[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [PATCH] Switch to poll() in cxenstored's IO loop
Poll() can support more file descriptors than select(). We've done this for xenconsoled, now do this for cxenstored as well. The code is taken from xenconsoled and modified to adapt to cxenstored. Signed-off-by: Wei Liu <wei.liu2@xxxxxxxxxx> --- tools/xenstore/xenstored_core.c | 140 ++++++++++++++++++++++++++------------- tools/xenstore/xenstored_core.h | 2 + 2 files changed, 96 insertions(+), 46 deletions(-) diff --git a/tools/xenstore/xenstored_core.c b/tools/xenstore/xenstored_core.c index bd44645..7079c5e 100644 --- a/tools/xenstore/xenstored_core.c +++ b/tools/xenstore/xenstored_core.c @@ -19,7 +19,7 @@ #include <sys/types.h> #include <sys/stat.h> -#include <sys/select.h> +#include <poll.h> #ifndef NO_SOCKETS #include <sys/socket.h> #include <sys/un.h> @@ -55,6 +55,12 @@ #include "hashtable.h" extern xc_evtchn *xce_handle; /* in xenstored_domain.c */ +static struct pollfd *xce_pollfd; +static struct pollfd *fds; +static unsigned int current_array_size; +static unsigned int nr_fds; + +#define ROUNDUP(_x, _w) (((unsigned long)(_x)+(1UL<<(_w))-1) & ~((1UL<<(_w))-1)) static bool verbose = false; LIST_HEAD(connections); @@ -62,6 +68,7 @@ static int tracefd = -1; static bool recovery = true; static bool remove_local = true; static int reopen_log_pipe[2]; +static struct pollfd *reopen_log_pipe0_pollfd; static char *tracefile = NULL; static TDB_CONTEXT *tdb_ctx = NULL; @@ -199,7 +206,7 @@ void trace_destroy(const void *data, const char *type) /** * Signal handler for SIGHUP, which requests that the trace log is reopened * (in the main loop). A single byte is written to reopen_log_pipe, to awaken - * the select() in the main loop. + * the poll() in the main loop. */ static void trigger_reopen_log(int signal __attribute__((unused))) { @@ -279,15 +286,12 @@ static int destroy_conn(void *_conn) /* Flush outgoing if possible, but don't block. */ if (!conn->domain) { - fd_set set; - struct timeval none; - - FD_ZERO(&set); - FD_SET(conn->fd, &set); - none.tv_sec = none.tv_usec = 0; + struct pollfd pfd; + pfd.fd = conn->fd; + pfd.events = POLLOUT; while (!list_empty(&conn->out_list) - && select(conn->fd+1, NULL, &set, NULL, &none) == 1) + && poll(&pfd, 1, 0) == 1) if (!write_messages(conn)) break; close(conn->fd); @@ -300,52 +304,74 @@ static int destroy_conn(void *_conn) } -static void set_fd(int fd, fd_set *set, int *max) +static struct pollfd *set_fd(int fd, short events) { - if (fd < 0) - return; - FD_SET(fd, set); - if (fd > *max) - *max = fd; -} + struct pollfd *ret; + if (current_array_size < nr_fds + 1) { + struct pollfd *new_fds = NULL; + unsigned long newsize; + + /* Round up to 2^8 boundary, in practice this just + * make newsize larger than current_array_size. + */ + newsize = ROUNDUP(nr_fds + 1, 8); + + new_fds = realloc(fds, sizeof(struct pollfd)*newsize); + if (!new_fds) + goto fail; + fds = new_fds; + + memset(&fds[0] + current_array_size, 0, + sizeof(struct pollfd ) * (newsize-current_array_size)); + current_array_size = newsize; + } + + fds[nr_fds].fd = fd; + fds[nr_fds].events = events; + ret = &fds[nr_fds]; + nr_fds++; + return ret; +fail: + syslog(LOG_ERR, "realloc failed, ignoring fd %d\n", fd); + return NULL; +} -static int initialize_set(fd_set *inset, fd_set *outset, int sock, int ro_sock, - struct timeval **ptimeout) +static void initialize_fds(int sock, struct pollfd **p_sock_pollfd, + int ro_sock, struct pollfd **p_ro_sock_pollfd, + int *ptimeout) { - static struct timeval zero_timeout = { 0 }; struct connection *conn; - int max = -1; - *ptimeout = NULL; + memset(fds, 0, sizeof(struct pollfd) * current_array_size); + nr_fds = 0; - FD_ZERO(inset); - FD_ZERO(outset); + *ptimeout = -1; if (sock != -1) - set_fd(sock, inset, &max); + *p_sock_pollfd = set_fd(sock, POLLIN|POLLPRI); if (ro_sock != -1) - set_fd(ro_sock, inset, &max); + *p_ro_sock_pollfd = set_fd(ro_sock, POLLIN|POLLPRI); if (reopen_log_pipe[0] != -1) - set_fd(reopen_log_pipe[0], inset, &max); + reopen_log_pipe0_pollfd = + set_fd(reopen_log_pipe[0], POLLIN|POLLPRI); if (xce_handle != NULL) - set_fd(xc_evtchn_fd(xce_handle), inset, &max); + xce_pollfd = set_fd(xc_evtchn_fd(xce_handle), POLLIN|POLLPRI); list_for_each_entry(conn, &connections, list) { if (conn->domain) { if (domain_can_read(conn) || (domain_can_write(conn) && !list_empty(&conn->out_list))) - *ptimeout = &zero_timeout; + *ptimeout = 0; } else { - set_fd(conn->fd, inset, &max); + short events = POLLIN|POLLPRI; if (!list_empty(&conn->out_list)) - FD_SET(conn->fd, outset); + events |= POLLOUT; + conn->pollfd = set_fd(conn->fd, events); } } - - return max; } /* Is child a subnode of parent, or equal? */ @@ -1770,14 +1796,14 @@ int priv_domid = 0; int main(int argc, char *argv[]) { - int opt, *sock, *ro_sock, max; - fd_set inset, outset; + int opt, *sock, *ro_sock; + struct pollfd *sock_pollfd = NULL, *ro_sock_pollfd = NULL; bool dofork = true; bool outputpid = false; bool no_domain_init = false; const char *pidfile = NULL; int evtchn_fd = -1; - struct timeval *timeout; + int timeout; while ((opt = getopt_long(argc, argv, "DE:F:HNPS:t:T:RLVW:", options, NULL)) != -1) { @@ -1884,7 +1910,8 @@ int main(int argc, char *argv[]) evtchn_fd = xc_evtchn_fd(xce_handle); /* Get ready to listen to the tools. */ - max = initialize_set(&inset, &outset, *sock, *ro_sock, &timeout); + initialize_fds(*sock, &sock_pollfd, *ro_sock, &ro_sock_pollfd, + &timeout); /* Tell the kernel we're up and running. */ xenbus_notify_running(); @@ -1893,27 +1920,42 @@ int main(int argc, char *argv[]) for (;;) { struct connection *conn, *next; - if (select(max+1, &inset, &outset, NULL, timeout) < 0) { + if (poll(fds, nr_fds, timeout) < 0) { if (errno == EINTR) continue; - barf_perror("Select failed"); + barf_perror("Poll failed"); } - if (reopen_log_pipe[0] != -1 && FD_ISSET(reopen_log_pipe[0], &inset)) { + if (reopen_log_pipe[0] != -1 && reopen_log_pipe0_pollfd && + !(reopen_log_pipe0_pollfd->revents & ~(POLLIN|POLLOUT|POLLPRI)) && + (reopen_log_pipe0_pollfd->revents & POLLIN)) { char c; if (read(reopen_log_pipe[0], &c, 1) != 1) barf_perror("read failed"); reopen_log(); + reopen_log_pipe0_pollfd = NULL; } - if (*sock != -1 && FD_ISSET(*sock, &inset)) + if (*sock != -1 && sock_pollfd && + !(sock_pollfd->revents & ~(POLLIN|POLLOUT|POLLPRI)) && + (sock_pollfd->revents & POLLIN)) { accept_connection(*sock, true); + sock_pollfd = NULL; + } - if (*ro_sock != -1 && FD_ISSET(*ro_sock, &inset)) + if (*ro_sock != -1 && ro_sock_pollfd && + !(ro_sock_pollfd->revents & ~(POLLIN|POLLOUT|POLLPRI)) && + (ro_sock_pollfd->revents & POLLIN)) { accept_connection(*ro_sock, false); + ro_sock_pollfd = NULL; + } - if (evtchn_fd != -1 && FD_ISSET(evtchn_fd, &inset)) + if (evtchn_fd != -1 && xce_pollfd && + !(xce_pollfd->revents & ~(POLLIN|POLLOUT|POLLPRI)) && + (xce_pollfd->revents & POLLIN)) { handle_event(); + xce_pollfd = NULL; + } next = list_entry(connections.next, typeof(*conn), list); if (&next->list != &connections) @@ -1939,21 +1981,27 @@ int main(int argc, char *argv[]) if (talloc_free(conn) == 0) continue; } else { - if (FD_ISSET(conn->fd, &inset)) + if (conn->pollfd && + !(conn->pollfd->revents & ~(POLLIN|POLLOUT|POLLPRI)) && + (conn->pollfd->revents & POLLIN)) handle_input(conn); if (talloc_free(conn) == 0) continue; talloc_increase_ref_count(conn); - if (FD_ISSET(conn->fd, &outset)) + if (conn->pollfd && + !(conn->pollfd->revents & ~(POLLIN|POLLOUT|POLLPRI)) && + (conn->pollfd->revents & POLLOUT)) handle_output(conn); if (talloc_free(conn) == 0) continue; + + conn->pollfd = NULL; } } - max = initialize_set(&inset, &outset, *sock, *ro_sock, - &timeout); + initialize_fds(*sock, &sock_pollfd, *ro_sock, &ro_sock_pollfd, + &timeout); } } diff --git a/tools/xenstore/xenstored_core.h b/tools/xenstore/xenstored_core.h index 492ca0d..f330a87 100644 --- a/tools/xenstore/xenstored_core.h +++ b/tools/xenstore/xenstored_core.h @@ -60,6 +60,8 @@ struct connection /* The file descriptor we came in on. */ int fd; + /* The pollfd corresponding to fd. */ + struct pollfd *pollfd; /* Who am I? 0 for socket connections. */ unsigned int id; -- 1.7.10.4 _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxx http://lists.xen.org/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |