*/
#include <linux/interrupt.h>
-#include <linux/pci.h> /* for Tavor hack below */
#include <linux/slab.h>
#include <asm/bitops.h>
return;
if (wc->opcode == IB_WC_FAST_REG_MR)
- frmr->r.frmr.state = FRMR_IS_VALID;
+ frmr->r.frmr.fr_state = FRMR_IS_VALID;
else if (wc->opcode == IB_WC_LOCAL_INV)
- frmr->r.frmr.state = FRMR_IS_INVALID;
+ frmr->r.frmr.fr_state = FRMR_IS_INVALID;
}
static int
rpcrdma_recvcq_poll(cq, ep);
}
+static void
+rpcrdma_flush_cqs(struct rpcrdma_ep *ep)
+{
+ rpcrdma_recvcq_upcall(ep->rep_attr.recv_cq, ep);
+ rpcrdma_sendcq_upcall(ep->rep_attr.send_cq, ep);
+}
+
#ifdef RPC_DEBUG
static const char * const conn[] = {
"address resolved",
/* Else will do memory reg/dereg for each chunk */
ia->ri_memreg_strategy = memreg;
+ rwlock_init(&ia->ri_qplock);
return 0;
out2:
rdma_destroy_id(ia->ri_id);
int
rpcrdma_ep_connect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
{
- struct rdma_cm_id *id;
+ struct rdma_cm_id *id, *old;
int rc = 0;
int retry_count = 0;
if (rc && rc != -ENOTCONN)
dprintk("RPC: %s: rpcrdma_ep_disconnect"
" status %i\n", __func__, rc);
-
- rpcrdma_clean_cq(ep->rep_attr.recv_cq);
- rpcrdma_clean_cq(ep->rep_attr.send_cq);
+ rpcrdma_flush_cqs(ep);
xprt = container_of(ia, struct rpcrdma_xprt, rx_ia);
id = rpcrdma_create_id(xprt, ia,
rc = -ENETUNREACH;
goto out;
}
- rdma_destroy_qp(ia->ri_id);
- rdma_destroy_id(ia->ri_id);
+
+ write_lock(&ia->ri_qplock);
+ old = ia->ri_id;
ia->ri_id = id;
+ write_unlock(&ia->ri_qplock);
+
+ rdma_destroy_qp(old);
+ rdma_destroy_id(old);
} else {
dprintk("RPC: %s: connecting...\n", __func__);
rc = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr);
}
}
-/* XXX Tavor device performs badly with 2K MTU! */
-if (strnicmp(ia->ri_id->device->dma_device->bus->name, "pci", 3) == 0) {
- struct pci_dev *pcid = to_pci_dev(ia->ri_id->device->dma_device);
- if (pcid->device == PCI_DEVICE_ID_MELLANOX_TAVOR &&
- (pcid->vendor == PCI_VENDOR_ID_MELLANOX ||
- pcid->vendor == PCI_VENDOR_ID_TOPSPIN)) {
- struct ib_qp_attr attr = {
- .path_mtu = IB_MTU_1024
- };
- rc = ib_modify_qp(ia->ri_id->qp, &attr, IB_QP_PATH_MTU);
- }
-}
-
ep->rep_connected = 0;
rc = rdma_connect(ia->ri_id, &ep->rep_remote_cma);
{
int rc;
- rpcrdma_clean_cq(ep->rep_attr.recv_cq);
- rpcrdma_clean_cq(ep->rep_attr.send_cq);
+ rpcrdma_flush_cqs(ep);
rc = rdma_disconnect(ia->ri_id);
if (!rc) {
/* returns without wait if not connected */
p += cdata->padding;
INIT_LIST_HEAD(&buf->rb_mws);
+ INIT_LIST_HEAD(&buf->rb_all);
r = (struct rpcrdma_mw *)p;
switch (ia->ri_memreg_strategy) {
case RPCRDMA_FRMR:
ib_dereg_mr(r->r.frmr.fr_mr);
goto out;
}
+ list_add(&r->mw_all, &buf->rb_all);
list_add(&r->mw_list, &buf->rb_mws);
++r;
}
" failed %i\n", __func__, rc);
goto out;
}
+ list_add(&r->mw_all, &buf->rb_all);
list_add(&r->mw_list, &buf->rb_mws);
++r;
}
while (!list_empty(&buf->rb_mws)) {
r = list_entry(buf->rb_mws.next,
struct rpcrdma_mw, mw_list);
+ list_del(&r->mw_all);
list_del(&r->mw_list);
switch (ia->ri_memreg_strategy) {
case RPCRDMA_FRMR:
kfree(buf->rb_pool);
}
+/* "*mw" can be NULL when rpcrdma_buffer_get_mrs() fails, leaving
+ * some req segments uninitialized.
