->resource.qid_table, i);
return 0;
}
/* nr_* must be power of 2 */
int c4iw_init_resource(struct c4iw_rdev *rdev, u32 nr_tpt, u32 nr_pdid)
{
int err = 0;
err = c4iw_id_table_alloc(&rdev->resource.tpt_table, 0, nr_tpt, 1,
C4IW_ID_TABLE_F_RANDOM);
if (err)
goto tpt_err;
err = c4iw_init_qid_table(rdev);
if (err)
goto qid_err;
err = c4iw_id_table_alloc(&rdev->resource.pdid_table, 0,
nr_pdid, 1, 0);
if (err)
goto pdid_err;
return 0;
pdid_err:
c4iw_id_table_free(&rdev->resource.qid_table);
qid_err:
c4iw_id_table_free(&rdev->resource.tpt_table);
tpt_err:
return -ENOMEM;
}
/*
* returns 0 if no resource available
*/
u32 c4iw_get_resource(struct c4iw_id_table *id_table)
{
u32 entry;
entry = c4iw_id_alloc(id_table);
if (entry == (u32)(-1)) {
return 0;
}
return entry;
}
void c4iw_put_resource(struct c4iw_id_table *id_table, u32 entry)
{
CTR2(KTR_IW_CXGBE, "%s entry 0x%x", __func__, entry);
c4iw_id_free(id_table, entry);
}
u32 c4iw_get_cqid(struct c4iw_rdev *rdev, struct c4iw_dev_ucontext *uctx)
{
struct c4iw_qid_list *entry;
u32 qid;
int i;
mutex_lock(&uctx->lock);
if (!list_empty(&uctx->cqids)) {
entry = list_entry(uctx->cqids.next, struct c4iw_qid_list,
entry);
list_del(&entry->entry);
qid = entry->qid;
kfree(entry);
} else {
qid = c4iw_get_resource(&rdev->resource.qid_table);
if (!qid)
goto out;
mutex_lock(&rdev->stats.lock);
rdev->stats.qid.cur += rdev->qpmask + 1;
mutex_unlock(&rdev->stats.lock);
for (i = qid+1; i & rdev->qpmask; i++) {
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
goto out;
entry->qid = i;
list_add_tail(&entry->entry, &uctx->cqids);
}
/*
* now put the same ids on the qp list since they all
* map to the same db/gts page.
*/
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
goto out;
entry->qid = qid;
list_add_tail(&entry->entry, &uctx->qpids);
for (i = qid+1; i & rdev->qpmask; i++) {
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
goto out;
entry->qid = i;
list_add_tail(&entry->entry, &uctx->qpids);
}
}
out:
mutex_unlock(&uctx->lock);
CTR2(KTR_IW_CXGBE, "%s: qid 0x%x", __func__, qid);
mutex_lock(&rdev->stats.lock);
if (rdev->stats.qid.cur > rdev->stats.qid.max)
rdev->stats.qid.max = rdev->stats.qid.cur;
mutex_unlock(&rdev->stats.lock);
return qid;
}
void c4iw_put_cqid(struct c4iw_rdev *rdev, u32 qid,
struct c4iw_dev_ucontext *uctx)
{
struct c4iw_qid_list *entry;
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
return;
CTR2(KTR_IW_CXGBE, "%s qid 0x%x", __func__, qid);
entry->qid = qid;
mutex_lock(&uctx->lock);
list_add_tail(&entry->entry, &uctx->cqids);
mutex_unlock(&uctx->lock);
}
u32 c4iw_get_qpid(struct c4iw_rdev *rdev, struct c4iw_dev_ucontext *uctx)
{
struct c4iw_qid_list *entry;
u32 qid;
int i;
mutex_lock(&uctx->lock);
if (!list_empty(&uctx->qpids)) {
entry = list_entry(uctx->qpids.next, struct c4iw_qid_list,
entry);
list_del(&entry->entry);
qid = entry->qid;
kfree(entry);
} else {
qid = c4iw_get_resource(&rdev->resource.qid_table);
if (!qid)
goto out;
mutex_lock(&rdev->stats.lock);
rdev->stats.qid.cur += rdev->qpmask + 1;
mutex_unlock(&rdev->stats.lock);
for (i = qid+1; i & rdev->qpmask; i++) {
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
goto out;
entry->qid = i;
list_add_tail(&entry->entry, &uctx->qpids);
}
/*
* now put the same ids on the cq list since they all
* map to the same db/gts page.
