[linux-2.6-block.git] / net / rds / message.c

/*
 * Copyright (c) 2006 Oracle.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/export.h>

#include "rds.h"

static unsigned int	rds_exthdr_size[__RDS_EXTHDR_MAX] = {
[RDS_EXTHDR_NONE]	= 0,
[RDS_EXTHDR_VERSION]	= sizeof(struct rds_ext_header_version),
[RDS_EXTHDR_RDMA]	= sizeof(struct rds_ext_header_rdma),
[RDS_EXTHDR_RDMA_DEST]	= sizeof(struct rds_ext_header_rdma_dest),
[RDS_EXTHDR_NPATHS]	= sizeof(u16),
};


void rds_message_addref(struct rds_message *rm)
{
	rdsdebug("addref rm %p ref %d\n", rm, atomic_read(&rm->m_refcount));
	atomic_inc(&rm->m_refcount);
}
EXPORT_SYMBOL_GPL(rds_message_addref);

/*
 * This relies on dma_map_sg() not touching sg[].page during merging.
 */
static void rds_message_purge(struct rds_message *rm)
{
	unsigned long i;

	if (unlikely(test_bit(RDS_MSG_PAGEVEC, &rm->m_flags)))
		return;

	for (i = 0; i < rm->data.op_nents; i++) {
		rdsdebug("putting data page %p\n", (void *)sg_page(&rm->data.op_sg[i]));
		/* XXX will have to put_page for page refs */
		__free_page(sg_page(&rm->data.op_sg[i]));
	}
	rm->data.op_nents = 0;

	if (rm->rdma.op_active)
		rds_rdma_free_op(&rm->rdma);
	if (rm->rdma.op_rdma_mr)
		rds_mr_put(rm->rdma.op_rdma_mr);

	if (rm->atomic.op_active)
		rds_atomic_free_op(&rm->atomic);
	if (rm->atomic.op_rdma_mr)
		rds_mr_put(rm->atomic.op_rdma_mr);
}

void rds_message_put(struct rds_message *rm)
{
	rdsdebug("put rm %p ref %d\n", rm, atomic_read(&rm->m_refcount));
	WARN(!atomic_read(&rm->m_refcount), "danger refcount zero on %p\n", rm);
	if (atomic_dec_and_test(&rm->m_refcount)) {
		BUG_ON(!list_empty(&rm->m_sock_item));
		BUG_ON(!list_empty(&rm->m_conn_item));
		rds_message_purge(rm);

		kfree(rm);
	}
}
EXPORT_SYMBOL_GPL(rds_message_put);

void rds_message_populate_header(struct rds_header *hdr, __be16 sport,
				 __be16 dport, u64 seq)
{
	hdr->h_flags = 0;
	hdr->h_sport = sport;
	hdr->h_dport = dport;
	hdr->h_sequence = cpu_to_be64(seq);
	hdr->h_exthdr[0] = RDS_EXTHDR_NONE;
}
EXPORT_SYMBOL_GPL(rds_message_populate_header);

int rds_message_add_extension(struct rds_header *hdr, unsigned int type,
			      const void *data, unsigned int len)
{
	unsigned int ext_len = sizeof(u8) + len;
	unsigned char *dst;

	/* For now, refuse to add more than one extension header */
	if (hdr->h_exthdr[0] != RDS_EXTHDR_NONE)
		return 0;

	if (type >= __RDS_EXTHDR_MAX || len != rds_exthdr_size[type])
		return 0;

	if (ext_len >= RDS_HEADER_EXT_SPACE)
		return 0;
	dst = hdr->h_exthdr;

	*dst++ = type;
	memcpy(dst, data, len);

	dst[len] = RDS_EXTHDR_NONE;
	return 1;
}
EXPORT_SYMBOL_GPL(rds_message_add_extension);

/*
 * If a message has extension headers, retrieve them here.
 * Call like this:
 *
 * unsigned int pos = 0;
 *
 * while (1) {
 *	buflen = sizeof(buffer);
 *	type = rds_message_next_extension(hdr, &pos, buffer, &buflen);
 *	if (type == RDS_EXTHDR_NONE)
 *		break;
 *	...
 * }
 */
int rds_message_next_extension(struct rds_header *hdr,
		unsigned int *pos, void *buf, unsigned int *buflen)
{
	unsigned int offset, ext_type, ext_len;
	u8 *src = hdr->h_exthdr;

	offset = *pos;
	if (offset >= RDS_HEADER_EXT_SPACE)
		goto none;

	/* Get the extension type and length. For now, the
	 * length is implied by the extension type. */
	ext_type = src[offset++];

	if (ext_type == RDS_EXTHDR_NONE || ext_type >= __RDS_EXTHDR_MAX)
		goto none;
	ext_len = rds_exthdr_size[ext_type];
	if (offset + ext_len > RDS_HEADER_EXT_SPACE)
		goto none;

	*pos = offset + ext_len;
	if (ext_len < *buflen)
		*buflen = ext_len;
	memcpy(buf, src + offset, *buflen);
	return ext_type;

none:
	*pos = RDS_HEADER_EXT_SPACE;
	*buflen = 0;
	return RDS_EXTHDR_NONE;
}

int rds_message_add_rdma_dest_extension(struct rds_header *hdr, u32 r_key, u32 offset)
{
	struct rds_ext_header_rdma_dest ext_hdr;

	ext_hdr.h_rdma_rkey = cpu_to_be32(r_key);
	ext_hdr.h_rdma_offset = cpu_to_be32(offset);
	return rds_message_add_extension(hdr, RDS_EXTHDR_RDMA_DEST, &ext_hdr, sizeof(ext_hdr));
}
EXPORT_SYMBOL_GPL(rds_message_add_rdma_dest_extension);

/*
 * Each rds_message is allocated with extra space for the scatterlist entries
 * rds ops will need. This is to minimize memory allocation count. Then, each rds op
 * can grab SGs when initializing its part of the rds_message.
 */
struct rds_message *rds_message_alloc(unsigned int extra_len, gfp_t gfp)
{
	struct rds_message *rm;

	if (extra_len > KMALLOC_MAX_SIZE - sizeof(struct rds_message))
		return NULL;

	rm = kzalloc(sizeof(struct rds_message) + extra_len, gfp);
	if (!rm)
		goto out;

	rm->m_used_sgs = 0;
	rm->m_total_sgs = extra_len / sizeof(struct scatterlist);

	atomic_set(&rm->m_refcount, 1);
	INIT_LIST_HEAD(&rm->m_sock_item);
	INIT_LIST_HEAD(&rm->m_conn_item);
	spin_lock_init(&rm->m_rs_lock);
	init_waitqueue_head(&rm->m_flush_wait);

out:
	return rm;
}

/*
 * RDS ops use this to grab SG entries from the rm's sg pool.
 */
struct scatterlist *rds_message_alloc_sgs(struct rds_message *rm, int nents)
{
	struct scatterlist *sg_first = (struct scatterlist *) &rm[1];
	struct scatterlist *sg_ret;

	WARN_ON(rm->m_used_sgs + nents > rm->m_total_sgs);
	WARN_ON(!nents);

	if (rm->m_used_sgs + nents > rm->m_total_sgs)
		return NULL;

	sg_ret = &sg_first[rm->m_used_sgs];
	sg_init_table(sg_ret, nents);
	rm->m_used_sgs += nents;

	return sg_ret;
}

struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned int total_len)
{
	struct rds_message *rm;
	unsigned int i;
	int num_sgs = ceil(total_len, PAGE_SIZE);
	int extra_bytes = num_sgs * sizeof(struct scatterlist);

	rm = rds_message_alloc(extra_bytes, GFP_NOWAIT);
	if (!rm)
		return ERR_PTR(-ENOMEM);

