xref: /openbmc/linux/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c (revision 5ef12cb4a3a78ffb331c03a795a15eea4ae35155)
1 /*
2  * Copyright (c) 2016, 2017 Oracle. All rights reserved.
3  * Copyright (c) 2014 Open Grid Computing, Inc. All rights reserved.
4  * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
5  *
6  * This software is available to you under a choice of one of two
7  * licenses.  You may choose to be licensed under the terms of the GNU
8  * General Public License (GPL) Version 2, available from the file
9  * COPYING in the main directory of this source tree, or the BSD-type
10  * license below:
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  *
16  *      Redistributions of source code must retain the above copyright
17  *      notice, this list of conditions and the following disclaimer.
18  *
19  *      Redistributions in binary form must reproduce the above
20  *      copyright notice, this list of conditions and the following
21  *      disclaimer in the documentation and/or other materials provided
22  *      with the distribution.
23  *
24  *      Neither the name of the Network Appliance, Inc. nor the names of
25  *      its contributors may be used to endorse or promote products
26  *      derived from this software without specific prior written
27  *      permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
30  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
31  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
32  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
33  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
34  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
35  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
36  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
37  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
38  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
39  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
40  *
41  * Author: Tom Tucker <tom@opengridcomputing.com>
42  */
43 
44 /* Operation
45  *
46  * The main entry point is svc_rdma_recvfrom. This is called from
47  * svc_recv when the transport indicates there is incoming data to
48  * be read. "Data Ready" is signaled when an RDMA Receive completes,
49  * or when a set of RDMA Reads complete.
50  *
51  * An svc_rqst is passed in. This structure contains an array of
52  * free pages (rq_pages) that will contain the incoming RPC message.
53  *
54  * Short messages are moved directly into svc_rqst::rq_arg, and
55  * the RPC Call is ready to be processed by the Upper Layer.
56  * svc_rdma_recvfrom returns the length of the RPC Call message,
57  * completing the reception of the RPC Call.
58  *
59  * However, when an incoming message has Read chunks,
60  * svc_rdma_recvfrom must post RDMA Reads to pull the RPC Call's
61  * data payload from the client. svc_rdma_recvfrom sets up the
62  * RDMA Reads using pages in svc_rqst::rq_pages, which are
63  * transferred to an svc_rdma_op_ctxt for the duration of the
64  * I/O. svc_rdma_recvfrom then returns zero, since the RPC message
65  * is still not yet ready.
66  *
67  * When the Read chunk payloads have become available on the
68  * server, "Data Ready" is raised again, and svc_recv calls
69  * svc_rdma_recvfrom again. This second call may use a different
70  * svc_rqst than the first one, thus any information that needs
71  * to be preserved across these two calls is kept in an
72  * svc_rdma_op_ctxt.
73  *
74  * The second call to svc_rdma_recvfrom performs final assembly
75  * of the RPC Call message, using the RDMA Read sink pages kept in
76  * the svc_rdma_op_ctxt. The xdr_buf is copied from the
77  * svc_rdma_op_ctxt to the second svc_rqst. The second call returns
78  * the length of the completed RPC Call message.
79  *
80  * Page Management
81  *
82  * Pages under I/O must be transferred from the first svc_rqst to an
83  * svc_rdma_op_ctxt before the first svc_rdma_recvfrom call returns.
84  *
85  * The first svc_rqst supplies pages for RDMA Reads. These are moved
86  * from rqstp::rq_pages into ctxt::pages. The consumed elements of
87  * the rq_pages array are set to NULL and refilled with the first
88  * svc_rdma_recvfrom call returns.
89  *
90  * During the second svc_rdma_recvfrom call, RDMA Read sink pages
91  * are transferred from the svc_rdma_op_ctxt to the second svc_rqst
92  * (see rdma_read_complete() below).
93  */
94 
95 #include <asm/unaligned.h>
96 #include <rdma/ib_verbs.h>
97 #include <rdma/rdma_cm.h>
98 
99 #include <linux/spinlock.h>
100 
101 #include <linux/sunrpc/xdr.h>
102 #include <linux/sunrpc/debug.h>
103 #include <linux/sunrpc/rpc_rdma.h>
104 #include <linux/sunrpc/svc_rdma.h>
105 
106 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
107 
108 /*
109  * Replace the pages in the rq_argpages array with the pages from the SGE in
110  * the RDMA_RECV completion. The SGL should contain full pages up until the
111  * last one.