*/
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
goto out;
entry->qid = qid;
list_add_tail(&entry->entry, &uctx->cqids);
for (i = qid; i & rdev->qpmask; i++) {
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
goto out;
entry->qid = i;
list_add_tail(&entry->entry, &uctx->cqids);
}
}
out:
mutex_unlock(&uctx->lock);
CTR2(KTR_IW_CXGBE, "%s qid 0x%x", __func__, qid);
mutex_lock(&rdev->stats.lock);
if (rdev->stats.qid.cur > rdev->stats.qid.max)
rdev->stats.qid.max = rdev->stats.qid.cur;
mutex_unlock(&rdev->stats.lock);
return qid;
}
void c4iw_put_qpid(struct c4iw_rdev *rdev, u32 qid,
struct c4iw_dev_ucontext *uctx)
{
struct c4iw_qid_list *entry;
entry = kmalloc(sizeof *entry, GFP_KERNEL);
if (!entry)
return;
CTR2(KTR_IW_CXGBE, "%s qid 0x%x", __func__, qid);
entry->qid = qid;
mutex_lock(&uctx->lock);
list_add_tail(&entry->entry, &uctx->qpids);
mutex_unlock(&uctx->lock);
}
void c4iw_destroy_resource(struct c4iw_resource *rscp)
{
c4iw_id_table_free(&rscp->tpt_table);
c4iw_id_table_free(&rscp->qid_table);
c4iw_id_table_free(&rscp->pdid_table);
}
/* PBL Memory Manager. Uses Linux generic allocator. */
#define MIN_PBL_SHIFT 8 /* 256B == min PBL size (32 entries) */
u32 c4iw_pblpool_alloc(struct c4iw_rdev *rdev, int size)
{
unsigned long addr = gen_pool_alloc(rdev->pbl_pool, size);
CTR3(KTR_IW_CXGBE, "%s addr 0x%x size %d", __func__, (u32)addr, size);
mutex_lock(&rdev->stats.lock);
if (addr) {
rdev->stats.pbl.cur += roundup(size, 1 << MIN_PBL_SHIFT);
if (rdev->stats.pbl.cur > rdev->stats.pbl.max)
rdev->stats.pbl.max = rdev->stats.pbl.cur;
} else
rdev->stats.pbl.fail++;
mutex_unlock(&rdev->stats.lock);
return (u32)addr;
}
void c4iw_pblpool_free(struct c4iw_rdev *rdev, u32 addr, int size)
{
CTR3(KTR_IW_CXGBE, "%s addr 0x%x size %d", __func__, addr, size);
mutex_lock(&rdev->stats.lock);
rdev->stats.pbl.cur -= roundup(size, 1 << MIN_PBL_SHIFT);
mutex_unlock(&rdev->stats.lock);
gen_pool_free(rdev->pbl_pool, (unsigned long)addr, size);
}
int c4iw_pblpool_create(struct c4iw_rdev *rdev)
{
rdev->pbl_pool = gen_pool_create(rdev->adap->vres.pbl.start,
MIN_PBL_SHIFT,
rdev->adap->vres.pbl.size);
if (!rdev->pbl_pool)
return -ENOMEM;
return 0;
}
void c4iw_pblpool_destroy(struct c4iw_rdev *rdev)
{
gen_pool_destroy(rdev->pbl_pool);
}
/* RQT Memory Manager. Uses Linux generic allocator. */
#define MIN_RQT_SHIFT 10 /* 1KB == min RQT size (16 entries) */
u32 c4iw_rqtpool_alloc(struct c4iw_rdev *rdev, int size)
{
unsigned long addr = gen_pool_alloc(rdev->rqt_pool, size << 6);
CTR3(KTR_IW_CXGBE, "%s addr 0x%x size %d", __func__, (u32)addr,
size << 6);
if (!addr)
printf("%s: Out of RQT memory\n",
device_get_nameunit(rdev->adap->dev));
mutex_lock(&rdev->stats.lock);
if (addr) {
rdev->stats.rqt.cur += roundup(size << 6, 1 << MIN_RQT_SHIFT);