	set_bit(RDS_MSG_PAGEVEC, &rm->m_flags);
	rm->m_inc.i_hdr.h_len = cpu_to_be32(total_len);
	rm->data.op_nents = ceil(total_len, PAGE_SIZE);
	rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
	if (!rm->data.op_sg) {
		rds_message_put(rm);
		return ERR_PTR(-ENOMEM);
	}

	for (i = 0; i < rm->data.op_nents; ++i) {
		sg_set_page(&rm->data.op_sg[i],
				virt_to_page(page_addrs[i]),
				PAGE_SIZE, 0);
	}

	return rm;
}

int rds_message_copy_from_user(struct rds_message *rm, struct iov_iter *from)
{
	unsigned long to_copy, nbytes;
	unsigned long sg_off;
	struct scatterlist *sg;
	int ret = 0;

	rm->m_inc.i_hdr.h_len = cpu_to_be32(iov_iter_count(from));

	/*
	 * now allocate and copy in the data payload.
	 */
	sg = rm->data.op_sg;
	sg_off = 0; /* Dear gcc, sg->page will be null from kzalloc. */

	while (iov_iter_count(from)) {
		if (!sg_page(sg)) {
			ret = rds_page_remainder_alloc(sg, iov_iter_count(from),
						       GFP_HIGHUSER);
			if (ret)
				return ret;
			rm->data.op_nents++;
			sg_off = 0;
		}

		to_copy = min_t(unsigned long, iov_iter_count(from),
				sg->length - sg_off);

		rds_stats_add(s_copy_from_user, to_copy);
		nbytes = copy_page_from_iter(sg_page(sg), sg->offset + sg_off,
					     to_copy, from);
		if (nbytes != to_copy)
			return -EFAULT;

		sg_off += to_copy;

		if (sg_off == sg->length)
			sg++;
	}

	return ret;
}

int rds_message_inc_copy_to_user(struct rds_incoming *inc, struct iov_iter *to)
{
	struct rds_message *rm;
	struct scatterlist *sg;
	unsigned long to_copy;
	unsigned long vec_off;
	int copied;
	int ret;
	u32 len;

	rm = container_of(inc, struct rds_message, m_inc);
	len = be32_to_cpu(rm->m_inc.i_hdr.h_len);

	sg = rm->data.op_sg;
	vec_off = 0;
	copied = 0;

	while (iov_iter_count(to) && copied < len) {
		to_copy = min_t(unsigned long, iov_iter_count(to),
				sg->length - vec_off);
		to_copy = min_t(unsigned long, to_copy, len - copied);

		rds_stats_add(s_copy_to_user, to_copy);
		ret = copy_page_to_iter(sg_page(sg), sg->offset + vec_off,
					to_copy, to);
		if (ret != to_copy)
			return -EFAULT;

		vec_off += to_copy;
		copied += to_copy;

		if (vec_off == sg->length) {
			vec_off = 0;
			sg++;
		}
	}

	return copied;
}

/*
 * If the message is still on the send queue, wait until the transport
 * is done with it. This is particularly important for RDMA operations.
 */
void rds_message_wait(struct rds_message *rm)
{
	wait_event_interruptible(rm->m_flush_wait,
			!test_bit(RDS_MSG_MAPPED, &rm->m_flags));
}

void rds_message_unmapped(struct rds_message *rm)
{
	clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
	wake_up_interruptible(&rm->m_flush_wait);
}
EXPORT_SYMBOL_GPL(rds_message_unmapped);
Commit	Line	Data
7875e18e AG	1	/*
	2	* Copyright (c) 2006 Oracle. All rights reserved.
	3	*
	4	* This software is available to you under a choice of one of two
	5	* licenses. You may choose to be licensed under the terms of the GNU
	6	* General Public License (GPL) Version 2, available from the file
	7	* COPYING in the main directory of this source tree, or the
	8	* OpenIB.org BSD license below:
	9	*
	10	* Redistribution and use in source and binary forms, with or
	11	* without modification, are permitted provided that the following
	12	* conditions are met:
	13	*
	14	* - Redistributions of source code must retain the above
	15	* copyright notice, this list of conditions and the following
	16	* disclaimer.
	17	*
	18	* - Redistributions in binary form must reproduce the above
	19	* copyright notice, this list of conditions and the following
	20	* disclaimer in the documentation and/or other materials
	21	* provided with the distribution.
	22	*
	23	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
	24	* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
	25	* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
	26	* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
	27	* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
	28	* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
	29	* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
	30	* SOFTWARE.
	31	*
	32	*/
	33	#include <linux/kernel.h>
5a0e3ad6	34	#include <linux/slab.h>
bc3b2d7f	35	#include <linux/export.h>
7875e18e AG	36
7875e18e AG	37	#include "rds.h"
7875e18e	38
7875e18e AG	39	static unsigned int rds_exthdr_size[__RDS_EXTHDR_MAX] = {
	40	[RDS_EXTHDR_NONE] = 0,
	41	[RDS_EXTHDR_VERSION] = sizeof(struct rds_ext_header_version),
	42	[RDS_EXTHDR_RDMA] = sizeof(struct rds_ext_header_rdma),
	43	[RDS_EXTHDR_RDMA_DEST] = sizeof(struct rds_ext_header_rdma_dest),
5916e2c1	44	[RDS_EXTHDR_NPATHS] = sizeof(u16),
7875e18e AG	45	};
	46
	47
	48	void rds_message_addref(struct rds_message *rm)
	49	{
	50	rdsdebug("addref rm %p ref %d\n", rm, atomic_read(&rm->m_refcount));
	51	atomic_inc(&rm->m_refcount);
	52	}
616b757a	53	EXPORT_SYMBOL_GPL(rds_message_addref);
7875e18e AG	54
	55	/*
	56	* This relies on dma_map_sg() not touching sg[].page during merging.
	57	*/
	58	static void rds_message_purge(struct rds_message *rm)
	59	{
	60	unsigned long i;
	61
	62	if (unlikely(test_bit(RDS_MSG_PAGEVEC, &rm->m_flags)))
	63	return;
	64
6c7cc6e4 AG	65	for (i = 0; i < rm->data.op_nents; i++) {
6c7cc6e4 AG	66	rdsdebug("putting data page %p\n", (void *)sg_page(&rm->data.op_sg[i]));
7875e18e	67	/* XXX will have to put_page for page refs */
6c7cc6e4	68	__free_page(sg_page(&rm->data.op_sg[i]));
7875e18e	69	}
6c7cc6e4	70	rm->data.op_nents = 0;
7875e18e	71
f8b3aaf2 AG	72	if (rm->rdma.op_active)
	73	rds_rdma_free_op(&rm->rdma);
	74	if (rm->rdma.op_rdma_mr)
	75	rds_mr_put(rm->rdma.op_rdma_mr);
