/* Send a backwards direction RPC call.
*
* Caller holds the connection's mutex and has already marshaled
* the RPC/RDMA request.
*
* This is similar to svc_rdma_reply, but takes an rpc_rqst
* instead, does not support chunks, and avoids blocking memory
* allocation.
*
* XXX: There is still an opportunity to block in svc_rdma_send()
* if there are no SQ entries to post the Send. This may occur if
* the adapter has a small maximum SQ depth.
*/
static int svc_rdma_bc_sendto(struct svcxprt_rdma *rdma,
struct rpc_rqst *rqst)
{
struct xdr_buf *sndbuf = &rqst->rq_snd_buf;
struct svc_rdma_op_ctxt *ctxt;
struct svc_rdma_req_map *vec;
struct ib_send_wr send_wr;
int ret;
vec = svc_rdma_get_req_map(rdma);
ret = svc_rdma_map_xdr(rdma, sndbuf, vec, false);
if (ret)
goto out_err;
ret = svc_rdma_repost_recv(rdma, GFP_NOIO);
if (ret)
goto out_err;
ctxt = svc_rdma_get_context(rdma);
ctxt->pages[0] = virt_to_page(rqst->rq_buffer);
ctxt->count = 1;
ctxt->direction = DMA_TO_DEVICE;
ctxt->sge[0].lkey = rdma->sc_pd->local_dma_lkey;
ctxt->sge[0].length = sndbuf->len;
ctxt->sge[0].addr =
ib_dma_map_page(rdma->sc_cm_id->device, ctxt->pages[0], 0,
sndbuf->len, DMA_TO_DEVICE);
if (ib_dma_mapping_error(rdma->sc_cm_id->device, ctxt->sge[0].addr)) {
ret = -EIO;
goto out_unmap;
}
svc_rdma_count_mappings(rdma, ctxt);
memset(&send_wr, 0, sizeof(send_wr));
ctxt->cqe.done = svc_rdma_wc_send;
send_wr.wr_cqe = &ctxt->cqe;
send_wr.sg_list = ctxt->sge;
send_wr.num_sge = 1;
send_wr.opcode = IB_WR_SEND;
send_wr.send_flags = IB_SEND_SIGNALED;
ret = svc_rdma_send(rdma, &send_wr);
if (ret) {
ret = -EIO;
goto out_unmap;
}
out_err:
svc_rdma_put_req_map(rdma, vec);
dprintk("svcrdma: %s returns %d\n", __func__, ret);
return ret;
out_unmap:
svc_rdma_unmap_dma(ctxt);
svc_rdma_put_context(ctxt, 1);
goto out_err;
}
/*
* Use RDMA_READ to read data from the advertised client buffer into the
* XDR stream starting at rq_arg.head[0].iov_base.
* Each chunk in the array
* contains the following fields:
* discrim - '1', This isn't used for data placement
* position - The xdr stream offset (the same for every chunk)
* handle - RMR for client memory region
* length - data transfer length
* offset - 64 bit tagged offset in remote memory region
*
* On our side, we need to read into a pagelist. The first page immediately
* follows the RPC header.
*
* This function returns:
* 0 - No error and no read-list found.
*
* 1 - Successful read-list processing. The data is not yet in
* the pagelist and therefore the RPC request must be deferred. The
* I/O completion will enqueue the transport again and
* svc_rdma_recvfrom will complete the request.
*
* <0 - Error processing/posting read-list.
*
* NOTE: The ctxt must not be touched after the last WR has been posted
* because the I/O completion processing may occur on another
* processor and free / modify the context. Ne touche pas!