if (rdev->stats.rqt.cur > rdev->stats.rqt.max)
rdev->stats.rqt.max = rdev->stats.rqt.cur;
} else
rdev->stats.rqt.fail++;
mutex_unlock(&rdev->stats.lock);
return (u32)addr;
}
void c4iw_rqtpool_free(struct c4iw_rdev *rdev, u32 addr, int size)
{
CTR3(KTR_IW_CXGBE, "%s addr 0x%x size %d", __func__, addr, size << 6);
mutex_lock(&rdev->stats.lock);
rdev->stats.rqt.cur -= roundup(size << 6, 1 << MIN_RQT_SHIFT);
mutex_unlock(&rdev->stats.lock);
gen_pool_free(rdev->rqt_pool, (unsigned long)addr, size << 6);
}
int c4iw_rqtpool_create(struct c4iw_rdev *rdev)
{
rdev->rqt_pool = gen_pool_create(rdev->adap->vres.rq.start,
MIN_RQT_SHIFT,
rdev->adap->vres.rq.size);
if (!rdev->rqt_pool)
return -ENOMEM;
return 0;
}
void c4iw_rqtpool_destroy(struct c4iw_rdev *rdev)
{
gen_pool_destroy(rdev->rqt_pool);
}
#endif

597
sys/dev/cxgbe/iw_cxgbe/t4.h Normal file
View File

@ -0,0 +1,597 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*
* $FreeBSD$
*/
#ifndef __T4_H__
#define __T4_H__
/*
* Fixme: Adding missing defines
*/
#define SGE_PF_KDOORBELL 0x0
#define QID_MASK 0xffff8000U
#define QID_SHIFT 15
#define QID(x) ((x) << QID_SHIFT)
#define DBPRIO 0x00004000U
#define PIDX_MASK 0x00003fffU
#define PIDX_SHIFT 0
#define PIDX(x) ((x) << PIDX_SHIFT)
#define SGE_PF_GTS 0x4
#define INGRESSQID_MASK 0xffff0000U
#define INGRESSQID_SHIFT 16
#define INGRESSQID(x) ((x) << INGRESSQID_SHIFT)
#define TIMERREG_MASK 0x0000e000U
#define TIMERREG_SHIFT 13
#define TIMERREG(x) ((x) << TIMERREG_SHIFT)
#define SEINTARM_MASK 0x00001000U
#define SEINTARM_SHIFT 12
#define SEINTARM(x) ((x) << SEINTARM_SHIFT)
#define CIDXINC_MASK 0x00000fffU
#define CIDXINC_SHIFT 0
#define CIDXINC(x) ((x) << CIDXINC_SHIFT)
#define T4_MAX_NUM_QP (1<<16)
#define T4_MAX_NUM_CQ (1<<15)
#define T4_MAX_NUM_PD (1<<15)
#define T4_EQ_STATUS_ENTRIES (L1_CACHE_BYTES > 64 ? 2 : 1)
#define T4_MAX_EQ_SIZE (65520 - T4_EQ_STATUS_ENTRIES)
#define T4_MAX_IQ_SIZE (65520 - 1)
#define T4_MAX_RQ_SIZE (8192 - T4_EQ_STATUS_ENTRIES)
#define T4_MAX_SQ_SIZE (T4_MAX_EQ_SIZE - 1)
#define T4_MAX_QP_DEPTH (T4_MAX_RQ_SIZE - 1)
#define T4_MAX_CQ_DEPTH (T4_MAX_IQ_SIZE - 1)
#define T4_MAX_NUM_STAG (1<<15)
#define T4_MAX_MR_SIZE (~0ULL - 1)
#define T4_PAGESIZE_MASK 0xffff000 /* 4KB-128MB */
#define T4_STAG_UNSET 0xffffffff
#define T4_FW_MAJ 0
#define T4_EQ_STATUS_ENTRIES (L1_CACHE_BYTES > 64 ? 2 : 1)
#define A_PCIE_MA_SYNC 0x30b4
struct t4_status_page {
__be32 rsvd1; /* flit 0 - hw owns */
__be16 rsvd2;
__be16 qid;
__be16 cidx;
__be16 pidx;
u8 qp_err; /* flit 1 - sw owns */
u8 db_off;
u8 pad;
u16 host_wq_pidx;
u16 host_cidx;
u16 host_pidx;
};
#define T4_EQ_ENTRY_SIZE 64
#define T4_SQ_NUM_SLOTS 5
#define T4_SQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_SQ_NUM_SLOTS)
#define T4_MAX_SEND_SGE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \
sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
#define T4_MAX_SEND_INLINE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \
sizeof(struct fw_ri_immd)))
#define T4_MAX_WRITE_INLINE ((T4_SQ_NUM_BYTES - \
sizeof(struct fw_ri_rdma_write_wr) - \
sizeof(struct fw_ri_immd)))
#define T4_MAX_WRITE_SGE ((T4_SQ_NUM_BYTES - \
sizeof(struct fw_ri_rdma_write_wr) - \
sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
#define T4_MAX_FR_IMMD ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_fr_nsmr_wr) - \
sizeof(struct fw_ri_immd)) & ~31UL)
#define T4_MAX_FR_DEPTH (T4_MAX_FR_IMMD / sizeof(u64))
#define T4_RQ_NUM_SLOTS 2
#define T4_RQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_RQ_NUM_SLOTS)
#define T4_MAX_RECV_SGE 4
union t4_wr {
struct fw_ri_res_wr res;
struct fw_ri_wr ri;
struct fw_ri_rdma_write_wr write;
struct fw_ri_send_wr send;
struct fw_ri_rdma_read_wr read;
struct fw_ri_bind_mw_wr bind;
struct fw_ri_fr_nsmr_wr fr;
struct fw_ri_inv_lstag_wr inv;
struct t4_status_page status;
__be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_SQ_NUM_SLOTS];
};
union t4_recv_wr {
struct fw_ri_recv_wr recv;
struct t4_status_page status;
__be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_RQ_NUM_SLOTS];
};
static inline void init_wr_hdr(union t4_wr *wqe, u16 wrid,
enum fw_wr_opcodes opcode, u8 flags, u8 len16)
{
wqe->send.opcode = (u8)opcode;
wqe->send.flags = flags;
wqe->send.wrid = wrid;
wqe->send.r1[0] = 0;
wqe->send.r1[1] = 0;
wqe->send.r1[2] = 0;
wqe->send.len16 = len16;
}
/* CQE/AE status codes */
#define T4_ERR_SUCCESS 0x0
#define T4_ERR_STAG 0x1 /* STAG invalid: either the */
/* STAG is offlimt, being 0, */
/* or STAG_key mismatch */
#define T4_ERR_PDID 0x2 /* PDID mismatch */
#define T4_ERR_QPID 0x3 /* QPID mismatch */
#define T4_ERR_ACCESS 0x4 /* Invalid access right */
#define T4_ERR_WRAP 0x5 /* Wrap error */
#define T4_ERR_BOUND 0x6 /* base and bounds voilation */
#define T4_ERR_INVALIDATE_SHARED_MR 0x7 /* attempt to invalidate a */
/* shared memory region */
#define T4_ERR_INVALIDATE_MR_WITH_MW_BOUND 0x8 /* attempt to invalidate a */
/* shared memory region */
#define T4_ERR_ECC 0x9 /* ECC error detected */
#define T4_ERR_ECC_PSTAG 0xA /* ECC error detected when */
/* reading PSTAG for a MW */
/* Invalidate */
#define T4_ERR_PBL_ADDR_BOUND 0xB /* pbl addr out of bounds: */
/* software error */
#define T4_ERR_SWFLUSH 0xC /* SW FLUSHED */
#define T4_ERR_CRC 0x10 /* CRC error */
#define T4_ERR_MARKER 0x11 /* Marker error */
#define T4_ERR_PDU_LEN_ERR 0x12 /* invalid PDU length */
#define T4_ERR_OUT_OF_RQE 0x13 /* out of RQE */
#define T4_ERR_DDP_VERSION 0x14 /* wrong DDP version */
#define T4_ERR_RDMA_VERSION 0x15 /* wrong RDMA version */
#define T4_ERR_OPCODE 0x16 /* invalid rdma opcode */