d0ab25a8 AG	76
	77	if (rm->atomic.op_active)
	78	rds_atomic_free_op(&rm->atomic);
	79	if (rm->atomic.op_rdma_mr)
	80	rds_mr_put(rm->atomic.op_rdma_mr);
7875e18e	81	}
7875e18e AG	82
	83	void rds_message_put(struct rds_message *rm)
	84	{
	85	rdsdebug("put rm %p ref %d\n", rm, atomic_read(&rm->m_refcount));
7dac1b51	86	WARN(!atomic_read(&rm->m_refcount), "danger refcount zero on %p\n", rm);
7875e18e AG	87	if (atomic_dec_and_test(&rm->m_refcount)) {
	88	BUG_ON(!list_empty(&rm->m_sock_item));
	89	BUG_ON(!list_empty(&rm->m_conn_item));
	90	rds_message_purge(rm);
	91
	92	kfree(rm);
	93	}
	94	}
616b757a	95	EXPORT_SYMBOL_GPL(rds_message_put);
7875e18e	96
7875e18e AG	97	void rds_message_populate_header(struct rds_header *hdr, __be16 sport,
	98	__be16 dport, u64 seq)
	99	{
	100	hdr->h_flags = 0;
	101	hdr->h_sport = sport;
	102	hdr->h_dport = dport;
	103	hdr->h_sequence = cpu_to_be64(seq);
	104	hdr->h_exthdr[0] = RDS_EXTHDR_NONE;
	105	}
616b757a	106	EXPORT_SYMBOL_GPL(rds_message_populate_header);
7875e18e	107
ff51bf84	108	int rds_message_add_extension(struct rds_header *hdr, unsigned int type,
ff51bf84	109	const void *data, unsigned int len)
7875e18e AG	110	{
	111	unsigned int ext_len = sizeof(u8) + len;
	112	unsigned char *dst;
	113
	114	/* For now, refuse to add more than one extension header */
	115	if (hdr->h_exthdr[0] != RDS_EXTHDR_NONE)
	116	return 0;
	117
f64f9e71	118	if (type >= __RDS_EXTHDR_MAX \|\| len != rds_exthdr_size[type])
7875e18e AG	119	return 0;
	120
	121	if (ext_len >= RDS_HEADER_EXT_SPACE)
	122	return 0;
	123	dst = hdr->h_exthdr;
	124
	125	*dst++ = type;
	126	memcpy(dst, data, len);
	127
	128	dst[len] = RDS_EXTHDR_NONE;
	129	return 1;
	130	}
616b757a	131	EXPORT_SYMBOL_GPL(rds_message_add_extension);
7875e18e AG	132
	133	/*
	134	* If a message has extension headers, retrieve them here.
	135	* Call like this:
	136	*
	137	* unsigned int pos = 0;
	138	*
	139	* while (1) {
	140	* buflen = sizeof(buffer);
	141	* type = rds_message_next_extension(hdr, &pos, buffer, &buflen);
	142	* if (type == RDS_EXTHDR_NONE)
	143	* break;
	144	* ...
	145	* }
	146	*/
	147	int rds_message_next_extension(struct rds_header *hdr,
	148	unsigned int pos, void buf, unsigned int *buflen)
	149	{
	150	unsigned int offset, ext_type, ext_len;
	151	u8 *src = hdr->h_exthdr;
	152
	153	offset = *pos;
	154	if (offset >= RDS_HEADER_EXT_SPACE)
	155	goto none;
	156
	157	/* Get the extension type and length. For now, the
	158	* length is implied by the extension type. */
	159	ext_type = src[offset++];
	160
	161	if (ext_type == RDS_EXTHDR_NONE \|\| ext_type >= __RDS_EXTHDR_MAX)
	162	goto none;
	163	ext_len = rds_exthdr_size[ext_type];
	164	if (offset + ext_len > RDS_HEADER_EXT_SPACE)
	165	goto none;
	166
	167	*pos = offset + ext_len;
	168	if (ext_len < *buflen)
	169	*buflen = ext_len;
	170	memcpy(buf, src + offset, *buflen);
	171	return ext_type;
	172
	173	none:
	174	*pos = RDS_HEADER_EXT_SPACE;
	175	*buflen = 0;
	176	return RDS_EXTHDR_NONE;
	177	}
	178
7875e18e AG	179	int rds_message_add_rdma_dest_extension(struct rds_header *hdr, u32 r_key, u32 offset)