112  */
113 static void svc_rdma_build_arg_xdr(struct svc_rqst *rqstp,
114 				   struct svc_rdma_op_ctxt *ctxt)
115 {
116 	struct page *page;
117 	int sge_no;
118 	u32 len;
119 
120 	/* The reply path assumes the Call's transport header resides
121 	 * in rqstp->rq_pages[0].
122 	 */
123 	page = ctxt->pages[0];
124 	put_page(rqstp->rq_pages[0]);
125 	rqstp->rq_pages[0] = page;
126 
127 	/* Set up the XDR head */
128 	rqstp->rq_arg.head[0].iov_base = page_address(page);
129 	rqstp->rq_arg.head[0].iov_len =
130 		min_t(size_t, ctxt->byte_len, ctxt->sge[0].length);
131 	rqstp->rq_arg.len = ctxt->byte_len;
132 	rqstp->rq_arg.buflen = ctxt->byte_len;
133 
134 	/* Compute bytes past head in the SGL */
135 	len = ctxt->byte_len - rqstp->rq_arg.head[0].iov_len;
136 
137 	/* If data remains, store it in the pagelist */
138 	rqstp->rq_arg.page_len = len;
139 	rqstp->rq_arg.page_base = 0;
140 
141 	sge_no = 1;
142 	while (len && sge_no < ctxt->count) {
143 		page = ctxt->pages[sge_no];
144 		put_page(rqstp->rq_pages[sge_no]);
145 		rqstp->rq_pages[sge_no] = page;
146 		len -= min_t(u32, len, ctxt->sge[sge_no].length);
147 		sge_no++;
148 	}
149 	rqstp->rq_respages = &rqstp->rq_pages[sge_no];
150 	rqstp->rq_next_page = rqstp->rq_respages + 1;
151 
152 	/* If not all pages were used from the SGL, free the remaining ones */
153 	len = sge_no;
154 	while (sge_no < ctxt->count) {
155 		page = ctxt->pages[sge_no++];
156 		put_page(page);
157 	}
158 	ctxt->count = len;
159 
160 	/* Set up tail */
161 	rqstp->rq_arg.tail[0].iov_base = NULL;
162 	rqstp->rq_arg.tail[0].iov_len = 0;
163 }
164 
165 /* This accommodates the largest possible Write chunk,
166  * in one segment.
167  */
168 #define MAX_BYTES_WRITE_SEG	((u32)(RPCSVC_MAXPAGES << PAGE_SHIFT))
169 
170 /* This accommodates the largest possible Position-Zero
171  * Read chunk or Reply chunk, in one segment.
172  */
173 #define MAX_BYTES_SPECIAL_SEG	((u32)((RPCSVC_MAXPAGES + 2) << PAGE_SHIFT))
174 
175 /* Sanity check the Read list.
176  *
177  * Implementation limits:
178  * - This implementation supports only one Read chunk.
179  *
180  * Sanity checks:
181  * - Read list does not overflow buffer.
182  * - Segment size limited by largest NFS data payload.
183  *
184  * The segment count is limited to how many segments can
185  * fit in the transport header without overflowing the
186  * buffer. That's about 40 Read segments for a 1KB inline
187  * threshold.
188  *
189  * Returns pointer to the following Write list.
190  */
191 static __be32 *xdr_check_read_list(__be32 *p, const __be32 *end)
192 {
193 	u32 position;
194 	bool first;
195 
196 	first = true;
197 	while (*p++ != xdr_zero) {
198 		if (first) {
199 			position = be32_to_cpup(p++);
200 			first = false;
201 		} else if (be32_to_cpup(p++) != position) {
202 			return NULL;
203 		}
204 		p++;	/* handle */
205 		if (be32_to_cpup(p++) > MAX_BYTES_SPECIAL_SEG)
206 			return NULL;
207 		p += 2;	/* offset */
208 
209 		if (p > end)
210 			return NULL;
211 	}
212 	return p;
213 }
214 
215 /* The segment count is limited to how many segments can
216  * fit in the transport header without overflowing the
217  * buffer. That's about 60 Write segments for a 1KB inline
218  * threshold.
219  */
220 static __be32 *xdr_check_write_chunk(__be32 *p, const __be32 *end,
221 				     u32 maxlen)
222 {
223 	u32 i, segcount;
224 
225 	segcount = be32_to_cpup(p++);
226 	for (i = 0; i < segcount; i++) {
227 		p++;	/* handle */
228 		if (be32_to_cpup(p++) > maxlen)
229 			return NULL;
230 		p += 2;	/* offset */
231 
232 		if (p > end)
233 			return NULL;
234 	}
235 
236 	return p;
237 }
238 
239 /* Sanity check the Write list.