#define T4_ERR_DDP_QUEUE_NUM 0x17 /* invalid ddp queue number */
#define T4_ERR_MSN 0x18 /* MSN error */
#define T4_ERR_TBIT 0x19 /* tag bit not set correctly */
#define T4_ERR_MO 0x1A /* MO not 0 for TERMINATE */
/* or READ_REQ */
#define T4_ERR_MSN_GAP 0x1B
#define T4_ERR_MSN_RANGE 0x1C
#define T4_ERR_IRD_OVERFLOW 0x1D
#define T4_ERR_RQE_ADDR_BOUND 0x1E /* RQE addr out of bounds: */
/* software error */
#define T4_ERR_INTERNAL_ERR 0x1F /* internal error (opcode */
/* mismatch) */
/*
* CQE defs
*/
struct t4_cqe {
__be32 header;
__be32 len;
union {
struct {
__be32 stag;
__be32 msn;
} rcqe;
struct {
u32 nada1;
u16 nada2;
u16 cidx;
} scqe;
struct {
__be32 wrid_hi;
__be32 wrid_low;
} gen;
} u;
__be64 reserved;
__be64 bits_type_ts;
};
/* macros for flit 0 of the cqe */
#define S_CQE_QPID 12
#define M_CQE_QPID 0xFFFFF
#define G_CQE_QPID(x) ((((x) >> S_CQE_QPID)) & M_CQE_QPID)
#define V_CQE_QPID(x) ((x)<<S_CQE_QPID)
#define S_CQE_SWCQE 11
#define M_CQE_SWCQE 0x1
#define G_CQE_SWCQE(x) ((((x) >> S_CQE_SWCQE)) & M_CQE_SWCQE)
#define V_CQE_SWCQE(x) ((x)<<S_CQE_SWCQE)
#define S_CQE_STATUS 5
#define M_CQE_STATUS 0x1F
#define G_CQE_STATUS(x) ((((x) >> S_CQE_STATUS)) & M_CQE_STATUS)
#define V_CQE_STATUS(x) ((x)<<S_CQE_STATUS)
#define S_CQE_TYPE 4
#define M_CQE_TYPE 0x1
#define G_CQE_TYPE(x) ((((x) >> S_CQE_TYPE)) & M_CQE_TYPE)
#define V_CQE_TYPE(x) ((x)<<S_CQE_TYPE)
#define S_CQE_OPCODE 0
#define M_CQE_OPCODE 0xF
#define G_CQE_OPCODE(x) ((((x) >> S_CQE_OPCODE)) & M_CQE_OPCODE)
#define V_CQE_OPCODE(x) ((x)<<S_CQE_OPCODE)
#define SW_CQE(x) (G_CQE_SWCQE(be32_to_cpu((x)->header)))
#define CQE_QPID(x) (G_CQE_QPID(be32_to_cpu((x)->header)))
#define CQE_TYPE(x) (G_CQE_TYPE(be32_to_cpu((x)->header)))
#define SQ_TYPE(x) (CQE_TYPE((x)))
#define RQ_TYPE(x) (!CQE_TYPE((x)))
#define CQE_STATUS(x) (G_CQE_STATUS(be32_to_cpu((x)->header)))
#define CQE_OPCODE(x) (G_CQE_OPCODE(be32_to_cpu((x)->header)))
#define CQE_SEND_OPCODE(x)(\
(G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND) || \
(G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE) || \
(G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_INV) || \
(G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE_INV))
#define CQE_LEN(x) (be32_to_cpu((x)->len))
/* used for RQ completion processing */
#define CQE_WRID_STAG(x) (be32_to_cpu((x)->u.rcqe.stag))
#define CQE_WRID_MSN(x) (be32_to_cpu((x)->u.rcqe.msn))
/* used for SQ completion processing */
#define CQE_WRID_SQ_IDX(x) ((x)->u.scqe.cidx)
/* generic accessor macros */
#define CQE_WRID_HI(x) ((x)->u.gen.wrid_hi)
#define CQE_WRID_LOW(x) ((x)->u.gen.wrid_low)
/* macros for flit 3 of the cqe */
#define S_CQE_GENBIT 63
#define M_CQE_GENBIT 0x1
#define G_CQE_GENBIT(x) (((x) >> S_CQE_GENBIT) & M_CQE_GENBIT)