	180	{
	181	struct rds_ext_header_rdma_dest ext_hdr;
	182
	183	ext_hdr.h_rdma_rkey = cpu_to_be32(r_key);
	184	ext_hdr.h_rdma_offset = cpu_to_be32(offset);
	185	return rds_message_add_extension(hdr, RDS_EXTHDR_RDMA_DEST, &ext_hdr, sizeof(ext_hdr));
	186	}
616b757a	187	EXPORT_SYMBOL_GPL(rds_message_add_rdma_dest_extension);
7875e18e	188
fc445084 AG	189	/*
	190	* Each rds_message is allocated with extra space for the scatterlist entries
	191	* rds ops will need. This is to minimize memory allocation count. Then, each rds op
	192	* can grab SGs when initializing its part of the rds_message.
	193	*/
	194	struct rds_message *rds_message_alloc(unsigned int extra_len, gfp_t gfp)
7875e18e AG	195	{
	196	struct rds_message *rm;
	197
ece6b0a2 CW	198	if (extra_len > KMALLOC_MAX_SIZE - sizeof(struct rds_message))
	199	return NULL;
	200
fc445084	201	rm = kzalloc(sizeof(struct rds_message) + extra_len, gfp);
7875e18e AG	202	if (!rm)
	203	goto out;
	204
fc445084 AG	205	rm->m_used_sgs = 0;
	206	rm->m_total_sgs = extra_len / sizeof(struct scatterlist);
	207
7875e18e AG	208	atomic_set(&rm->m_refcount, 1);
	209	INIT_LIST_HEAD(&rm->m_sock_item);
	210	INIT_LIST_HEAD(&rm->m_conn_item);
	211	spin_lock_init(&rm->m_rs_lock);
c83188dc	212	init_waitqueue_head(&rm->m_flush_wait);
7875e18e AG	213
	214	out:
	215	return rm;
	216	}
	217
fc445084 AG	218	/*
	219	* RDS ops use this to grab SG entries from the rm's sg pool.
	220	*/
	221	struct scatterlist rds_message_alloc_sgs(struct rds_message rm, int nents)
	222	{
	223	struct scatterlist sg_first = (struct scatterlist ) &rm[1];
	224	struct scatterlist *sg_ret;
	225
	226	WARN_ON(rm->m_used_sgs + nents > rm->m_total_sgs);
ee4c7b47	227	WARN_ON(!nents);
fc445084	228
d139ff09 AG	229	if (rm->m_used_sgs + nents > rm->m_total_sgs)
	230	return NULL;
	231
fc445084	232	sg_ret = &sg_first[rm->m_used_sgs];
f4dd96f7	233	sg_init_table(sg_ret, nents);
fc445084 AG	234	rm->m_used_sgs += nents;
	235
	236	return sg_ret;
	237	}
	238
7875e18e AG	239	struct rds_message rds_message_map_pages(unsigned long page_addrs, unsigned int total_len)
	240	{
	241	struct rds_message *rm;
	242	unsigned int i;
ff87e97a AG	243	int num_sgs = ceil(total_len, PAGE_SIZE);
ff87e97a AG	244	int extra_bytes = num_sgs * sizeof(struct scatterlist);
7875e18e	245
f2ec76f2	246	rm = rds_message_alloc(extra_bytes, GFP_NOWAIT);
8690bfa1	247	if (!rm)
7875e18e AG	248	return ERR_PTR(-ENOMEM);
	249
	250	set_bit(RDS_MSG_PAGEVEC, &rm->m_flags);
	251	rm->m_inc.i_hdr.h_len = cpu_to_be32(total_len);
6c7cc6e4 AG	252	rm->data.op_nents = ceil(total_len, PAGE_SIZE);
6c7cc6e4 AG	253	rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
aa58163a PE	254	if (!rm->data.op_sg) {
aa58163a PE	255	rds_message_put(rm);
d139ff09	256	return ERR_PTR(-ENOMEM);
aa58163a	257	}
7875e18e	258
6c7cc6e4 AG	259	for (i = 0; i < rm->data.op_nents; ++i) {
6c7cc6e4 AG	260	sg_set_page(&rm->data.op_sg[i],
7875e18e AG	261	virt_to_page(page_addrs[i]),
	262	PAGE_SIZE, 0);
	263	}
	264
	265	return rm;
	266	}
	267
