static int kcm_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
{
struct sk_buff_head *list = &sk->sk_receive_queue;
if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
return -ENOMEM;
if (!sk_rmem_schedule(sk, skb, skb->truesize))
return -ENOBUFS;
skb->dev = NULL;
skb_orphan(skb);
skb->sk = sk;
skb->destructor = kcm_rfree;
atomic_add(skb->truesize, &sk->sk_rmem_alloc);
sk_mem_charge(sk, skb->truesize);
skb_queue_tail(list, skb);
if (!sock_flag(sk, SOCK_DEAD))
sk->sk_data_ready(sk);
return 0;
}
static int alloc_sg(struct sock *sk, int len, struct scatterlist *sg,
int *sg_num_elem, unsigned int *sg_size,
int first_coalesce)
{
struct page_frag *pfrag;
unsigned int size = *sg_size;
int num_elem = *sg_num_elem, use = 0, rc = 0;
struct scatterlist *sge;
unsigned int orig_offset;
len -= size;
pfrag = sk_page_frag(sk);
while (len > 0) {
if (!sk_page_frag_refill(sk, pfrag)) {
rc = -ENOMEM;
goto out;
}
use = min_t(int, len, pfrag->size - pfrag->offset);
if (!sk_wmem_schedule(sk, use)) {
rc = -ENOMEM;
goto out;
}
sk_mem_charge(sk, use);
size += use;
orig_offset = pfrag->offset;
pfrag->offset += use;
sge = sg + num_elem - 1;
if (num_elem > first_coalesce && sg_page(sg) == pfrag->page &&
sg->offset + sg->length == orig_offset) {
sg->length += use;
} else {
sge++;
sg_unmark_end(sge);
sg_set_page(sge, pfrag->page, use, orig_offset);
get_page(pfrag->page);
++num_elem;
if (num_elem == MAX_SKB_FRAGS) {
rc = -ENOSPC;
break;
}
}
len -= use;
}
goto out;
out:
*sg_size = size;
*sg_num_elem = num_elem;
return rc;
}
static int zerocopy_from_iter(struct sock *sk, struct iov_iter *from,
int length)
{
struct tls_context *tls_ctx = tls_get_ctx(sk);
struct tls_sw_context *ctx = tls_sw_ctx(tls_ctx);
struct page *pages[MAX_SKB_FRAGS];
size_t offset;
ssize_t copied, use;
int i = 0;
unsigned int size = ctx->sg_plaintext_size;
int num_elem = ctx->sg_plaintext_num_elem;
int rc = 0;
int maxpages;
while (length > 0) {
i = 0;
maxpages = ARRAY_SIZE(ctx->sg_plaintext_data) - num_elem;
if (maxpages == 0) {
rc = -EFAULT;
goto out;
}
copied = iov_iter_get_pages(from, pages,
length,
maxpages, &offset);
if (copied <= 0) {
rc = -EFAULT;
goto out;
}
iov_iter_advance(from, copied);
length -= copied;
size += copied;
while (copied) {
use = min_t(int, copied, PAGE_SIZE - offset);
sg_set_page(&ctx->sg_plaintext_data[num_elem],
pages[i], use, offset);
sg_unmark_end(&ctx->sg_plaintext_data[num_elem]);
sk_mem_charge(sk, use);
offset = 0;
copied -= use;
++i;
++num_elem;
}
}
out:
ctx->sg_plaintext_size = size;
ctx->sg_plaintext_num_elem = num_elem;
return rc;
}
static void
ss_skb_entail(struct sock *sk, struct sk_buff *skb)
{
struct tcp_sock *tp = tcp_sk(sk);
struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
skb->csum = 0;
tcb->seq = tcb->end_seq = tp->write_seq;
tcb->tcp_flags = TCPHDR_ACK;
tcb->sacked = 0;
skb_header_release(skb);
tcp_add_write_queue_tail(sk, skb);
sk->sk_wmem_queued += skb->truesize;
sk_mem_charge(sk, skb->truesize);
if (tp->nonagle & TCP_NAGLE_PUSH)
tp->nonagle &= ~TCP_NAGLE_PUSH;
}
int tls_sw_sendpage(struct sock *sk, struct page *page,
int offset, size_t size, int flags)
{
struct tls_context *tls_ctx = tls_get_ctx(sk);
struct tls_sw_context *ctx = tls_sw_ctx(tls_ctx);
int ret = 0;
long timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
bool eor;
size_t orig_size = size;
unsigned char record_type = TLS_RECORD_TYPE_DATA;
struct scatterlist *sg;
bool full_record;
int record_room;
if (flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL |
MSG_SENDPAGE_NOTLAST))
return -ENOTSUPP;
/* No MSG_EOR from splice, only look at MSG_MORE */
eor = !(flags & (MSG_MORE | MSG_SENDPAGE_NOTLAST));
lock_sock(sk);
sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
if (tls_complete_pending_work(sk, tls_ctx, flags, &timeo))
goto sendpage_end;
/* Call the sk_stream functions to manage the sndbuf mem. */
while (size > 0) {
size_t copy, required_size;
if (sk->sk_err) {
ret = sk->sk_err;
goto sendpage_end;
}
full_record = false;
record_room = TLS_MAX_PAYLOAD_SIZE - ctx->sg_plaintext_size;
copy = size;
if (copy >= record_room) {
copy = record_room;
full_record = true;
}
required_size = ctx->sg_plaintext_size + copy +
tls_ctx->overhead_size;
if (!sk_stream_memory_free(sk))
goto wait_for_sndbuf;
alloc_payload:
ret = alloc_encrypted_sg(sk, required_size);
if (ret) {
if (ret != -ENOSPC)
goto wait_for_memory;
/* Adjust copy according to the amount that was
* actually allocated. The difference is due
* to max sg elements limit
*/
copy -= required_size - ctx->sg_plaintext_size;
full_record = true;
}
get_page(page);
sg = ctx->sg_plaintext_data + ctx->sg_plaintext_num_elem;
sg_set_page(sg, page, copy, offset);
ctx->sg_plaintext_num_elem++;
sk_mem_charge(sk, copy);
offset += copy;
size -= copy;
ctx->sg_plaintext_size += copy;
tls_ctx->pending_open_record_frags = ctx->sg_plaintext_num_elem;
if (full_record || eor ||
ctx->sg_plaintext_num_elem ==
ARRAY_SIZE(ctx->sg_plaintext_data)) {
push_record:
ret = tls_push_record(sk, flags, record_type);
if (ret) {
if (ret == -ENOMEM)
goto wait_for_memory;
goto sendpage_end;
}
}
continue;
wait_for_sndbuf:
set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
wait_for_memory:
ret = sk_stream_wait_memory(sk, &timeo);
if (ret) {
trim_both_sgl(sk, ctx->sg_plaintext_size);
goto sendpage_end;
}
if (tls_is_pending_closed_record(tls_ctx))
goto push_record;
goto alloc_payload;
}
sendpage_end:
if (orig_size > size)
ret = orig_size - size;
else
ret = sk_stream_error(sk, flags, ret);
release_sock(sk);
return ret;
}
void sdp_post_send(struct sdp_sock *ssk, struct sk_buff *skb)
{
struct sdp_buf *tx_req;
struct sdp_bsdh *h = (struct sdp_bsdh *)skb_transport_header(skb);
unsigned long mseq = ring_head(ssk->tx_ring);
int i, rc, frags;
u64 addr;
struct ib_device *dev;
struct ib_send_wr *bad_wr;
struct ib_sge ibsge[SDP_MAX_SEND_SGES];
struct ib_sge *sge = ibsge;
struct ib_send_wr tx_wr = { NULL };
u32 send_flags = IB_SEND_SIGNALED;
SDPSTATS_COUNTER_MID_INC(post_send, h->mid);
SDPSTATS_HIST(send_size, skb->len);
if (!ssk->qp_active)
goto err;
ssk->tx_packets++;
if (h->mid != SDP_MID_SRCAVAIL &&
h->mid != SDP_MID_DATA &&
h->mid != SDP_MID_SRCAVAIL_CANCEL) {
struct sock *sk = sk_ssk(ssk);
sk->sk_wmem_queued += skb->truesize;
sk_mem_charge(sk, skb->truesize);
}
if (unlikely(h->mid == SDP_MID_SRCAVAIL)) {
struct tx_srcavail_state *tx_sa = TX_SRCAVAIL_STATE(skb);
if (ssk->tx_sa != tx_sa) {
sdp_dbg_data(sk_ssk(ssk), "SrcAvail cancelled "
"before being sent!\n");
SDP_WARN_ON(1);
sk_wmem_free_skb(sk_ssk(ssk), skb);
return;
}
TX_SRCAVAIL_STATE(skb)->mseq = mseq;
}
if (unlikely(SDP_SKB_CB(skb)->flags & TCPHDR_URG))
h->flags = SDP_OOB_PRES | SDP_OOB_PEND;
else
h->flags = 0;
h->bufs = htons(rx_ring_posted(ssk));
h->len = htonl(skb->len);
h->mseq = htonl(mseq);
h->mseq_ack = htonl(mseq_ack(ssk));
sdp_prf(sk_ssk(ssk), skb, "TX: %s bufs: %d mseq:%ld ack:%d c: %d",
mid2str(h->mid), rx_ring_posted(ssk), mseq,
ntohl(h->mseq_ack), tx_credits(ssk));
SDP_DUMP_PACKET(sk_ssk(ssk), "TX", skb, h);
tx_req = &ssk->tx_ring.buffer[mseq & (SDP_TX_SIZE - 1)];
tx_req->skb = skb;
dev = ssk->ib_device;
if (skb->len <= ssk->inline_thresh && !skb_shinfo(skb)->nr_frags) {
SDPSTATS_COUNTER_INC(inline_sends);
sge->addr = (u64) skb->data;
sge->length = skb->len;
sge->lkey = 0;
frags = 0;
tx_req->mapping[0] = 0; /* Nothing to be cleaned up by sdp_cleanup_sdp_buf() */
send_flags |= IB_SEND_INLINE;
} else {
addr = ib_dma_map_single(dev, skb->data, skb->len - skb->data_len,
DMA_TO_DEVICE);
tx_req->mapping[0] = addr;
/* TODO: proper error handling */
BUG_ON(ib_dma_mapping_error(dev, addr));
sge->addr = addr;
sge->length = skb->len - skb->data_len;
sge->lkey = ssk->sdp_dev->mr->lkey;
frags = skb_shinfo(skb)->nr_frags;
for (i = 0; i < frags; ++i) {
++sge;
addr = ib_dma_map_page(dev,
skb_shinfo(skb)->frags[i].page.p,
skb_shinfo(skb)->frags[i].page_offset,
skb_shinfo(skb)->frags[i].size,
DMA_TO_DEVICE);
BUG_ON(ib_dma_mapping_error(dev, addr));
tx_req->mapping[i + 1] = addr;
sge->addr = addr;
sge->length = skb_shinfo(skb)->frags[i].size;
sge->lkey = ssk->sdp_dev->mr->lkey;
}
}
tx_wr.next = NULL;
tx_wr.wr_id = ring_head(ssk->tx_ring) | SDP_OP_SEND;
tx_wr.sg_list = ibsge;
tx_wr.num_sge = frags + 1;
tx_wr.opcode = IB_WR_SEND;
tx_wr.send_flags = send_flags;
if (unlikely(SDP_SKB_CB(skb)->flags & TCPHDR_URG))
tx_wr.send_flags |= IB_SEND_SOLICITED;
rc = ib_post_send(ssk->qp, &tx_wr, &bad_wr);
if (unlikely(rc)) {
sdp_dbg(sk_ssk(ssk),
"ib_post_send failed with status %d.\n", rc);
sdp_cleanup_sdp_buf(ssk, tx_req, skb->len - skb->data_len, DMA_TO_DEVICE);
sdp_set_error(sk_ssk(ssk), -ECONNRESET);
goto err;
}
atomic_inc(&ssk->tx_ring.head);
atomic_dec(&ssk->tx_ring.credits);
atomic_set(&ssk->remote_credits, rx_ring_posted(ssk));
return;
err:
sk_wmem_free_skb(sk_ssk(ssk), skb);
}
int sdp_post_sends(struct sdp_sock *ssk, gfp_t gfp)