#define V_CQE_GENBIT(x) ((x)<<S_CQE_GENBIT)
#define S_CQE_OVFBIT 62
#define M_CQE_OVFBIT 0x1
#define G_CQE_OVFBIT(x) ((((x) >> S_CQE_OVFBIT)) & M_CQE_OVFBIT)
#define S_CQE_IQTYPE 60
#define M_CQE_IQTYPE 0x3
#define G_CQE_IQTYPE(x) ((((x) >> S_CQE_IQTYPE)) & M_CQE_IQTYPE)
#define M_CQE_TS 0x0fffffffffffffffULL
#define G_CQE_TS(x) ((x) & M_CQE_TS)
#define CQE_OVFBIT(x) ((unsigned)G_CQE_OVFBIT(be64_to_cpu((x)->bits_type_ts)))
#define CQE_GENBIT(x) ((unsigned)G_CQE_GENBIT(be64_to_cpu((x)->bits_type_ts)))
#define CQE_TS(x) (G_CQE_TS(be64_to_cpu((x)->bits_type_ts)))
struct t4_swsqe {
u64 wr_id;
struct t4_cqe cqe;
int read_len;
int opcode;
int complete;
int signaled;
u16 idx;
};
struct t4_sq {
union t4_wr *queue;
bus_addr_t dma_addr;
DECLARE_PCI_UNMAP_ADDR(mapping);
unsigned long phys_addr;
struct t4_swsqe *sw_sq;
struct t4_swsqe *oldest_read;
u64 udb;
size_t memsize;
u32 qid;
u16 in_use;
u16 size;
u16 cidx;
u16 pidx;
u16 wq_pidx;
u16 flags;
};
struct t4_swrqe {
u64 wr_id;
};
struct t4_rq {
union t4_recv_wr *queue;
bus_addr_t dma_addr;
DECLARE_PCI_UNMAP_ADDR(mapping);
struct t4_swrqe *sw_rq;
u64 udb;
size_t memsize;
u32 qid;
u32 msn;
u32 rqt_hwaddr;
u16 rqt_size;
u16 in_use;
u16 size;
u16 cidx;
u16 pidx;
u16 wq_pidx;
};
struct t4_wq {
struct t4_sq sq;
struct t4_rq rq;
void __iomem *db;
void __iomem *gts;
struct c4iw_rdev *rdev;
};
static inline int t4_rqes_posted(struct t4_wq *wq)
{
return wq->rq.in_use;
}
static inline int t4_rq_empty(struct t4_wq *wq)
{
return wq->rq.in_use == 0;
}
static inline int t4_rq_full(struct t4_wq *wq)
{
return wq->rq.in_use == (wq->rq.size - 1);
}
static inline u32 t4_rq_avail(struct t4_wq *wq)
{
return wq->rq.size - 1 - wq->rq.in_use;
}
static inline void t4_rq_produce(struct t4_wq *wq, u8 len16)
{
wq->rq.in_use++;
if (++wq->rq.pidx == wq->rq.size)
wq->rq.pidx = 0;
wq->rq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE);
if (wq->rq.wq_pidx >= wq->rq.size * T4_RQ_NUM_SLOTS)
wq->rq.wq_pidx %= wq->rq.size * T4_RQ_NUM_SLOTS;
}
static inline void t4_rq_consume(struct t4_wq *wq)
{
wq->rq.in_use--;
wq->rq.msn++;
if (++wq->rq.cidx == wq->rq.size)
wq->rq.cidx = 0;
}
static inline u16 t4_rq_host_wq_pidx(struct t4_wq *wq)
{
return wq->rq.queue[wq->rq.size].status.host_wq_pidx;
}
static inline u16 t4_rq_wq_size(struct t4_wq *wq)
{
return wq->rq.size * T4_RQ_NUM_SLOTS;
}
static inline int t4_sq_empty(struct t4_wq *wq)
{
return wq->sq.in_use == 0;
}
static inline int t4_sq_full(struct t4_wq *wq)
{
return wq->sq.in_use == (wq->sq.size - 1);
}
static inline u32 t4_sq_avail(struct t4_wq *wq)
{
return wq->sq.size - 1 - wq->sq.in_use;
}
static inline void t4_sq_produce(struct t4_wq *wq, u8 len16)
{
wq->sq.in_use++;
if (++wq->sq.pidx == wq->sq.size)
wq->sq.pidx = 0;
wq->sq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE);
