diff options
Diffstat (limited to 'lib/libtle_glue/select.c')
| -rw-r--r-- | lib/libtle_glue/select.c | 201 |
1 files changed, 201 insertions, 0 deletions
diff --git a/lib/libtle_glue/select.c b/lib/libtle_glue/select.c new file mode 100644 index 0000000..b3b8539 --- /dev/null +++ b/lib/libtle_glue/select.c @@ -0,0 +1,201 @@ +/* + * Copyright (c) 2018 Ant Financial Services Group. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#ifndef _GNU_SOURCE +#define _GNU_SOURCE +#endif +#include <signal.h> +#include <sys/select.h> +#include <sys/time.h> +#include <sys/types.h> +#include <unistd.h> + +#include "fd.h" +#include "ctx.h" +#include "sym.h" +#include "log.h" +#include "util.h" +#include "internal.h" +#include "tle_glue.h" + +#define FD_ZERO_N(s, n) do { memset((s)->fds_bits, 0, n/sizeof(long)); } while(0) + +static int +fdset_to_events_user(int nfds, fd_set *fdset, int *total, int event) +{ + int i, num = 0; + struct sock *so; + const struct tle_event *ev; + + for (i = fd_table.fd_base; i < nfds; ++i) { + if (!FD_ISSET(i, fdset)) + continue; + + so = fd2sock(i); /* fix me: check if fd is opened */ + + switch (event) { + case EPOLLIN: + ev = &so->rxev; + break; + case EPOLLOUT: + ev = &so->txev; + break; + case EPOLLERR: + ev = &so->erev; + break; + default: + rte_panic("non-sense value\n"); + } + /* Check event is ready */ + if (TLE_SEV_UP == tle_event_state(ev)) { + *total = *total + 1; + } else { + FD_CLR(i, fdset); + num++; + } + + /* We fill sock->event here as we need this when + * we filter events in poll_common(). But it was + * originally set by epoll_ctl(). Now we have to + * assume that there are no application which + * uses epoll/poll/select at the same time. + */ + so->event.events |= event; + so->event.data.u32 = i; + } + + return num; +} + +static int +fdset_to_events_kernel(int nfds, fd_set *fdset, int efd, int event) +{ + int i, num = 0; + struct epoll_event k_ev; + + for (i = 0; i < nfds; ++i) { + if (!FD_ISSET(i, fdset)) + continue; + + k_ev.events = event; + k_ev.data.u32 = i; + k_epoll_ctl(efd, EPOLL_CTL_ADD, i, &k_ev); + num++; + } + + return num; +} + +int +PRE(select)(int nfds, fd_set *readfds, fd_set *writefds, + fd_set *exceptfds, struct timeval *timeout) +{ + int to; + struct glue_ctx *ctx; + int j, efd, total = 0, max = 0; + + /* thread <> context binding happens here */ + if (RTE_PER_LCORE(glue_ctx) == NULL) { + ctx = &ctx_array[glue_ctx_alloc()]; + RTE_PER_LCORE(glue_ctx) = ctx; + } else + ctx = RTE_PER_LCORE(glue_ctx); + + /* step 0, process some packets */ + be_process(ctx); + + /* step 1, check if any userspace events are ready */ + + if (readfds) + max += fdset_to_events_user(nfds, readfds, + &total, EPOLLIN); + if (writefds) + max += fdset_to_events_user(nfds, writefds, + &total, EPOLLOUT); + if (exceptfds) + max += fdset_to_events_user(nfds, writefds, + &total, EPOLLERR); + if (total > 0) { + /* userspace events go firstly */ + if (readfds) + FD_ZERO_N(readfds, fd_table.fd_base); + if (writefds) + FD_ZERO_N(writefds, fd_table.fd_base); + if (exceptfds) + FD_ZERO_N(exceptfds, fd_table.fd_base); + + return total; + } + + /* step 2, only wait for kernel events? */ + if (max == 0) + return k_select(nfds, readfds, writefds, exceptfds, timeout); + + /* step 3, slow path: wait for I/O and kernel events */ + efd = k_epoll_create(1); + if (efd < 0) + rte_panic("k_epoll_create failed %d", errno); + + nfds = RTE_MIN(nfds, fd_table.fd_base); + if (readfds) + max += fdset_to_events_kernel(nfds, readfds, + efd, EPOLLIN); + if (writefds) + max += fdset_to_events_kernel(nfds, writefds, + efd, EPOLLOUT); + if (exceptfds) + max += fdset_to_events_kernel(nfds, exceptfds, + efd, EPOLLERR); + + struct epoll_event events[max]; + + if (timeout) + to = timeout->tv_sec * 1000 + timeout->tv_usec / 1000; + else + to = -1; + total = poll_common(ctx, events, max, to, efd); + + k_close(efd); + for (j = 0; j < total; ++j) { + if (events[j].events & EPOLLIN) + FD_SET(events[j].data.fd, readfds); + + if (events[j].events & EPOLLOUT) + FD_SET(events[j].data.fd, writefds); + + if ((events[j].events & (EPOLLHUP | EPOLLERR)) && exceptfds) + FD_SET(events[j].data.fd, exceptfds); + } + return total; +} + +int +PRE(pselect)(int nfds, fd_set *readfds, fd_set *writefds, fd_set *exceptfds, + const struct timespec *timeout, const sigset_t *sigmask) +{ + struct timeval tv, *tv_to; + + if (sigmask != NULL) + rte_panic("pselect with signal is not supported"); + + if (timeout) { + tv.tv_usec = timeout->tv_nsec / 1000; + tv.tv_sec = timeout->tv_sec; + tv_to = &tv; + } else + tv_to = NULL; + + return select(nfds, readfds, writefds, exceptfds, tv_to); +} |