+ */
+static void
+rpcrdma_buffer_put_mr(struct rpcrdma_mw **mw, struct rpcrdma_buffer *buf)
+{
+ if (*mw) {
+ list_add_tail(&(*mw)->mw_list, &buf->rb_mws);
+ *mw = NULL;
+ }
+}
+
+/* Cycle mw's back in reverse order, and "spin" them.
+ * This delays and scrambles reuse as much as possible.
+ */
+static void
+rpcrdma_buffer_put_mrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf)
+{
+ struct rpcrdma_mr_seg *seg = req->rl_segments;
+ struct rpcrdma_mr_seg *seg1 = seg;
+ int i;
+
+ for (i = 1, seg++; i < RPCRDMA_MAX_SEGS; seg++, i++)
+ rpcrdma_buffer_put_mr(&seg->mr_chunk.rl_mw, buf);
+ rpcrdma_buffer_put_mr(&seg1->mr_chunk.rl_mw, buf);
+}
+
+static void
+rpcrdma_buffer_put_sendbuf(struct rpcrdma_req *req, struct rpcrdma_buffer *buf)
+{
+ buf->rb_send_bufs[--buf->rb_send_index] = req;
+ req->rl_niovs = 0;
+ if (req->rl_reply) {
+ buf->rb_recv_bufs[--buf->rb_recv_index] = req->rl_reply;
+ req->rl_reply->rr_func = NULL;
+ req->rl_reply = NULL;
+ }
+}
+
+static struct rpcrdma_req *
+rpcrdma_buffer_get_mrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf)
+{
+ struct rpcrdma_mw *r;
+ int i;
+
+ i = RPCRDMA_MAX_SEGS - 1;
+ while (!list_empty(&buf->rb_mws)) {
+ r = list_entry(buf->rb_mws.next,
+ struct rpcrdma_mw, mw_list);
+ list_del(&r->mw_list);
+ req->rl_segments[i].mr_chunk.rl_mw = r;
+ if (unlikely(i-- == 0))
+ return req; /* Success */
+ }
+
+ /* Not enough entries on rb_mws for this req */
+ rpcrdma_buffer_put_sendbuf(req, buf);
+ rpcrdma_buffer_put_mrs(req, buf);
+ return NULL;
+}
+
/*
* Get a set of request/reply buffers.
*
struct rpcrdma_req *
rpcrdma_buffer_get(struct rpcrdma_buffer *buffers)
{
+ struct rpcrdma_ia *ia = rdmab_to_ia(buffers);
struct rpcrdma_req *req;
unsigned long flags;
- int i;
- struct rpcrdma_mw *r;
spin_lock_irqsave(&buffers->rb_lock, flags);
if (buffers->rb_send_index == buffers->rb_max_requests) {
buffers->rb_recv_bufs[buffers->rb_recv_index++] = NULL;
}
buffers->rb_send_bufs[buffers->rb_send_index++] = NULL;
- if (!list_empty(&buffers->rb_mws)) {
- i = RPCRDMA_MAX_SEGS - 1;
- do {
- r = list_entry(buffers->rb_mws.next,
- struct rpcrdma_mw, mw_list);
- list_del(&r->mw_list);
- req->rl_segments[i].mr_chunk.rl_mw = r;
- } while (--i >= 0);
+ switch (ia->ri_memreg_strategy) {
+ case RPCRDMA_FRMR:
+ case RPCRDMA_MTHCAFMR:
+ req = rpcrdma_buffer_get_mrs(req, buffers);
+ break;
+ default:
+ break;
}
spin_unlock_irqrestore(&buffers->rb_lock, flags);
return req;
{
struct rpcrdma_buffer *buffers = req->rl_buffer;
struct rpcrdma_ia *ia = rdmab_to_ia(buffers);
- int i;
unsigned long flags;
- BUG_ON(req->rl_nchunks != 0);
spin_lock_irqsave(&buffers->rb_lock, flags);
- buffers->rb_send_bufs[--buffers->rb_send_index] = req;
- req->rl_niovs = 0;
- if (req->rl_reply) {
- buffers->rb_recv_bufs[--buffers->rb_recv_index] = req->rl_reply;
- req->rl_reply->rr_func = NULL;
- req->rl_reply = NULL;
- }
+ rpcrdma_buffer_put_sendbuf(req, buffers);
switch (ia->ri_memreg_strategy) {
case RPCRDMA_FRMR:
case RPCRDMA_MTHCAFMR:
- /*
- * Cycle mw's back in reverse order, and "spin" them.
