/**
* rvt_post_receive - post a receive on a QP
* @ibqp: the QP to post the receive on
* @wr: the WR to post
* @bad_wr: the first bad WR is put here
*
* This may be called from interrupt context.
*
* Return: 0 on success otherwise errno
*/
int rvt_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr)
{
struct rvt_qp *qp = ibqp_to_rvtqp(ibqp);
struct rvt_rwq *wq = qp->r_rq.wq;
unsigned long flags;
int qp_err_flush = (ib_rvt_state_ops[qp->state] & RVT_FLUSH_RECV) &&
!qp->ibqp.srq;
/* Check that state is OK to post receive. */
if (!(ib_rvt_state_ops[qp->state] & RVT_POST_RECV_OK) || !wq) {
*bad_wr = wr;
return -EINVAL;
}
for (; wr; wr = wr->next) {
struct rvt_rwqe *wqe;
u32 next;
int i;
if ((unsigned)wr->num_sge > qp->r_rq.max_sge) {
*bad_wr = wr;
return -EINVAL;
}
spin_lock_irqsave(&qp->r_rq.lock, flags);
next = wq->head + 1;
if (next >= qp->r_rq.size)
next = 0;
if (next == wq->tail) {
spin_unlock_irqrestore(&qp->r_rq.lock, flags);
*bad_wr = wr;
return -ENOMEM;
}
if (unlikely(qp_err_flush)) {
struct ib_wc wc;
memset(&wc, 0, sizeof(wc));
wc.qp = &qp->ibqp;
wc.opcode = IB_WC_RECV;
wc.wr_id = wr->wr_id;
wc.status = IB_WC_WR_FLUSH_ERR;
rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
} else {
wqe = rvt_get_rwqe_ptr(&qp->r_rq, wq->head);
wqe->wr_id = wr->wr_id;
wqe->num_sge = wr->num_sge;
for (i = 0; i < wr->num_sge; i++)
wqe->sg_list[i] = wr->sg_list[i];
/*
* Make sure queue entry is written
* before the head index.
*/
smp_wmb();
wq->head = next;
}
spin_unlock_irqrestore(&qp->r_rq.lock, flags);
}
return 0;
}
/**
* rvt_post_srq_receive - post a receive on a shared receive queue
* @ibsrq: the SRQ to post the receive on
* @wr: the list of work requests to post
* @bad_wr: A pointer to the first WR to cause a problem is put here
*
* This may be called from interrupt context.
*
* Return: 0 on success else errno
*/
int rvt_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr)
{
struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
struct rvt_rwq *wq;
unsigned long flags;
for (; wr; wr = wr->next) {
struct rvt_rwqe *wqe;
u32 next;
int i;
if ((unsigned)wr->num_sge > srq->rq.max_sge) {
*bad_wr = wr;
return -EINVAL;
}
spin_lock_irqsave(&srq->rq.lock, flags);
wq = srq->rq.wq;
next = wq->head + 1;
if (next >= srq->rq.size)
next = 0;
if (next == wq->tail) {
spin_unlock_irqrestore(&srq->rq.lock, flags);
*bad_wr = wr;
return -ENOMEM;
}
wqe = rvt_get_rwqe_ptr(&srq->rq, wq->head);
wqe->wr_id = wr->wr_id;
wqe->num_sge = wr->num_sge;
for (i = 0; i < wr->num_sge; i++)
wqe->sg_list[i] = wr->sg_list[i];
/* Make sure queue entry is written before the head index. */
smp_wmb();
wq->head = next;
spin_unlock_irqrestore(&srq->rq.lock, flags);
}
return 0;
}
/**
* qib_get_rwqe - copy the next RWQE into the QP's RWQE
* @qp: the QP
* @wr_id_only: update qp->r_wr_id only, not qp->r_sge
*
* Return -1 if there is a local error, 0 if no RWQE is available,
* otherwise return 1.
*
* Can be called from interrupt level.
*/
int qib_get_rwqe(struct rvt_qp *qp, int wr_id_only)
{
unsigned long flags;
struct rvt_rq *rq;
struct rvt_rwq *wq;
struct rvt_srq *srq;
struct rvt_rwqe *wqe;
void (*handler)(struct ib_event *, void *);
u32 tail;
int ret;
if (qp->ibqp.srq) {
srq = ibsrq_to_rvtsrq(qp->ibqp.srq);
handler = srq->ibsrq.event_handler;
rq = &srq->rq;
} else {
srq = NULL;
handler = NULL;
rq = &qp->r_rq;
}
spin_lock_irqsave(&rq->lock, flags);
if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK)) {
ret = 0;
goto unlock;
}
wq = rq->wq;
tail = wq->tail;
/* Validate tail before using it since it is user writable. */
if (tail >= rq->size)
tail = 0;
if (unlikely(tail == wq->head)) {
ret = 0;
goto unlock;
}
/* Make sure entry is read after head index is read. */
smp_rmb();
wqe = rvt_get_rwqe_ptr(rq, tail);
/*
* Even though we update the tail index in memory, the verbs
* consumer is not supposed to post more entries until a
* completion is generated.
*/
if (++tail >= rq->size)
tail = 0;
wq->tail = tail;
if (!wr_id_only && !qib_init_sge(qp, wqe)) {
ret = -1;
goto unlock;
}
qp->r_wr_id = wqe->wr_id;
ret = 1;
set_bit(RVT_R_WRID_VALID, &qp->r_aflags);
if (handler) {
u32 n;
/*
* Validate head pointer value and compute
* the number of remaining WQEs.
*/
n = wq->head;
if (n >= rq->size)
n = 0;
if (n < tail)
n += rq->size - tail;
else
n -= tail;
if (n < srq->limit) {
struct ib_event ev;
srq->limit = 0;
spin_unlock_irqrestore(&rq->lock, flags);
ev.device = qp->ibqp.device;
ev.element.srq = qp->ibqp.srq;
ev.event = IB_EVENT_SRQ_LIMIT_REACHED;
handler(&ev, srq->ibsrq.srq_context);
goto bail;
}
}
unlock:
spin_unlock_irqrestore(&rq->lock, flags);
bail:
return ret;
}
/**
* rvt_error_qp - put a QP into the error state
* @qp: the QP to put into the error state
* @err: the receive completion error to signal if a RWQE is active
*
* Flushes both send and receive work queues.
*
* Return: true if last WQE event should be generated.
* The QP r_lock and s_lock should be held and interrupts disabled.
* If we are already in error state, just return.
*/
int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err)
{
struct ib_wc wc;
int ret = 0;
struct rvt_dev_info *rdi = ib_to_rvt(qp->ibqp.device);
if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
goto bail;
qp->state = IB_QPS_ERR;
if (qp->s_flags & (RVT_S_TIMER | RVT_S_WAIT_RNR)) {
qp->s_flags &= ~(RVT_S_TIMER | RVT_S_WAIT_RNR);
del_timer(&qp->s_timer);
}
if (qp->s_flags & RVT_S_ANY_WAIT_SEND)
qp->s_flags &= ~RVT_S_ANY_WAIT_SEND;
rdi->driver_f.notify_error_qp(qp);
/* Schedule the sending tasklet to drain the send work queue. */
if (ACCESS_ONCE(qp->s_last) != qp->s_head)
rdi->driver_f.schedule_send(qp);
rvt_clear_mr_refs(qp, 0);
memset(&wc, 0, sizeof(wc));
wc.qp = &qp->ibqp;
wc.opcode = IB_WC_RECV;
if (test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags)) {
wc.wr_id = qp->r_wr_id;
wc.status = err;
rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
}
wc.status = IB_WC_WR_FLUSH_ERR;
if (qp->r_rq.wq) {
struct rvt_rwq *wq;
u32 head;
u32 tail;
spin_lock(&qp->r_rq.lock);
/* sanity check pointers before trusting them */
wq = qp->r_rq.wq;
head = wq->head;
if (head >= qp->r_rq.size)
head = 0;
tail = wq->tail;
if (tail >= qp->r_rq.size)
tail = 0;
while (tail != head) {
wc.wr_id = rvt_get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
if (++tail >= qp->r_rq.size)
tail = 0;
rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc, 1);
}
wq->tail = tail;
spin_unlock(&qp->r_rq.lock);
} else if (qp->ibqp.event_handler) {
ret = 1;
}
bail:
return ret;
}
/**
* rvt_modify_srq - modify a shared receive queue
* @ibsrq: the SRQ to modify
* @attr: the new attributes of the SRQ
* @attr_mask: indicates which attributes to modify
* @udata: user data for libibverbs.so
*
* Return: 0 on success
*/
int rvt_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
enum ib_srq_attr_mask attr_mask,
struct ib_udata *udata)
{
struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
struct rvt_dev_info *dev = ib_to_rvt(ibsrq->device);
struct rvt_rwq *wq;
int ret = 0;
if (attr_mask & IB_SRQ_MAX_WR) {
struct rvt_rwq *owq;
struct rvt_rwqe *p;
u32 sz, size, n, head, tail;
/* Check that the requested sizes are below the limits. */
if ((attr->max_wr > dev->dparms.props.max_srq_wr) ||
((attr_mask & IB_SRQ_LIMIT) ?