if (wq->sq.wq_pidx >= wq->sq.size * T4_SQ_NUM_SLOTS)
wq->sq.wq_pidx %= wq->sq.size * T4_SQ_NUM_SLOTS;
}
static inline void t4_sq_consume(struct t4_wq *wq)
{
wq->sq.in_use--;
if (++wq->sq.cidx == wq->sq.size)
wq->sq.cidx = 0;
}
static inline u16 t4_sq_host_wq_pidx(struct t4_wq *wq)
{
return wq->sq.queue[wq->sq.size].status.host_wq_pidx;
}
static inline u16 t4_sq_wq_size(struct t4_wq *wq)
{
return wq->sq.size * T4_SQ_NUM_SLOTS;
}
static inline void t4_ring_sq_db(struct t4_wq *wq, u16 inc)
{
wmb();
writel(QID(wq->sq.qid) | PIDX(inc), wq->db);
}
static inline void t4_ring_rq_db(struct t4_wq *wq, u16 inc)
{
wmb();
writel(QID(wq->rq.qid) | PIDX(inc), wq->db);
}
static inline int t4_wq_in_error(struct t4_wq *wq)
{
return wq->rq.queue[wq->rq.size].status.qp_err;
}
static inline void t4_set_wq_in_error(struct t4_wq *wq)
{
wq->rq.queue[wq->rq.size].status.qp_err = 1;
}
static inline void t4_disable_wq_db(struct t4_wq *wq)
{
wq->rq.queue[wq->rq.size].status.db_off = 1;
}
static inline void t4_enable_wq_db(struct t4_wq *wq)
{
wq->rq.queue[wq->rq.size].status.db_off = 0;
}
static inline int t4_wq_db_enabled(struct t4_wq *wq)
{
return !wq->rq.queue[wq->rq.size].status.db_off;
}
struct t4_cq {
struct t4_cqe *queue;
bus_addr_t dma_addr;
DECLARE_PCI_UNMAP_ADDR(mapping);
struct t4_cqe *sw_queue;
void __iomem *gts;
struct c4iw_rdev *rdev;
u64 ugts;
size_t memsize;
__be64 bits_type_ts;
u32 cqid;
u16 size; /* including status page */
u16 cidx;
u16 sw_pidx;
u16 sw_cidx;
u16 sw_in_use;
u16 cidx_inc;
u8 gen;
u8 error;
};
static inline int t4_arm_cq(struct t4_cq *cq, int se)
{
u32 val;
while (cq->cidx_inc > CIDXINC_MASK) {
val = SEINTARM(0) | CIDXINC(CIDXINC_MASK) | TIMERREG(7) |
INGRESSQID(cq->cqid);
writel(val, cq->gts);
cq->cidx_inc -= CIDXINC_MASK;
}
val = SEINTARM(se) | CIDXINC(cq->cidx_inc) | TIMERREG(6) |
INGRESSQID(cq->cqid);
writel(val, cq->gts);
cq->cidx_inc = 0;
return 0;
}
static inline void t4_swcq_produce(struct t4_cq *cq)
{
cq->sw_in_use++;
if (++cq->sw_pidx == cq->size)
cq->sw_pidx = 0;
}
static inline void t4_swcq_consume(struct t4_cq *cq)
{
cq->sw_in_use--;
if (++cq->sw_cidx == cq->size)
cq->sw_cidx = 0;
}
static inline void t4_hwcq_consume(struct t4_cq *cq)
{
cq->bits_type_ts = cq->queue[cq->cidx].bits_type_ts;
if (++cq->cidx_inc == (cq->size >> 4)) {
u32 val;
val = SEINTARM(0) | CIDXINC(cq->cidx_inc) | TIMERREG(7) |
INGRESSQID(cq->cqid);
writel(val, cq->gts);
cq->cidx_inc = 0;
}
if (++cq->cidx == cq->size) {
cq->cidx = 0;
cq->gen ^= 1;
}
}
static inline int t4_valid_cqe(struct t4_cq *cq, struct t4_cqe *cqe)
{
return (CQE_GENBIT(cqe) == cq->gen);
}
static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
{
int ret;
u16 prev_cidx;
if (cq->cidx == 0)
prev_cidx = cq->size - 1;
else
prev_cidx = cq->cidx - 1;