{
/* TODO: nonagle? */
struct sk_buff *skb;
int post_count = 0;
struct sock *sk = sk_ssk(ssk);
if (unlikely(!ssk->id)) {
if (sk->sk_send_head) {
sdp_dbg(sk, "Send on socket without cmid ECONNRESET\n");
/* TODO: flush send queue? */
sdp_reset(sk);
}
return -ECONNRESET;
}
again:
if (sdp_tx_ring_slots_left(ssk) < SDP_TX_SIZE / 2)
sdp_xmit_poll(ssk, 1);
/* Run out of credits, check if got a credit update */
if (unlikely(tx_credits(ssk) <= SDP_MIN_TX_CREDITS)) {
sdp_poll_rx_cq(ssk);
if (unlikely(sdp_should_rearm(sk) || !posts_handler(ssk)))
sdp_arm_rx_cq(sk);
}
if (unlikely((ssk->sa_post_rdma_rd_compl || ssk->sa_post_sendsm) &&
tx_credits(ssk) < SDP_MIN_TX_CREDITS)) {
sdp_dbg_data(sk, "Run out of credits, can't abort SrcAvail. "
"RdmaRdCompl: %d SendSm: %d\n",
ssk->sa_post_rdma_rd_compl, ssk->sa_post_sendsm);
}
if (ssk->sa_post_rdma_rd_compl && tx_credits(ssk) >= SDP_MIN_TX_CREDITS) {
int unreported = ssk->sa_post_rdma_rd_compl;
skb = sdp_alloc_skb_rdmardcompl(sk, unreported, gfp);
if (!skb)
goto no_mem;
sdp_post_send(ssk, skb);
post_count++;
ssk->sa_post_rdma_rd_compl = 0;
}
if (ssk->sa_post_sendsm && tx_credits(ssk) >= SDP_MIN_TX_CREDITS) {
skb = sdp_alloc_skb_sendsm(sk, gfp);
if (unlikely(!skb))
goto no_mem;
sdp_post_send(ssk, skb);
ssk->sa_post_sendsm = 0;
post_count++;
}
if (ssk->recv_request &&
ring_tail(ssk->rx_ring) >= SDP_MIN_TX_CREDITS &&
tx_credits(ssk) >= SDP_MIN_TX_CREDITS &&
sdp_tx_ring_slots_left(ssk)) {
skb = sdp_alloc_skb_chrcvbuf_ack(sk,
ssk->recv_frags * PAGE_SIZE, gfp);
if (!skb)
goto no_mem;
ssk->recv_request = 0;
sdp_post_send(ssk, skb);
post_count++;
}
if (tx_credits(ssk) <= SDP_MIN_TX_CREDITS &&
sdp_tx_ring_slots_left(ssk) &&
sk->sk_send_head &&
sdp_nagle_off(ssk, sk->sk_send_head)) {
SDPSTATS_COUNTER_INC(send_miss_no_credits);
}
while (tx_credits(ssk) > SDP_MIN_TX_CREDITS &&
sdp_tx_ring_slots_left(ssk) &&
(skb = sk->sk_send_head) &&
sdp_nagle_off(ssk, skb)) {
update_send_head(sk, skb);
__skb_dequeue(&sk->sk_write_queue);
sdp_post_send(ssk, skb);
post_count++;
}
if (credit_update_needed(ssk) &&
likely((1 << sk->sk_state) &
(TCPF_ESTABLISHED | TCPF_FIN_WAIT1))) {
skb = sdp_alloc_skb_data(sk, 0, gfp);
if (!skb)
goto no_mem;
sk->sk_wmem_queued += skb->truesize;
sk_mem_charge(sk, skb->truesize);
sdp_post_send(ssk, skb);
SDPSTATS_COUNTER_INC(post_send_credits);
post_count++;
}
/* send DisConn if needed
* Do not send DisConn if there is only 1 credit. Compliance with CA4-82
* If one credit is available, an implementation shall only send SDP
* messages that provide additional credits and also do not contain ULP
* payload. */
if (unlikely(ssk->sdp_disconnect) &&
!sk->sk_send_head &&
tx_credits(ssk) >= SDP_MIN_TX_CREDITS) {
skb = sdp_alloc_skb_disconnect(sk, gfp);
if (!skb)
goto no_mem;
ssk->sdp_disconnect = 0;
sdp_post_send(ssk, skb);
post_count++;
}
if (!sdp_tx_ring_slots_left(ssk) || post_count) {
if (sdp_xmit_poll(ssk, 1))
goto again;
}
no_mem:
return post_count;
}
