linux/drivers/infiniband/sw/rxe/rxe_queue.c
Bob Pearson ae6e843fe0 RDMA/rxe: Add memory barriers to kernel queues
Earlier patches added memory barriers to protect user space to kernel
space communications. The user space queues were previously shown to have
occasional memory synchonization errors which were removed by adding
smp_load_acquire, smp_store_release barriers.  This patch extends that to
the case where queues are used between kernel space threads.

This patch also extends the queue types to include kernel ULP queues which
access the other end of the queues in kernel verbs calls like poll_cq and
post_send/recv.

Link: https://lore.kernel.org/r/20210914164206.19768-2-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2021-09-24 10:14:59 -03:00

200 lines
4.4 KiB
C

// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
/*
* Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
* Copyright (c) 2015 System Fabric Works, Inc. All rights reserved.
*/
#include <linux/vmalloc.h>
#include "rxe.h"
#include "rxe_loc.h"
#include "rxe_queue.h"
int do_mmap_info(struct rxe_dev *rxe, struct mminfo __user *outbuf,
struct ib_udata *udata, struct rxe_queue_buf *buf,
size_t buf_size, struct rxe_mmap_info **ip_p)
{
int err;
struct rxe_mmap_info *ip = NULL;
if (outbuf) {
ip = rxe_create_mmap_info(rxe, buf_size, udata, buf);
if (IS_ERR(ip)) {
err = PTR_ERR(ip);
goto err1;
}
if (copy_to_user(outbuf, &ip->info, sizeof(ip->info))) {
err = -EFAULT;
goto err2;
}
spin_lock_bh(&rxe->pending_lock);
list_add(&ip->pending_mmaps, &rxe->pending_mmaps);
spin_unlock_bh(&rxe->pending_lock);
}
*ip_p = ip;
return 0;
err2:
kfree(ip);
err1:
return err;
}
inline void rxe_queue_reset(struct rxe_queue *q)
{
/* queue is comprised from header and the memory
* of the actual queue. See "struct rxe_queue_buf" in rxe_queue.h
* reset only the queue itself and not the management header
*/
memset(q->buf->data, 0, q->buf_size - sizeof(struct rxe_queue_buf));
}
struct rxe_queue *rxe_queue_init(struct rxe_dev *rxe, int *num_elem,
unsigned int elem_size, enum queue_type type)
{
struct rxe_queue *q;
size_t buf_size;
unsigned int num_slots;
/* num_elem == 0 is allowed, but uninteresting */
if (*num_elem < 0)
goto err1;
q = kzalloc(sizeof(*q), GFP_KERNEL);
if (!q)
goto err1;
q->rxe = rxe;
q->type = type;
/* used in resize, only need to copy used part of queue */
q->elem_size = elem_size;
/* pad element up to at least a cacheline and always a power of 2 */
if (elem_size < cache_line_size())
elem_size = cache_line_size();
elem_size = roundup_pow_of_two(elem_size);
q->log2_elem_size = order_base_2(elem_size);
num_slots = *num_elem + 1;
num_slots = roundup_pow_of_two(num_slots);
q->index_mask = num_slots - 1;
buf_size = sizeof(struct rxe_queue_buf) + num_slots * elem_size;
q->buf = vmalloc_user(buf_size);
if (!q->buf)
goto err2;
q->buf->log2_elem_size = q->log2_elem_size;
q->buf->index_mask = q->index_mask;
q->buf_size = buf_size;
*num_elem = num_slots - 1;
return q;
err2:
kfree(q);
err1:
return NULL;
}
/* copies elements from original q to new q and then swaps the contents of the
* two q headers. This is so that if anyone is holding a pointer to q it will
* still work
*/
static int resize_finish(struct rxe_queue *q, struct rxe_queue *new_q,
unsigned int num_elem)
{
enum queue_type type = q->type;
u32 prod;
u32 cons;
if (!queue_empty(q, q->type) && (num_elem < queue_count(q, type)))
return -EINVAL;
prod = queue_get_producer(new_q, type);
cons = queue_get_consumer(q, type);
while (!queue_empty(q, type)) {
memcpy(queue_addr_from_index(new_q, prod),
queue_addr_from_index(q, cons), new_q->elem_size);
prod = queue_next_index(new_q, prod);
cons = queue_next_index(q, cons);
}
new_q->buf->producer_index = prod;
q->buf->consumer_index = cons;
/* update private index copies */
if (type == QUEUE_TYPE_TO_CLIENT)
new_q->index = new_q->buf->producer_index;
else
q->index = q->buf->consumer_index;
/* exchange rxe_queue headers */
swap(*q, *new_q);
return 0;
}
int rxe_queue_resize(struct rxe_queue *q, unsigned int *num_elem_p,
unsigned int elem_size, struct ib_udata *udata,
struct mminfo __user *outbuf, spinlock_t *producer_lock,
spinlock_t *consumer_lock)
{
struct rxe_queue *new_q;
unsigned int num_elem = *num_elem_p;
int err;
unsigned long flags = 0, flags1;
new_q = rxe_queue_init(q->rxe, &num_elem, elem_size, q->type);
if (!new_q)
return -ENOMEM;
err = do_mmap_info(new_q->rxe, outbuf, udata, new_q->buf,
new_q->buf_size, &new_q->ip);
if (err) {
vfree(new_q->buf);
kfree(new_q);
goto err1;
}
spin_lock_irqsave(consumer_lock, flags1);
if (producer_lock) {
spin_lock_irqsave(producer_lock, flags);
err = resize_finish(q, new_q, num_elem);
spin_unlock_irqrestore(producer_lock, flags);
} else {
err = resize_finish(q, new_q, num_elem);
}
spin_unlock_irqrestore(consumer_lock, flags1);
rxe_queue_cleanup(new_q); /* new/old dep on err */
if (err)
goto err1;
*num_elem_p = num_elem;
return 0;
err1:
return err;
}
void rxe_queue_cleanup(struct rxe_queue *q)
{
if (q->ip)
kref_put(&q->ip->ref, rxe_mmap_release);
else
vfree(q->buf);
kfree(q);
}