240  *
241  * Implementation limits:
242  * - This implementation supports only one Write chunk.
243  *
244  * Sanity checks:
245  * - Write list does not overflow buffer.
246  * - Segment size limited by largest NFS data payload.
247  *
248  * Returns pointer to the following Reply chunk.
249  */
250 static __be32 *xdr_check_write_list(__be32 *p, const __be32 *end)
251 {
252 	u32 chcount;
253 
254 	chcount = 0;
255 	while (*p++ != xdr_zero) {
256 		p = xdr_check_write_chunk(p, end, MAX_BYTES_WRITE_SEG);
257 		if (!p)
258 			return NULL;
259 		if (chcount++ > 1)
260 			return NULL;
261 	}
262 	return p;
263 }
264 
265 /* Sanity check the Reply chunk.
266  *
267  * Sanity checks:
268  * - Reply chunk does not overflow buffer.
269  * - Segment size limited by largest NFS data payload.
270  *
271  * Returns pointer to the following RPC header.
272  */
273 static __be32 *xdr_check_reply_chunk(__be32 *p, const __be32 *end)
274 {
275 	if (*p++ != xdr_zero) {
276 		p = xdr_check_write_chunk(p, end, MAX_BYTES_SPECIAL_SEG);
277 		if (!p)
278 			return NULL;
279 	}
280 	return p;
281 }
282 
283 /* On entry, xdr->head[0].iov_base points to first byte in the
284  * RPC-over-RDMA header.
285  *
286  * On successful exit, head[0] points to first byte past the
287  * RPC-over-RDMA header. For RDMA_MSG, this is the RPC message.
288  * The length of the RPC-over-RDMA header is returned.
289  *
290  * Assumptions:
291  * - The transport header is entirely contained in the head iovec.
292  */
293 static int svc_rdma_xdr_decode_req(struct xdr_buf *rq_arg)
294 {
295 	__be32 *p, *end, *rdma_argp;
296 	unsigned int hdr_len;
297 	char *proc;
298 
299 	/* Verify that there's enough bytes for header + something */
300 	if (rq_arg->len <= RPCRDMA_HDRLEN_ERR)
301 		goto out_short;
302 
303 	rdma_argp = rq_arg->head[0].iov_base;
304 	if (*(rdma_argp + 1) != rpcrdma_version)
305 		goto out_version;
306 
307 	switch (*(rdma_argp + 3)) {
308 	case rdma_msg:
309 		proc = "RDMA_MSG";
310 		break;
311 	case rdma_nomsg:
312 		proc = "RDMA_NOMSG";
313 		break;
314 
315 	case rdma_done:
316 		goto out_drop;
317 
318 	case rdma_error:
319 		goto out_drop;
320 
321 	default:
322 		goto out_proc;
323 	}
324 
325 	end = (__be32 *)((unsigned long)rdma_argp + rq_arg->len);
326 	p = xdr_check_read_list(rdma_argp + 4, end);
327 	if (!p)
328 		goto out_inval;
329 	p = xdr_check_write_list(p, end);
330 	if (!p)
331 		goto out_inval;
332 	p = xdr_check_reply_chunk(p, end);
333 	if (!p)
334 		goto out_inval;
335 	if (p > end)
336 		goto out_inval;
337 
338 	rq_arg->head[0].iov_base = p;
339 	hdr_len = (unsigned long)p - (unsigned long)rdma_argp;
340 	rq_arg->head[0].iov_len -= hdr_len;
341 	rq_arg->len -= hdr_len;
342 	dprintk("svcrdma: received %s request for XID 0x%08x, hdr_len=%u\n",
343 		proc, be32_to_cpup(rdma_argp), hdr_len);
344 	return hdr_len;
345 
346 out_short:
347 	dprintk("svcrdma: header too short = %d\n", rq_arg->len);
348 	return -EINVAL;
349 
350 out_version:
351 	dprintk("svcrdma: bad xprt version: %u\n",
352 		be32_to_cpup(rdma_argp + 1));
353 	return -EPROTONOSUPPORT;
354 
355 out_drop:
356 	dprintk("svcrdma: dropping RDMA_DONE/ERROR message\n");
357 	return 0;
358 
359 out_proc:
360 	dprintk("svcrdma: bad rdma procedure (%u)\n",
361 		be32_to_cpup(rdma_argp + 3));
362 	return -EINVAL;
363 
364 out_inval:
365 	dprintk("svcrdma: failed to parse transport header\n");
366 	return -EINVAL;
367 }
368 
369 static void rdma_read_complete(struct svc_rqst *rqstp,