attr->srq_limit : srq->limit) > attr->max_wr)
return -EINVAL;
sz = sizeof(struct rvt_rwqe) +
srq->rq.max_sge * sizeof(struct ib_sge);
size = attr->max_wr + 1;
wq = udata ?
vmalloc_user(sizeof(struct rvt_rwq) + size * sz) :
vzalloc_node(sizeof(struct rvt_rwq) + size * sz,
dev->dparms.node);
if (!wq)
return -ENOMEM;
/* Check that we can write the offset to mmap. */
if (udata && udata->inlen >= sizeof(__u64)) {
__u64 offset_addr;
__u64 offset = 0;
ret = ib_copy_from_udata(&offset_addr, udata,
sizeof(offset_addr));
if (ret)
goto bail_free;
udata->outbuf = (void __user *)
(unsigned long)offset_addr;
ret = ib_copy_to_udata(udata, &offset,
sizeof(offset));
if (ret)
goto bail_free;
}
spin_lock_irq(&srq->rq.lock);
/*
* validate head and tail pointer values and compute
* the number of remaining WQEs.
*/
owq = srq->rq.wq;
head = owq->head;
tail = owq->tail;
if (head >= srq->rq.size || tail >= srq->rq.size) {
ret = -EINVAL;
goto bail_unlock;
}
n = head;
if (n < tail)
n += srq->rq.size - tail;
else
n -= tail;
if (size <= n) {
ret = -EINVAL;
goto bail_unlock;
}
n = 0;
p = wq->wq;
while (tail != head) {
struct rvt_rwqe *wqe;
int i;
wqe = rvt_get_rwqe_ptr(&srq->rq, tail);
p->wr_id = wqe->wr_id;
p->num_sge = wqe->num_sge;
for (i = 0; i < wqe->num_sge; i++)
p->sg_list[i] = wqe->sg_list[i];
n++;
p = (struct rvt_rwqe *)((char *)p + sz);
if (++tail >= srq->rq.size)
tail = 0;
}
srq->rq.wq = wq;
srq->rq.size = size;
wq->head = n;
wq->tail = 0;
if (attr_mask & IB_SRQ_LIMIT)
srq->limit = attr->srq_limit;
spin_unlock_irq(&srq->rq.lock);
vfree(owq);
if (srq->ip) {
struct rvt_mmap_info *ip = srq->ip;
struct rvt_dev_info *dev = ib_to_rvt(srq->ibsrq.device);
u32 s = sizeof(struct rvt_rwq) + size * sz;
rvt_update_mmap_info(dev, ip, s, wq);
/*
* Return the offset to mmap.
* See rvt_mmap() for details.
*/
if (udata && udata->inlen >= sizeof(__u64)) {
ret = ib_copy_to_udata(udata, &ip->offset,
sizeof(ip->offset));
if (ret)
return ret;
}
/*
* Put user mapping info onto the pending list
* unless it already is on the list.
*/
spin_lock_irq(&dev->pending_lock);
if (list_empty(&ip->pending_mmaps))
list_add(&ip->pending_mmaps,
&dev->pending_mmaps);
spin_unlock_irq(&dev->pending_lock);
}
} else if (attr_mask & IB_SRQ_LIMIT) {
spin_lock_irq(&srq->rq.lock);
if (attr->srq_limit >= srq->rq.size)
ret = -EINVAL;
else
srq->limit = attr->srq_limit;
spin_unlock_irq(&srq->rq.lock);
}
return ret;
bail_unlock:
spin_unlock_irq(&srq->rq.lock);
bail_free:
vfree(wq);
return ret;
}