*/
static int rdma_read_xdr(struct svcxprt_rdma *xprt,
struct rpcrdma_msg *rmsgp,
struct svc_rqst *rqstp,
struct svc_rdma_op_ctxt *hdr_ctxt)
{
struct ib_send_wr read_wr;
struct ib_send_wr inv_wr;
int err = 0;
int ch_no;
int ch_count;
int byte_count;
int sge_count;
u64 sgl_offset;
struct rpcrdma_read_chunk *ch;
struct svc_rdma_op_ctxt *ctxt = NULL;
struct svc_rdma_req_map *rpl_map;
struct svc_rdma_req_map *chl_map;
/* If no read list is present, return 0 */
ch = svc_rdma_get_read_chunk(rmsgp);
if (!ch)
return 0;
svc_rdma_rcl_chunk_counts(ch, &ch_count, &byte_count);
if (ch_count > RPCSVC_MAXPAGES)
return -EINVAL;
/* Allocate temporary reply and chunk maps */
rpl_map = svc_rdma_get_req_map();
chl_map = svc_rdma_get_req_map();
if (!xprt->sc_frmr_pg_list_len)
sge_count = map_read_chunks(xprt, rqstp, hdr_ctxt, rmsgp,
rpl_map, chl_map, ch_count,
byte_count);
else
sge_count = fast_reg_read_chunks(xprt, rqstp, hdr_ctxt, rmsgp,
rpl_map, chl_map, ch_count,
byte_count);
if (sge_count < 0) {
err = -EIO;
goto out;
}
sgl_offset = 0;
ch_no = 0;
for (ch = (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
ch->rc_discrim != 0; ch++, ch_no++) {
next_sge:
ctxt = svc_rdma_get_context(xprt);
ctxt->direction = DMA_FROM_DEVICE;
ctxt->frmr = hdr_ctxt->frmr;
ctxt->read_hdr = NULL;
clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
clear_bit(RDMACTXT_F_FAST_UNREG, &ctxt->flags);
/* Prepare READ WR */
memset(&read_wr, 0, sizeof read_wr);
read_wr.wr_id = (unsigned long)ctxt;
read_wr.opcode = IB_WR_RDMA_READ;
ctxt->wr_op = read_wr.opcode;
read_wr.send_flags = IB_SEND_SIGNALED;
read_wr.wr.rdma.rkey = ch->rc_target.rs_handle;
read_wr.wr.rdma.remote_addr =
get_unaligned(&(ch->rc_target.rs_offset)) +
sgl_offset;
read_wr.sg_list = ctxt->sge;
read_wr.num_sge =
rdma_read_max_sge(xprt, chl_map->ch[ch_no].count);
err = rdma_set_ctxt_sge(xprt, ctxt, hdr_ctxt->frmr,
&rpl_map->sge[chl_map->ch[ch_no].start],
&sgl_offset,
read_wr.num_sge);
if (err) {
svc_rdma_unmap_dma(ctxt);
svc_rdma_put_context(ctxt, 0);
goto out;
}
if (((ch+1)->rc_discrim == 0) &&
(read_wr.num_sge == chl_map->ch[ch_no].count)) {
/*
* Mark the last RDMA_READ with a bit to
* indicate all RPC data has been fetched from
* the client and the RPC needs to be enqueued.
*/
set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
if (hdr_ctxt->frmr) {
set_bit(RDMACTXT_F_FAST_UNREG, &ctxt->flags);
/*
* Invalidate the local MR used to map the data
* sink.
*/
if (xprt->sc_dev_caps &
SVCRDMA_DEVCAP_READ_W_INV) {
read_wr.opcode =
IB_WR_RDMA_READ_WITH_INV;
ctxt->wr_op = read_wr.opcode;
read_wr.ex.invalidate_rkey =
ctxt->frmr->mr->lkey;
} else {
/* Prepare INVALIDATE WR */
memset(&inv_wr, 0, sizeof inv_wr);
inv_wr.opcode = IB_WR_LOCAL_INV;
inv_wr.send_flags = IB_SEND_SIGNALED;
inv_wr.ex.invalidate_rkey =
hdr_ctxt->frmr->mr->lkey;
read_wr.next = &inv_wr;
}
}
ctxt->read_hdr = hdr_ctxt;
}
/* Post the read */
err = svc_rdma_send(xprt, &read_wr);
if (err) {
printk(KERN_ERR "svcrdma: Error %d posting RDMA_READ\n",
err);
set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
svc_rdma_unmap_dma(ctxt);
svc_rdma_put_context(ctxt, 0);
goto out;
}
atomic_inc(&rdma_stat_read);
if (read_wr.num_sge < chl_map->ch[ch_no].count) {
chl_map->ch[ch_no].count -= read_wr.num_sge;
chl_map->ch[ch_no].start += read_wr.num_sge;
goto next_sge;
}
sgl_offset = 0;
err = 1;
}
out:
svc_rdma_put_req_map(rpl_map);
svc_rdma_put_req_map(chl_map);
/* Detach arg pages. svc_recv will replenish them */
for (ch_no = 0; &rqstp->rq_pages[ch_no] < rqstp->rq_respages; ch_no++)
rqstp->rq_pages[ch_no] = NULL;
/*
* Detach res pages. svc_release must see a resused count of
* zero or it will attempt to put them.