083735f4	268	int rds_message_copy_from_user(struct rds_message rm, struct iov_iter from)
7875e18e	269	{
d0a47d32	270	unsigned long to_copy, nbytes;
7875e18e	271	unsigned long sg_off;
7875e18e	272	struct scatterlist *sg;
fc445084	273	int ret = 0;
7875e18e	274
083735f4	275	rm->m_inc.i_hdr.h_len = cpu_to_be32(iov_iter_count(from));
7875e18e AG	276
	277	/*
	278	* now allocate and copy in the data payload.
	279	*/
6c7cc6e4	280	sg = rm->data.op_sg;
7875e18e AG	281	sg_off = 0; /* Dear gcc, sg->page will be null from kzalloc. */
7875e18e AG	282
083735f4	283	while (iov_iter_count(from)) {
8690bfa1	284	if (!sg_page(sg)) {
083735f4	285	ret = rds_page_remainder_alloc(sg, iov_iter_count(from),
7875e18e AG	286	GFP_HIGHUSER);
7875e18e AG	287	if (ret)
083735f4	288	return ret;
6c7cc6e4	289	rm->data.op_nents++;
7875e18e AG	290	sg_off = 0;
	291	}
	292
083735f4 AV	293	to_copy = min_t(unsigned long, iov_iter_count(from),
083735f4 AV	294	sg->length - sg_off);
7875e18e	295
083735f4	296	rds_stats_add(s_copy_from_user, to_copy);
d0a47d32 SV	297	nbytes = copy_page_from_iter(sg_page(sg), sg->offset + sg_off,
	298	to_copy, from);
	299	if (nbytes != to_copy)
083735f4	300	return -EFAULT;
7875e18e	301
7875e18e AG	302	sg_off += to_copy;
	303
	304	if (sg_off == sg->length)
	305	sg++;
	306	}
	307
fc445084	308	return ret;
7875e18e AG	309	}
7875e18e AG	310
c310e72c	311	int rds_message_inc_copy_to_user(struct rds_incoming inc, struct iov_iter to)
7875e18e AG	312	{
7875e18e AG	313	struct rds_message *rm;
7875e18e AG	314	struct scatterlist *sg;
7875e18e AG	315	unsigned long to_copy;
7875e18e AG	316	unsigned long vec_off;
	317	int copied;
	318	int ret;
	319	u32 len;
	320
	321	rm = container_of(inc, struct rds_message, m_inc);
	322	len = be32_to_cpu(rm->m_inc.i_hdr.h_len);
	323
6c7cc6e4	324	sg = rm->data.op_sg;
7875e18e AG	325	vec_off = 0;
	326	copied = 0;
	327
c310e72c	328	while (iov_iter_count(to) && copied < len) {
6ff4a8ad GU	329	to_copy = min_t(unsigned long, iov_iter_count(to),
6ff4a8ad GU	330	sg->length - vec_off);
7875e18e AG	331	to_copy = min_t(unsigned long, to_copy, len - copied);
7875e18e AG	332
c310e72c AV	333	rds_stats_add(s_copy_to_user, to_copy);
	334	ret = copy_page_to_iter(sg_page(sg), sg->offset + vec_off,
	335	to_copy, to);
	336	if (ret != to_copy)
	337	return -EFAULT;
7875e18e	338
7875e18e AG	339	vec_off += to_copy;
	340	copied += to_copy;
	341
	342	if (vec_off == sg->length) {
	343	vec_off = 0;
	344	sg++;
	345	}
	346	}
	347
	348	return copied;
	349	}
	350
	351	/*
	352	* If the message is still on the send queue, wait until the transport
	353	* is done with it. This is particularly important for RDMA operations.
	354	*/
	355	void rds_message_wait(struct rds_message *rm)
	356	{
c83188dc	357	wait_event_interruptible(rm->m_flush_wait,
7875e18e AG	358	!test_bit(RDS_MSG_MAPPED, &rm->m_flags));
	359	}
	360
	361	void rds_message_unmapped(struct rds_message *rm)
	362	{
	363	clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
c83188dc	364	wake_up_interruptible(&rm->m_flush_wait);
7875e18e	365	}
616b757a	366	EXPORT_SYMBOL_GPL(rds_message_unmapped);
7875e18e	367