- * This delays and scrambles reuse as much as possible.
- */
- i = 1;
- do {
- struct rpcrdma_mw **mw;
- mw = &req->rl_segments[i].mr_chunk.rl_mw;
- list_add_tail(&(*mw)->mw_list, &buffers->rb_mws);
- *mw = NULL;
- } while (++i < RPCRDMA_MAX_SEGS);
- list_add_tail(&req->rl_segments[0].mr_chunk.rl_mw->mw_list,
- &buffers->rb_mws);
- req->rl_segments[0].mr_chunk.rl_mw = NULL;
+ rpcrdma_buffer_put_mrs(req, buffers);
break;
default:
break;
*/
iov->addr = ib_dma_map_single(ia->ri_id->device,
va, len, DMA_BIDIRECTIONAL);
+ if (ib_dma_mapping_error(ia->ri_id->device, iov->addr))
+ return -ENOMEM;
+
iov->length = len;
if (ia->ri_have_dma_lkey) {
struct rpcrdma_xprt *r_xprt)
{
struct rpcrdma_mr_seg *seg1 = seg;
+ struct rpcrdma_mw *mw = seg1->mr_chunk.rl_mw;
+ struct rpcrdma_frmr *frmr = &mw->r.frmr;
+ struct ib_mr *mr = frmr->fr_mr;
struct ib_send_wr invalidate_wr, frmr_wr, *bad_wr, *post_wr;
u8 key;
rpcrdma_map_one(ia, seg, writing);
pa = seg->mr_dma;
for (seg_len = seg->mr_len; seg_len > 0; seg_len -= PAGE_SIZE) {
- seg1->mr_chunk.rl_mw->r.frmr.fr_pgl->
- page_list[page_no++] = pa;
+ frmr->fr_pgl->page_list[page_no++] = pa;
pa += PAGE_SIZE;
}
len += seg->mr_len;
break;
}
dprintk("RPC: %s: Using frmr %p to map %d segments\n",
- __func__, seg1->mr_chunk.rl_mw, i);
+ __func__, mw, i);
- if (unlikely(seg1->mr_chunk.rl_mw->r.frmr.state == FRMR_IS_VALID)) {
+ if (unlikely(frmr->fr_state == FRMR_IS_VALID)) {
dprintk("RPC: %s: frmr %x left valid, posting invalidate.\n",
- __func__,
- seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey);
+ __func__, mr->rkey);
/* Invalidate before using. */
memset(&invalidate_wr, 0, sizeof invalidate_wr);
- invalidate_wr.wr_id = (unsigned long)(void *)seg1->mr_chunk.rl_mw;
+ invalidate_wr.wr_id = (unsigned long)(void *)mw;
invalidate_wr.next = &frmr_wr;
invalidate_wr.opcode = IB_WR_LOCAL_INV;
invalidate_wr.send_flags = IB_SEND_SIGNALED;
- invalidate_wr.ex.invalidate_rkey =
- seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey;
+ invalidate_wr.ex.invalidate_rkey = mr->rkey;
DECR_CQCOUNT(&r_xprt->rx_ep);
post_wr = &invalidate_wr;
} else
post_wr = &frmr_wr;
- /* Bump the key */
- key = (u8)(seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey & 0x000000FF);
- ib_update_fast_reg_key(seg1->mr_chunk.rl_mw->r.frmr.fr_mr, ++key);
-
/* Prepare FRMR WR */
memset(&frmr_wr, 0, sizeof frmr_wr);
- frmr_wr.wr_id = (unsigned long)(void *)seg1->mr_chunk.rl_mw;
+ frmr_wr.wr_id = (unsigned long)(void *)mw;
frmr_wr.opcode = IB_WR_FAST_REG_MR;
frmr_wr.send_flags = IB_SEND_SIGNALED;
frmr_wr.wr.fast_reg.iova_start = seg1->mr_dma;
- frmr_wr.wr.fast_reg.page_list = seg1->mr_chunk.rl_mw->r.frmr.fr_pgl;
+ frmr_wr.wr.fast_reg.page_list = frmr->fr_pgl;
frmr_wr.wr.fast_reg.page_list_len = page_no;
frmr_wr.wr.fast_reg.page_shift = PAGE_SHIFT;
frmr_wr.wr.fast_reg.length = page_no << PAGE_SHIFT;
- BUG_ON(frmr_wr.wr.fast_reg.length < len);
+ if (frmr_wr.wr.fast_reg.length < len) {
+ rc = -EIO;
+ goto out_err;
+ }
+
+ /* Bump the key */
+ key = (u8)(mr->rkey & 0x000000FF);
+ ib_update_fast_reg_key(mr, ++key);
+
frmr_wr.wr.fast_reg.access_flags = (writing ?
IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE :
IB_ACCESS_REMOTE_READ);
- frmr_wr.wr.fast_reg.rkey = seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey;
+ frmr_wr.wr.fast_reg.rkey = mr->rkey;
DECR_CQCOUNT(&r_xprt->rx_ep);
rc = ib_post_send(ia->ri_id->qp, post_wr, &bad_wr);
if (rc) {
dprintk("RPC: %s: failed ib_post_send for register,"
" status %i\n", __func__, rc);
- while (i--)
- rpcrdma_unmap_one(ia, --seg);
+ ib_update_fast_reg_key(mr, --key);
+ goto out_err;
} else {
- seg1->mr_rkey = seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey;
+ seg1->mr_rkey = mr->rkey;
seg1->mr_base = seg1->mr_dma + pageoff;
seg1->mr_nsegs = i;
seg1->mr_len = len;
}
*nsegs = i;
+ return 0;
+out_err:
+ while (i--)
+ rpcrdma_unmap_one(ia, --seg);
return rc;
}
struct ib_send_wr invalidate_wr, *bad_wr;
int rc;
- while (seg1->mr_nsegs--)
- rpcrdma_unmap_one(ia, seg++);
-
memset(&invalidate_wr, 0, sizeof invalidate_wr);
invalidate_wr.wr_id = (unsigned long)(void *)seg1->mr_chunk.rl_mw;
invalidate_wr.opcode = IB_WR_LOCAL_INV;
invalidate_wr.ex.invalidate_rkey = seg1->mr_chunk.rl_mw->r.frmr.fr_mr->rkey;
DECR_CQCOUNT(&r_xprt->rx_ep);
+ read_lock(&ia->ri_qplock);
+ while (seg1->mr_nsegs--)
+ rpcrdma_unmap_one(ia, seg++);
rc = ib_post_send(ia->ri_id->qp, &invalidate_wr, &bad_wr);
+ read_unlock(&ia->ri_qplock);
if (rc)
dprintk("RPC: %s: failed ib_post_send for invalidate,"
" status %i\n", __func__, rc);
list_add(&seg1->mr_chunk.rl_mw->r.fmr->list, &l);
rc = ib_unmap_fmr(&l);
+ read_lock(&ia->ri_qplock);
while (seg1->mr_nsegs--)
rpcrdma_unmap_one(ia, seg++);
+ read_unlock(&ia->ri_qplock);
if (rc)
dprintk("RPC: %s: failed ib_unmap_fmr,"
" status %i\n", __func__, rc);
#if RPCRDMA_PERSISTENT_REGISTRATION
case RPCRDMA_ALLPHYSICAL:
- BUG_ON(nsegs != 1);
+ read_lock(&ia->ri_qplock);
rpcrdma_unmap_one(ia, seg);
- rc = 0;
+ read_unlock(&ia->ri_qplock);
break;
#endif
rc);
return rc;
}
+
+/* Physical mapping means one Read/Write list entry per-page.
+ * All list entries must fit within an inline buffer
+ *
+ * NB: The server must return a Write list for NFS READ,
+ * which has the same constraint. Factor in the inline
+ * rsize as well.
+ */
+static size_t
+rpcrdma_physical_max_payload(struct rpcrdma_xprt *r_xprt)
+{
+ struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
+ unsigned int inline_size, pages;
+
+ inline_size = min_t(unsigned int,
+ cdata->inline_wsize, cdata->inline_rsize);
+ inline_size -= RPCRDMA_HDRLEN_MIN;
+ pages = inline_size / sizeof(struct rpcrdma_segment);
+ return pages << PAGE_SHIFT;
+}
+
+static size_t
+rpcrdma_mr_max_payload(struct rpcrdma_xprt *r_xprt)
+{
+ return RPCRDMA_MAX_DATA_SEGS << PAGE_SHIFT;
+}
+
+size_t
+rpcrdma_max_payload(struct rpcrdma_xprt *r_xprt)
+{
+ size_t result;
+
+ switch (r_xprt->rx_ia.ri_memreg_strategy) {
+ case RPCRDMA_ALLPHYSICAL:
+ result = rpcrdma_physical_max_payload(r_xprt);
+ break;
+ default:
+ result = rpcrdma_mr_max_payload(r_xprt);
+ }
+ return result;
+}
projects / karo-tx-linux.git / blobdiff