*/
while (rqstp->rq_resused)
rqstp->rq_respages[--rqstp->rq_resused] = NULL;
return err;
}
int svc_rdma_sendto(struct svc_rqst *rqstp)
{
struct svc_xprt *xprt = rqstp->rq_xprt;
struct svcxprt_rdma *rdma =
container_of(xprt, struct svcxprt_rdma, sc_xprt);
struct rpcrdma_msg *rdma_argp;
struct rpcrdma_msg *rdma_resp;
struct rpcrdma_write_array *reply_ary;
enum rpcrdma_proc reply_type;
int ret;
int inline_bytes;
struct page *res_page;
struct svc_rdma_op_ctxt *ctxt;
struct svc_rdma_req_map *vec;
dprintk("svcrdma: sending response for rqstp=%p\n", rqstp);
/* Get the RDMA request header. */
rdma_argp = xdr_start(&rqstp->rq_arg);
/* Build an req vec for the XDR */
ctxt = svc_rdma_get_context(rdma);
ctxt->direction = DMA_TO_DEVICE;
vec = svc_rdma_get_req_map();
ret = map_xdr(rdma, &rqstp->rq_res, vec);
if (ret)
goto err0;
inline_bytes = rqstp->rq_res.len;
/* Create the RDMA response header */
res_page = svc_rdma_get_page();
rdma_resp = page_address(res_page);
reply_ary = svc_rdma_get_reply_array(rdma_argp);
if (reply_ary)
reply_type = RDMA_NOMSG;
else
reply_type = RDMA_MSG;
svc_rdma_xdr_encode_reply_header(rdma, rdma_argp,
rdma_resp, reply_type);
/* Send any write-chunk data and build resp write-list */
ret = send_write_chunks(rdma, rdma_argp, rdma_resp,
rqstp, vec);
if (ret < 0) {
printk(KERN_ERR "svcrdma: failed to send write chunks, rc=%d\n",
ret);
goto err1;
}
inline_bytes -= ret;
/* Send any reply-list data and update resp reply-list */
ret = send_reply_chunks(rdma, rdma_argp, rdma_resp,
rqstp, vec);
if (ret < 0) {
printk(KERN_ERR "svcrdma: failed to send reply chunks, rc=%d\n",
ret);
goto err1;
}
inline_bytes -= ret;
ret = send_reply(rdma, rqstp, res_page, rdma_resp, ctxt, vec,
inline_bytes);
svc_rdma_put_req_map(vec);
dprintk("svcrdma: send_reply returns %d\n", ret);
return ret;
err1:
put_page(res_page);
err0:
svc_rdma_put_req_map(vec);
svc_rdma_put_context(ctxt, 0);
return ret;
}
static int rdma_read_xdr(struct svcxprt_rdma *xprt,
struct rpcrdma_msg *rmsgp,
struct svc_rqst *rqstp,
struct svc_rdma_op_ctxt *hdr_ctxt)
{
struct ib_send_wr read_wr;
struct ib_send_wr inv_wr;
int err = 0;
int ch_no;
int ch_count;
int byte_count;
int sge_count;
u64 sgl_offset;
struct rpcrdma_read_chunk *ch;
struct svc_rdma_op_ctxt *ctxt = NULL;
struct svc_rdma_req_map *rpl_map;
struct svc_rdma_req_map *chl_map;
/* */
ch = svc_rdma_get_read_chunk(rmsgp);
if (!ch)
return 0;
svc_rdma_rcl_chunk_counts(ch, &ch_count, &byte_count);
if (ch_count > RPCSVC_MAXPAGES)
return -EINVAL;
/* */
rpl_map = svc_rdma_get_req_map();
chl_map = svc_rdma_get_req_map();
if (!xprt->sc_frmr_pg_list_len)
