thebestvibedev/tromcho.net
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Allow to send more than 1 operation at a time

Browse Source
This commit is contained in:
Vitaliy Filippov
2023-02-26 02:01:04 +03:00
parent 138ffe4032
commit c3e80abad7
4 changed files with 49 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/msgr_rdma.cpp
+44 -26
View File
@@ -368,9 +368,8 @@ static void try_send_rdma_wr(osd_client_t *cl, ibv_sge *sge, int op_sge)
bool osd_messenger_t::try_send_rdma(osd_client_t *cl)
{
auto rc = cl->rdma_conn;
if (!cl->send_list.size() || rc->cur_send > 0)
if (!cl->send_list.size() || rc->cur_send >= rc->max_send)
{
// Only send one batch at a time
return true;
}
uint64_t op_size = 0, op_sge = 0;
@@ -380,6 +379,7 @@ bool osd_messenger_t::try_send_rdma(osd_client_t *cl)
iovec & iov = cl->send_list[rc->send_pos];
if (op_size >= rc->max_msg || op_sge >= rc->max_sge)
{
rc->send_sizes.push_back(op_size);
try_send_rdma_wr(cl, sge, op_sge);
op_sge = 0;
op_size = 0;
@@ -405,6 +405,7 @@ bool osd_messenger_t::try_send_rdma(osd_client_t *cl)
}
if (op_sge > 0)
{
rc->send_sizes.push_back(op_size);
try_send_rdma_wr(cl, sge, op_sge);
}
return true;
@@ -476,6 +477,7 @@ void osd_messenger_t::handle_rdma_events()
continue;
}
osd_client_t *cl = cl_it->second;
auto rc = cl->rdma_conn;
if (wc[i].status != IBV_WC_SUCCESS)
{
fprintf(stderr, "RDMA work request failed for client %d", client_id);
@@ -489,43 +491,59 @@ void osd_messenger_t::handle_rdma_events()
}
if (!is_send)
{
cl->rdma_conn->cur_recv--;
if (!handle_read_buffer(cl, cl->rdma_conn->recv_buffers[cl->rdma_conn->next_recv_buf], wc[i].byte_len))
rc->cur_recv--;
if (!handle_read_buffer(cl, rc->recv_buffers[rc->next_recv_buf], wc[i].byte_len))
{
// handle_read_buffer may stop the client
continue;
}
try_recv_rdma_wr(cl, cl->rdma_conn->recv_buffers[cl->rdma_conn->next_recv_buf]);
cl->rdma_conn->next_recv_buf = (cl->rdma_conn->next_recv_buf+1) % cl->rdma_conn->recv_buffers.size();
try_recv_rdma_wr(cl, rc->recv_buffers[rc->next_recv_buf]);
rc->next_recv_buf = (rc->next_recv_buf+1) % rc->recv_buffers.size();
}
else
{
cl->rdma_conn->cur_send--;
if (!cl->rdma_conn->cur_send)
rc->cur_send--;
uint64_t sent_size = rc->send_sizes.at(0);
rc->send_sizes.erase(rc->send_sizes.begin(), rc->send_sizes.begin()+1);
int send_pos = 0, send_buf_pos = 0;
while (sent_size > 0)
{
// Wait for the whole batch
for (int i = 0; i < cl->rdma_conn->send_pos; i++)
if (sent_size >= cl->send_list.at(send_pos).iov_len)
{
if (cl->outbox[i].flags & MSGR_SENDP_FREE)
{
// Reply fully sent
delete cl->outbox[i].op;
}
sent_size -= cl->send_list[send_pos].iov_len;
send_pos++;
}
if (cl->rdma_conn->send_pos > 0)
else
{
cl->send_list.erase(cl->send_list.begin(), cl->send_list.begin()+cl->rdma_conn->send_pos);
cl->outbox.erase(cl->outbox.begin(), cl->outbox.begin()+cl->rdma_conn->send_pos);
cl->rdma_conn->send_pos = 0;
send_buf_pos = sent_size;
sent_size = 0;
}
if (cl->rdma_conn->send_buf_pos > 0)
{
cl->send_list[0].iov_base = (uint8_t*)cl->send_list[0].iov_base + cl->rdma_conn->send_buf_pos;
cl->send_list[0].iov_len -= cl->rdma_conn->send_buf_pos;
cl->rdma_conn->send_buf_pos = 0;
}
try_send_rdma(cl);
}
assert(rc->send_pos >= send_pos);
if (rc->send_pos == send_pos)
{
rc->send_buf_pos -= send_buf_pos;
}
rc->send_pos -= send_pos;
for (int i = 0; i < send_pos; i++)
{
if (cl->outbox[i].flags & MSGR_SENDP_FREE)
{
// Reply fully sent
delete cl->outbox[i].op;
}
}
if (send_pos > 0)
{
cl->send_list.erase(cl->send_list.begin(), cl->send_list.begin()+send_pos);
cl->outbox.erase(cl->outbox.begin(), cl->outbox.begin()+send_pos);
}
if (send_buf_pos > 0)
{
cl->send_list[0].iov_base = (uint8_t*)cl->send_list[0].iov_base + send_buf_pos;
cl->send_list[0].iov_len -= send_buf_pos;
}
try_send_rdma(cl);
}
}
} while (event_count > 0);