if (cq->queue[prev_cidx].bits_type_ts != cq->bits_type_ts) {
ret = -EOVERFLOW;
cq->error = 1;
printk(KERN_ERR MOD "cq overflow cqid %u\n", cq->cqid);
} else if (t4_valid_cqe(cq, &cq->queue[cq->cidx])) {
*cqe = &cq->queue[cq->cidx];
ret = 0;
} else
ret = -ENODATA;
return ret;
}
static inline struct t4_cqe *t4_next_sw_cqe(struct t4_cq *cq)
{
if (cq->sw_in_use)
return &cq->sw_queue[cq->sw_cidx];
return NULL;
}
static inline int t4_next_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
{
int ret = 0;
if (cq->error)
ret = -ENODATA;
else if (cq->sw_in_use)
*cqe = &cq->sw_queue[cq->sw_cidx];
else
ret = t4_next_hw_cqe(cq, cqe);
return ret;
}
static inline int t4_cq_in_error(struct t4_cq *cq)
{
return ((struct t4_status_page *)&cq->queue[cq->size])->qp_err;
}
static inline void t4_set_cq_in_error(struct t4_cq *cq)
{
((struct t4_status_page *)&cq->queue[cq->size])->qp_err = 1;
}
#endif

View File

@ -0,0 +1,70 @@
/*
* Copyright (c) 2009-2013 Chelsio, Inc. 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.
*
* $FreeBSD$
*/
#ifndef __C4IW_USER_H__
#define __C4IW_USER_H__
#define C4IW_UVERBS_ABI_VERSION 2
/*
* Make sure that all structs defined in this file remain laid out so
* that they pack the same way on 32-bit and 64-bit architectures (to
* avoid incompatibility between 32-bit userspace and 64-bit kernels).
* In particular do not use pointer types -- pass pointers in __u64
* instead.
*/
struct c4iw_create_cq_resp {
__u64 key;
__u64 gts_key;
__u64 memsize;
__u32 cqid;
__u32 size;
__u32 qid_mask;
};
struct c4iw_create_qp_resp {
__u64 ma_sync_key;
__u64 sq_key;
__u64 rq_key;
__u64 sq_db_gts_key;
__u64 rq_db_gts_key;
__u64 sq_memsize;
__u64 rq_memsize;
__u32 sqid;
__u32 rqid;
__u32 sq_size;
__u32 rq_size;
__u32 qid_mask;
__u32 flags;
};
#endif

View File

@ -6,9 +6,16 @@ SUBDIR = if_cxgbe
SUBDIR+= t4_firmware
SUBDIR+= t5_firmware
SUBDIR+= ${_tom}
SUBDIR+= ${_iw_cxgbe}
.if ${MACHINE_CPUARCH} == "amd64" || ${MACHINE_CPUARCH} == "i386"
.if ${MACHINE_CPUARCH} == "amd64"
_tom= tom
_iw_cxgbe= iw_cxgbe
.endif
.if ${MACHINE_CPUARCH} == "i386"
_tom= tom
.endif
.include <bsd.subdir.mk>

View File

@ -0,0 +1,27 @@
# $FreeBSD$
.include <bsd.own.mk>
CXGBE = ${.CURDIR}/../../../dev/cxgbe
.PATH: ${CXGBE}/iw_cxgbe
KMOD= iw_cxgbe
SRCS= device.c cm.c provider.c mem.c cq.c qp.c resource.c ev.c id_table.c
SRCS+= bus_if.h device_if.h opt_sched.h pci_if.h pcib_if.h opt_ktr.h
SRCS+= opt_inet.h opt_ofed.h vnode_if.h
CFLAGS+= -I${CXGBE} -I${.CURDIR}/../../../ofed/include -DLINUX_TYPES_DEFINED
.if !defined(KERNBUILDDIR)
.if ${MK_INET_SUPPORT} != "no"
opt_inet.h:
@echo "#define INET 1" > ${.TARGET}
@echo "#define TCP_OFFLOAD 1" >> ${.TARGET}
.endif
.if ${MK_INET6_SUPPORT} != "no"
opt_inet6.h:
@echo "#define INET6 1" > ${.TARGET}
.endif
.endif
.include <bsd.kmod.mk>