sge_count = map_read_chunks(xprt, rqstp, hdr_ctxt, rmsgp,
rpl_map, chl_map, ch_count,
byte_count);
else
sge_count = fast_reg_read_chunks(xprt, rqstp, hdr_ctxt, rmsgp,
rpl_map, chl_map, ch_count,
byte_count);
if (sge_count < 0) {
err = -EIO;
goto out;
}
sgl_offset = 0;
ch_no = 0;
for (ch = (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
ch->rc_discrim != 0; ch++, ch_no++) {
u64 rs_offset;
next_sge:
ctxt = svc_rdma_get_context(xprt);
ctxt->direction = DMA_FROM_DEVICE;
ctxt->frmr = hdr_ctxt->frmr;
ctxt->read_hdr = NULL;
clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
clear_bit(RDMACTXT_F_FAST_UNREG, &ctxt->flags);
/* */
memset(&read_wr, 0, sizeof read_wr);
read_wr.wr_id = (unsigned long)ctxt;
read_wr.opcode = IB_WR_RDMA_READ;
ctxt->wr_op = read_wr.opcode;
read_wr.send_flags = IB_SEND_SIGNALED;
read_wr.wr.rdma.rkey = ntohl(ch->rc_target.rs_handle);
xdr_decode_hyper((__be32 *)&ch->rc_target.rs_offset,
&rs_offset);
read_wr.wr.rdma.remote_addr = rs_offset + sgl_offset;
read_wr.sg_list = ctxt->sge;
read_wr.num_sge =
rdma_read_max_sge(xprt, chl_map->ch[ch_no].count);
err = rdma_set_ctxt_sge(xprt, ctxt, hdr_ctxt->frmr,
&rpl_map->sge[chl_map->ch[ch_no].start],
&sgl_offset,
read_wr.num_sge);
if (err) {
svc_rdma_unmap_dma(ctxt);
svc_rdma_put_context(ctxt, 0);
goto out;
}
if (((ch+1)->rc_discrim == 0) &&
(read_wr.num_sge == chl_map->ch[ch_no].count)) {
/*
*/
set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
if (hdr_ctxt->frmr) {
set_bit(RDMACTXT_F_FAST_UNREG, &ctxt->flags);
/*
*/
if (xprt->sc_dev_caps &
SVCRDMA_DEVCAP_READ_W_INV) {
read_wr.opcode =
IB_WR_RDMA_READ_WITH_INV;
ctxt->wr_op = read_wr.opcode;
read_wr.ex.invalidate_rkey =
ctxt->frmr->mr->lkey;
} else {
/* */
memset(&inv_wr, 0, sizeof inv_wr);
inv_wr.opcode = IB_WR_LOCAL_INV;
inv_wr.send_flags = IB_SEND_SIGNALED;
inv_wr.ex.invalidate_rkey =
hdr_ctxt->frmr->mr->lkey;
read_wr.next = &inv_wr;
}
}
ctxt->read_hdr = hdr_ctxt;
}
/* */
err = svc_rdma_send(xprt, &read_wr);
if (err) {
printk(KERN_ERR "svcrdma: Error %d posting RDMA_READ\n",
err);
set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
svc_rdma_unmap_dma(ctxt);
svc_rdma_put_context(ctxt, 0);
goto out;
}
atomic_inc(&rdma_stat_read);
if (read_wr.num_sge < chl_map->ch[ch_no].count) {
chl_map->ch[ch_no].count -= read_wr.num_sge;
chl_map->ch[ch_no].start += read_wr.num_sge;
goto next_sge;
}
sgl_offset = 0;
err = 1;
}
out:
svc_rdma_put_req_map(rpl_map);
svc_rdma_put_req_map(chl_map);
/* */
for (ch_no = 0; &rqstp->rq_pages[ch_no] < rqstp->rq_respages; ch_no++)
rqstp->rq_pages[ch_no] = NULL;
/*
*/
while (rqstp->rq_resused)
rqstp->rq_respages[--rqstp->rq_resused] = NULL;
return err;
}