370 			       struct svc_rdma_op_ctxt *head)
371 {
372 	int page_no;
373 
374 	/* Copy RPC pages */
375 	for (page_no = 0; page_no < head->count; page_no++) {
376 		put_page(rqstp->rq_pages[page_no]);
377 		rqstp->rq_pages[page_no] = head->pages[page_no];
378 	}
379 
380 	/* Point rq_arg.pages past header */
381 	rqstp->rq_arg.pages = &rqstp->rq_pages[head->hdr_count];
382 	rqstp->rq_arg.page_len = head->arg.page_len;
383 
384 	/* rq_respages starts after the last arg page */
385 	rqstp->rq_respages = &rqstp->rq_pages[page_no];
386 	rqstp->rq_next_page = rqstp->rq_respages + 1;
387 
388 	/* Rebuild rq_arg head and tail. */
389 	rqstp->rq_arg.head[0] = head->arg.head[0];
390 	rqstp->rq_arg.tail[0] = head->arg.tail[0];
391 	rqstp->rq_arg.len = head->arg.len;
392 	rqstp->rq_arg.buflen = head->arg.buflen;
393 }
394 
395 static void svc_rdma_send_error(struct svcxprt_rdma *xprt,
396 				__be32 *rdma_argp, int status)
397 {
398 	struct svc_rdma_op_ctxt *ctxt;
399 	__be32 *p, *err_msgp;
400 	unsigned int length;
401 	struct page *page;
402 	int ret;
403 
404 	page = alloc_page(GFP_KERNEL);
405 	if (!page)
406 		return;
407 	err_msgp = page_address(page);
408 
409 	p = err_msgp;
410 	*p++ = *rdma_argp;
411 	*p++ = *(rdma_argp + 1);
412 	*p++ = xprt->sc_fc_credits;
413 	*p++ = rdma_error;
414 	if (status == -EPROTONOSUPPORT) {
415 		*p++ = err_vers;
416 		*p++ = rpcrdma_version;
417 		*p++ = rpcrdma_version;
418 	} else {
419 		*p++ = err_chunk;
420 	}
421 	length = (unsigned long)p - (unsigned long)err_msgp;
422 
423 	/* Map transport header; no RPC message payload */
424 	ctxt = svc_rdma_get_context(xprt);
425 	ret = svc_rdma_map_reply_hdr(xprt, ctxt, err_msgp, length);
426 	if (ret) {
427 		dprintk("svcrdma: Error %d mapping send for protocol error\n",
428 			ret);
429 		return;
430 	}
431 
432 	ret = svc_rdma_post_send_wr(xprt, ctxt, 1, 0);
433 	if (ret) {
434 		dprintk("svcrdma: Error %d posting send for protocol error\n",
435 			ret);
436 		svc_rdma_unmap_dma(ctxt);
437 		svc_rdma_put_context(ctxt, 1);
438 	}
439 }
440 
441 /* By convention, backchannel calls arrive via rdma_msg type
442  * messages, and never populate the chunk lists. This makes
443  * the RPC/RDMA header small and fixed in size, so it is
444  * straightforward to check the RPC header's direction field.
445  */
446 static bool svc_rdma_is_backchannel_reply(struct svc_xprt *xprt,
447 					  __be32 *rdma_resp)
448 {
449 	__be32 *p;
450 
451 	if (!xprt->xpt_bc_xprt)
452 		return false;
453 
454 	p = rdma_resp + 3;
455 	if (*p++ != rdma_msg)
456 		return false;
457 
458 	if (*p++ != xdr_zero)
459 		return false;
460 	if (*p++ != xdr_zero)
461 		return false;
462 	if (*p++ != xdr_zero)
463 		return false;
464 
465 	/* XID sanity */
466 	if (*p++ != *rdma_resp)
467 		return false;
468 	/* call direction */
469 	if (*p == cpu_to_be32(RPC_CALL))
470 		return false;
471 
472 	return true;
473 }
474 
475 /**
476  * svc_rdma_recvfrom - Receive an RPC call
477  * @rqstp: request structure into which to receive an RPC Call
478  *
479  * Returns:
480  *	The positive number of bytes in the RPC Call message,
481  *	%0 if there were no Calls ready to return,
482  *	%-EINVAL if the Read chunk data is too large,
483  *	%-ENOMEM if rdma_rw context pool was exhausted,
484  *	%-ENOTCONN if posting failed (connection is lost),
485  *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
486  *
487  * Called in a loop when XPT_DATA is set. XPT_DATA is cleared only
488  * when there are no remaining ctxt's to process.
