#include <asm/page.h>
#include <linux/io.h>
#include <linux/wait.h>
#include <rdma/ib_addr.h>
#include <rdma/ib_smi.h>
#include <rdma/ib_user_verbs.h>
#include <rdma/uverbs_ioctl.h>
#include "pvrdma.h"
int pvrdma_req_notify_cq(struct ib_cq *ibcq,
enum ib_cq_notify_flags notify_flags)
{
struct pvrdma_dev *dev = to_vdev(ibcq->device);
struct pvrdma_cq *cq = to_vcq(ibcq);
u32 val = cq->cq_handle;
unsigned long flags;
int has_data = 0;
val |= (notify_flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
PVRDMA_UAR_CQ_ARM_SOL : PVRDMA_UAR_CQ_ARM;
spin_lock_irqsave(&cq->cq_lock, flags);
pvrdma_write_uar_cq(dev, val);
if (notify_flags & IB_CQ_REPORT_MISSED_EVENTS) {
unsigned int head;
has_data = pvrdma_idx_ring_has_data(&cq->ring_state->rx,
cq->ibcq.cqe, &head);
if (unlikely(has_data == PVRDMA_INVALID_IDX))
dev_err(&dev->pdev->dev, "CQ ring state invalid\n");
}
spin_unlock_irqrestore(&cq->cq_lock, flags);
return has_data;
}
int pvrdma_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
struct ib_udata *udata)
{
struct ib_device *ibdev = ibcq->device;
int entries = attr->cqe;
struct pvrdma_dev *dev = to_vdev(ibdev);
struct pvrdma_cq *cq = to_vcq(ibcq);
int ret;
int npages;
unsigned long flags;
union pvrdma_cmd_req req;
union pvrdma_cmd_resp rsp;
struct pvrdma_cmd_create_cq *cmd = &req.create_cq;
struct pvrdma_cmd_create_cq_resp *resp = &rsp.create_cq_resp;
struct pvrdma_create_cq_resp cq_resp = {};
struct pvrdma_create_cq ucmd;
struct pvrdma_ucontext *context = rdma_udata_to_drv_context(
udata, struct pvrdma_ucontext, ibucontext);
BUILD_BUG_ON(sizeof(struct pvrdma_cqe) != 64);
if (attr->flags)
return -EOPNOTSUPP;
entries = roundup_pow_of_two(entries);
if (entries < 1 || entries > dev->dsr->caps.max_cqe)
return -EINVAL;
if (!atomic_add_unless(&dev->num_cqs, 1, dev->dsr->caps.max_cq))
return -ENOMEM;
cq->ibcq.cqe = entries;
cq->is_kernel = !udata;
if (!cq->is_kernel) {
if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) {
ret = -EFAULT;
goto err_cq;
}
cq->umem = ib_umem_get(ibdev, ucmd.buf_addr, ucmd.buf_size,
IB_ACCESS_LOCAL_WRITE);
if (IS_ERR(cq->umem)) {
ret = PTR_ERR(cq->umem);
goto err_cq;
}
npages = ib_umem_num_dma_blocks(cq->umem, PAGE_SIZE);
} else {
npages = 1 + (entries * sizeof(struct pvrdma_cqe) +
PAGE_SIZE - 1) / PAGE_SIZE;
cq->offset = PAGE_SIZE;
}
if (npages < 0 || npages > PVRDMA_PAGE_DIR_MAX_PAGES) {
dev_warn(&dev->pdev->dev,
"overflow pages in completion queue\n");
ret = -EINVAL;
goto err_umem;
}
ret = pvrdma_page_dir_init(dev, &cq->pdir, npages, cq->is_kernel);
if (ret) {
dev_warn(&dev->pdev->dev,
"could not allocate page directory\n");
goto err_umem;
}
if (cq->is_kernel)
cq->ring_state = cq->pdir.pages[0];
else
pvrdma_page_dir_insert_umem(&cq->pdir, cq->umem, 0);
refcount_set(&cq->refcnt, 1);
init_completion(&cq->free);
spin_lock_init(&cq->cq_lock);
memset(cmd, 0, sizeof(*cmd));
cmd->hdr.cmd = PVRDMA_CMD_CREATE_CQ;
cmd->nchunks = npages;
cmd->ctx_handle = context ? context->ctx_handle : 0;
cmd->cqe = entries;
cmd->pdir_dma = cq->pdir.dir_dma;
ret = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_CREATE_CQ_RESP);
if (ret < 0) {
dev_warn(&dev->pdev->dev,
"could not create completion queue, error: %d\n", ret);
goto err_page_dir;
}
cq->ibcq.cqe = resp->cqe;
cq->cq_handle = resp->cq_handle;
cq_resp.cqn = resp->cq_handle;
spin_lock_irqsave(&dev->cq_tbl_lock, flags);
dev->cq_tbl[cq->cq_handle % dev->dsr->caps.max_cq] = cq;
spin_unlock_irqrestore(&dev->cq_tbl_lock, flags);
if (!cq->is_kernel) {
cq->uar = &context->uar;
if (ib_copy_to_udata(udata, &cq_resp, sizeof(cq_resp))) {
dev_warn(&dev->pdev->dev,
"failed to copy back udata\n");
pvrdma_destroy_cq(&cq->ibcq, udata);
return -EINVAL;
}
}
return 0;
err_page_dir:
pvrdma_page_dir_cleanup(dev, &cq->pdir);
err_umem:
ib_umem_release(cq->umem);
err_cq:
atomic_dec(&dev->num_cqs);
return ret;
}
static void pvrdma_free_cq(struct pvrdma_dev *dev, struct pvrdma_cq *cq)
{
if (refcount_dec_and_test(&cq->refcnt))
complete(&cq->free);
wait_for_completion(&cq->free);
ib_umem_release(cq->umem);
pvrdma_page_dir_cleanup(dev, &cq->pdir);
}
int pvrdma_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
{
struct pvrdma_cq *vcq = to_vcq(cq);
union pvrdma_cmd_req req;
struct pvrdma_cmd_destroy_cq *cmd = &req.destroy_cq;
struct pvrdma_dev *dev = to_vdev(cq->device);
unsigned long flags;
int ret;
memset(cmd, 0, sizeof(*cmd));
cmd->hdr.cmd = PVRDMA_CMD_DESTROY_CQ;
cmd->cq_handle = vcq->cq_handle;
ret = pvrdma_cmd_post(dev, &req, NULL, 0);
if (ret < 0)
dev_warn(&dev->pdev->dev,
"could not destroy completion queue, error: %d\n",
ret);
spin_lock_irqsave(&dev->cq_tbl_lock, flags);
dev->cq_tbl[vcq->cq_handle] = NULL;
spin_unlock_irqrestore(&dev->cq_tbl_lock, flags);
pvrdma_free_cq(dev, vcq);
atomic_dec(&dev->num_cqs);
return 0;
}
static inline struct pvrdma_cqe *get_cqe(struct pvrdma_cq *cq, int i)
{
return (struct pvrdma_cqe *)pvrdma_page_dir_get_ptr(
&cq->pdir,
cq->offset +
sizeof(struct pvrdma_cqe) * i);
}
void _pvrdma_flush_cqe(struct pvrdma_qp *qp, struct pvrdma_cq *cq)
{
unsigned int head;
int has_data;
if (!cq->is_kernel)
return;
has_data = pvrdma_idx_ring_has_data(&cq->ring_state->rx,
cq->ibcq.cqe, &head);
if (unlikely(has_data > 0)) {
int items;
int curr;
int tail = pvrdma_idx(&cq->ring_state->rx.prod_tail,
cq->ibcq.cqe);
struct pvrdma_cqe *cqe;
struct pvrdma_cqe *curr_cqe;
items = (tail > head) ? (tail - head) :
(cq->ibcq.cqe - head + tail);
curr = --tail;
while (items-- > 0) {
if (curr < 0)
curr = cq->ibcq.cqe - 1;
if (tail < 0)
tail = cq->ibcq.cqe - 1;
curr_cqe = get_cqe(cq, curr);
if ((curr_cqe->qp & 0xFFFF) != qp->qp_handle) {
if (curr != tail) {
cqe = get_cqe(cq, tail);
*cqe = *curr_cqe;
}
tail--;
} else {
pvrdma_idx_ring_inc(
&cq->ring_state->rx.cons_head,
cq->ibcq.cqe);
}
curr--;
}
}
}
static int pvrdma_poll_one(struct pvrdma_cq *cq, struct pvrdma_qp **cur_qp,
struct ib_wc *wc)
{
struct pvrdma_dev *dev = to_vdev(cq->ibcq.device);
int has_data;
unsigned int head;
bool tried = false;
struct pvrdma_cqe *cqe;
retry:
has_data = pvrdma_idx_ring_has_data(&cq->ring_state->rx,
cq->ibcq.cqe, &head);
if (has_data == 0) {
if (tried)
return -EAGAIN;
pvrdma_write_uar_cq(dev, cq->cq_handle | PVRDMA_UAR_CQ_POLL);
tried = true;
goto retry;
} else if (has_data == PVRDMA_INVALID_IDX) {
dev_err(&dev->pdev->dev, "CQ ring state invalid\n");
return -EAGAIN;
}
cqe = get_cqe(cq, head);
rmb();
if (dev->qp_tbl[cqe->qp & 0xffff])
*cur_qp = (struct pvrdma_qp *)dev->qp_tbl[cqe->qp & 0xffff];
else
return -EAGAIN;
wc->opcode = pvrdma_wc_opcode_to_ib(cqe->opcode);
wc->status = pvrdma_wc_status_to_ib(cqe->status);
wc->wr_id = cqe->wr_id;
wc->qp = &(*cur_qp)->ibqp;
wc->byte_len = cqe->byte_len;
wc->ex.imm_data = cqe->imm_data;
wc->src_qp = cqe->src_qp;
wc->wc_flags = pvrdma_wc_flags_to_ib(cqe->wc_flags);
wc->pkey_index = cqe->pkey_index;
wc->slid = cqe->slid;
wc->sl = cqe->sl;
wc->dlid_path_bits = cqe->dlid_path_bits;
wc->port_num = cqe->port_num;
wc->vendor_err = cqe->vendor_err;
wc->network_hdr_type = pvrdma_network_type_to_ib(cqe->network_hdr_type);
pvrdma_idx_ring_inc(&cq->ring_state->rx.cons_head, cq->ibcq.cqe);
return 0;
}
int pvrdma_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
{
struct pvrdma_cq *cq = to_vcq(ibcq);
struct pvrdma_qp *cur_qp = NULL;
unsigned long flags;
int npolled;
if (num_entries < 1 || wc == NULL)
return 0;
spin_lock_irqsave(&cq->cq_lock, flags);
for (npolled = 0; npolled < num_entries; ++npolled) {
if (pvrdma_poll_one(cq, &cur_qp, wc + npolled))
break;
}
spin_unlock_irqrestore(&cq->cq_lock, flags);
return npolled;
}