linux/net/rds/message.c
Sowmini Varadhan 905dd4184e RDS: TCP: Track peer's connection generation number
The RDS transport has to be able to distinguish between
two types of failure events:
(a) when the transport fails (e.g., TCP connection reset)
    but the RDS socket/connection layer on both sides stays
    the same
(b) when the peer's RDS layer itself resets (e.g., due to module
    reload or machine reboot at the peer)
In case (a) both sides must reconnect and continue the RDS messaging
without any message loss or disruption to the message sequence numbers,
and this is achieved by rds_send_path_reset().

In case (b) we should reset all rds_connection state to the
new incarnation of the peer. Examples of state that needs to
be reset are next expected rx sequence number from, or messages to be
retransmitted to, the new incarnation of the peer.

To achieve this, the RDS handshake probe added as part of
commit 5916e2c155 ("RDS: TCP: Enable multipath RDS for TCP")
is enhanced so that sender and receiver of the RDS ping-probe
will add a generation number as part of the RDS_EXTHDR_GEN_NUM
extension header. Each peer stores local and remote generation
numbers as part of each rds_connection. Changes in generation
number will be detected via incoming handshake probe ping
request or response and will allow the receiver to reset rds_connection
state.

Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-17 13:35:18 -05:00

369 lines
9.5 KiB
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),
[RDS_EXTHDR_GEN_NUM] = sizeof(u32),
};
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);