489  *
490  * The next ctxt is removed from the "receive" lists.
491  *
492  * - If the ctxt completes a Read, then finish assembling the Call
493  *   message and return the number of bytes in the message.
494  *
495  * - If the ctxt completes a Receive, then construct the Call
496  *   message from the contents of the Receive buffer.
497  *
498  *   - If there are no Read chunks in this message, then finish
499  *     assembling the Call message and return the number of bytes
500  *     in the message.
501  *
502  *   - If there are Read chunks in this message, post Read WRs to
503  *     pull that payload and return 0.
504  */
505 int svc_rdma_recvfrom(struct svc_rqst *rqstp)
506 {
507 	struct svc_xprt *xprt = rqstp->rq_xprt;
508 	struct svcxprt_rdma *rdma_xprt =
509 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
510 	struct svc_rdma_op_ctxt *ctxt;
511 	__be32 *p;
512 	int ret;
513 
514 	spin_lock(&rdma_xprt->sc_rq_dto_lock);
515 	if (!list_empty(&rdma_xprt->sc_read_complete_q)) {
516 		ctxt = list_first_entry(&rdma_xprt->sc_read_complete_q,
517 					struct svc_rdma_op_ctxt, list);
518 		list_del(&ctxt->list);
519 		spin_unlock(&rdma_xprt->sc_rq_dto_lock);
520 		rdma_read_complete(rqstp, ctxt);
521 		goto complete;
522 	} else if (!list_empty(&rdma_xprt->sc_rq_dto_q)) {
523 		ctxt = list_first_entry(&rdma_xprt->sc_rq_dto_q,
524 					struct svc_rdma_op_ctxt, list);
525 		list_del(&ctxt->list);
526 	} else {
527 		/* No new incoming requests, terminate the loop */
528 		clear_bit(XPT_DATA, &xprt->xpt_flags);
529 		spin_unlock(&rdma_xprt->sc_rq_dto_lock);
530 		return 0;
531 	}
532 	spin_unlock(&rdma_xprt->sc_rq_dto_lock);
533 
534 	dprintk("svcrdma: recvfrom: ctxt=%p on xprt=%p, rqstp=%p\n",
535 		ctxt, rdma_xprt, rqstp);
536 	atomic_inc(&rdma_stat_recv);
537 
538 	svc_rdma_build_arg_xdr(rqstp, ctxt);
539 
540 	p = (__be32 *)rqstp->rq_arg.head[0].iov_base;
541 	ret = svc_rdma_xdr_decode_req(&rqstp->rq_arg);
542 	if (ret < 0)
543 		goto out_err;
544 	if (ret == 0)
545 		goto out_drop;
546 	rqstp->rq_xprt_hlen = ret;
547 
548 	if (svc_rdma_is_backchannel_reply(xprt, p)) {
549 		ret = svc_rdma_handle_bc_reply(xprt->xpt_bc_xprt, p,
550 					       &rqstp->rq_arg);
551 		svc_rdma_put_context(ctxt, 0);
552 		return ret;
553 	}
554 
555 	p += rpcrdma_fixed_maxsz;
556 	if (*p != xdr_zero)
557 		goto out_readchunk;
558 
559 complete:
560 	svc_rdma_put_context(ctxt, 0);
561 	dprintk("svcrdma: recvfrom: xprt=%p, rqstp=%p, rq_arg.len=%u\n",
562 		rdma_xprt, rqstp, rqstp->rq_arg.len);
563 	rqstp->rq_prot = IPPROTO_MAX;
564 	svc_xprt_copy_addrs(rqstp, xprt);
565 	return rqstp->rq_arg.len;
566 
567 out_readchunk:
568 	ret = svc_rdma_recv_read_chunk(rdma_xprt, rqstp, ctxt, p);
569 	if (ret < 0)
570 		goto out_postfail;
571 	return 0;
572 
573 out_err:
574 	svc_rdma_send_error(rdma_xprt, p, ret);
575 	svc_rdma_put_context(ctxt, 0);
576 	return 0;
577 
578 out_postfail:
579 	if (ret == -EINVAL)
580 		svc_rdma_send_error(rdma_xprt, p, ret);
581 	svc_rdma_put_context(ctxt, 1);
582 	return ret;
583 
584 out_drop:
585 	svc_rdma_put_context(ctxt, 1);
586 	return 0;
587 }
588