Skip to content

Commit

Permalink
svcrdma: Simplify RDMA_READ deferral buffer management
Browse files Browse the repository at this point in the history
An NFS_WRITE requires a set of RDMA_READ requests to fetch the write
data from the client. There are two principal pieces of data that
need to be tracked: the list of pages that comprise the completed RPC
and the SGE of dma mapped pages to refer to this list of pages. Previously
this whole bit was managed as a linked list of contexts with the
context containing the page list buried in this list. This patch
simplifies this processing by not keeping a linked list, but rather only
a pionter from the last submitted RDMA_READ's context to the context
that maps the set of pages that describe the RPC.  This significantly
simplifies this code path. SGE contexts are cleaned up inline in the DTO
path instead of at read completion time.

Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
  • Loading branch information
Tom Tucker committed May 19, 2008
1 parent 10a38c3 commit 02e7452
Show file tree
Hide file tree
Showing 3 changed files with 16 additions and 48 deletions.
1 change: 1 addition & 0 deletions include/linux/sunrpc/svc_rdma.h
Original file line number Diff line number Diff line change
Expand Up @@ -71,6 +71,7 @@ extern atomic_t rdma_stat_sq_prod;
* completes.
*/
struct svc_rdma_op_ctxt {
struct svc_rdma_op_ctxt *read_hdr;
struct svc_rdma_op_ctxt *next;
struct xdr_buf arg;
struct list_head dto_q;
Expand Down
58 changes: 11 additions & 47 deletions net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
Original file line number Diff line number Diff line change
Expand Up @@ -289,7 +289,6 @@ static int rdma_read_xdr(struct svcxprt_rdma *xprt,
u64 sgl_offset;
struct rpcrdma_read_chunk *ch;
struct svc_rdma_op_ctxt *ctxt = NULL;
struct svc_rdma_op_ctxt *head;
struct svc_rdma_op_ctxt *tmp_sge_ctxt;
struct svc_rdma_op_ctxt *tmp_ch_ctxt;
struct chunk_sge *ch_sge_ary;
Expand All @@ -310,20 +309,13 @@ static int rdma_read_xdr(struct svcxprt_rdma *xprt,
sge_count = rdma_rcl_to_sge(xprt, rqstp, hdr_ctxt, rmsgp,
sge, ch_sge_ary,
ch_count, byte_count);
head = svc_rdma_get_context(xprt);
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:
if (!ctxt)
ctxt = head;
else {
ctxt->next = svc_rdma_get_context(xprt);
ctxt = ctxt->next;
}
ctxt->next = NULL;
ctxt = svc_rdma_get_context(xprt);
ctxt->direction = DMA_FROM_DEVICE;
clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);

Expand Down Expand Up @@ -351,20 +343,15 @@ static int rdma_read_xdr(struct svcxprt_rdma *xprt,
* the client and the RPC needs to be enqueued.
*/
set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
ctxt->next = hdr_ctxt;
hdr_ctxt->next = head;
ctxt->read_hdr = hdr_ctxt;
}
/* Post the read */
err = svc_rdma_send(xprt, &read_wr);
if (err) {
printk(KERN_ERR "svcrdma: Error posting send = %d\n",
printk(KERN_ERR "svcrdma: Error %d posting RDMA_READ\n",
err);
/*
* Break the circular list so free knows when
* to stop if the error happened to occur on
* the last read
*/
ctxt->next = NULL;
set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
svc_rdma_put_context(ctxt, 0);
goto out;
}
atomic_inc(&rdma_stat_read);
Expand All @@ -375,7 +362,7 @@ static int rdma_read_xdr(struct svcxprt_rdma *xprt,
goto next_sge;
}
sgl_offset = 0;
err = 0;
err = 1;
}

out:
Expand All @@ -393,25 +380,12 @@ static int rdma_read_xdr(struct svcxprt_rdma *xprt,
while (rqstp->rq_resused)
rqstp->rq_respages[--rqstp->rq_resused] = NULL;

if (err) {
printk(KERN_ERR "svcrdma : RDMA_READ error = %d\n", err);
set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
/* Free the linked list of read contexts */
while (head != NULL) {
ctxt = head->next;
svc_rdma_put_context(head, 1);
head = ctxt;
}
return err;
}

return 1;
return err;
}

static int rdma_read_complete(struct svc_rqst *rqstp,
struct svc_rdma_op_ctxt *data)
struct svc_rdma_op_ctxt *head)
{
struct svc_rdma_op_ctxt *head = data->next;
int page_no;
int ret;

Expand All @@ -437,22 +411,12 @@ static int rdma_read_complete(struct svc_rqst *rqstp,
rqstp->rq_arg.len = head->arg.len;
rqstp->rq_arg.buflen = head->arg.buflen;

/* Free the context */
svc_rdma_put_context(head, 0);

/* XXX: What should this be? */
rqstp->rq_prot = IPPROTO_MAX;

/*
* Free the contexts we used to build the RDMA_READ. We have
* to be careful here because the context list uses the same
* next pointer used to chain the contexts associated with the
* RDMA_READ
*/
data->next = NULL; /* terminate circular list */
do {
data = head->next;
svc_rdma_put_context(head, 0);
head = data;
} while (head != NULL);

ret = rqstp->rq_arg.head[0].iov_len
+ rqstp->rq_arg.page_len
+ rqstp->rq_arg.tail[0].iov_len;
Expand Down
5 changes: 4 additions & 1 deletion net/sunrpc/xprtrdma/svc_rdma_transport.c
Original file line number Diff line number Diff line change
Expand Up @@ -352,13 +352,16 @@ static void sq_cq_reap(struct svcxprt_rdma *xprt)

case IB_WR_RDMA_READ:
if (test_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags)) {
struct svc_rdma_op_ctxt *read_hdr = ctxt->read_hdr;
BUG_ON(!read_hdr);
set_bit(XPT_DATA, &xprt->sc_xprt.xpt_flags);
spin_lock_bh(&xprt->sc_read_complete_lock);
list_add_tail(&ctxt->dto_q,
list_add_tail(&read_hdr->dto_q,
&xprt->sc_read_complete_q);
spin_unlock_bh(&xprt->sc_read_complete_lock);
svc_xprt_enqueue(&xprt->sc_xprt);
}
svc_rdma_put_context(ctxt, 0);
break;

default:
Expand Down

0 comments on commit 02e7452

Please sign in to comment.