Date: Thu, 29 Apr 2010 15:36:32 +0000 (UTC) From: Pawel Jakub Dawidek <pjd@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r207371 - head/sbin/hastd Message-ID: <201004291536.o3TFaW19085308@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: pjd Date: Thu Apr 29 15:36:32 2010 New Revision: 207371 URL: http://svn.freebsd.org/changeset/base/207371 Log: Fix a problem where hastd will stuck in recv(2) after sending request to secondary, which died between send(2) and recv(2). Do it by adding timeout to recv(2) for primary incoming and outgoing sockets and secondary outgoing socket. Reported by: Mikolaj Golub <to.my.trociny@gmail.com> Tested by: Mikolaj Golub <to.my.trociny@gmail.com> MFC after: 3 days Modified: head/sbin/hastd/hast.conf.5 head/sbin/hastd/hast.h head/sbin/hastd/hastd.c head/sbin/hastd/parse.y head/sbin/hastd/primary.c head/sbin/hastd/proto.c head/sbin/hastd/proto.h head/sbin/hastd/proto_common.c head/sbin/hastd/secondary.c head/sbin/hastd/token.l Modified: head/sbin/hastd/hast.conf.5 ============================================================================== --- head/sbin/hastd/hast.conf.5 Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/hast.conf.5 Thu Apr 29 15:36:32 2010 (r207371) @@ -58,6 +58,7 @@ file is following: control <addr> listen <addr> replication <mode> +timeout <seconds> on <node> { # Node section @@ -76,6 +77,7 @@ resource <name> { replication <mode> name <name> local <path> + timeout <seconds> on <node> { # Resource-node section @@ -194,6 +196,11 @@ The .Ic async replication mode is currently not implemented. .El +.It Ic timeout Aq seconds +.Pp +Connection timeout in seconds. +The default value is +.Va 5 . .It Ic name Aq name .Pp GEOM provider name that will appear as Modified: head/sbin/hastd/hast.h ============================================================================== --- head/sbin/hastd/hast.h Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/hast.h Thu Apr 29 15:36:32 2010 (r207371) @@ -75,6 +75,7 @@ #define HIO_DELETE 3 #define HIO_FLUSH 4 +#define HAST_TIMEOUT 5 #define HAST_CONFIG "/etc/hast.conf" #define HAST_CONTROL "/var/run/hastctl" #define HASTD_PORT 8457 @@ -148,6 +149,8 @@ struct hast_resource { /* Token to verify both in and out connection are coming from the same node (not necessarily from the same address). */ unsigned char hr_token[HAST_TOKEN_SIZE]; + /* Connection timeout. */ + int hr_timeout; /* Resource unique identifier. */ uint64_t hr_resuid; Modified: head/sbin/hastd/hastd.c ============================================================================== --- head/sbin/hastd/hastd.c Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/hastd.c Thu Apr 29 15:36:32 2010 (r207371) @@ -187,6 +187,10 @@ listen_accept(void) proto_remote_address(conn, raddr, sizeof(raddr)); pjdlog_info("Connection from %s to %s.", laddr, raddr); + /* Error in setting timeout is not critical, but why should it fail? */ + if (proto_timeout(conn, HAST_TIMEOUT) < 0) + pjdlog_errno(LOG_WARNING, "Unable to set connection timeout"); + nvin = nvout = nverr = NULL; /* Modified: head/sbin/hastd/parse.y ============================================================================== --- head/sbin/hastd/parse.y Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/parse.y Thu Apr 29 15:36:32 2010 (r207371) @@ -58,6 +58,7 @@ static bool mynode; static char depth0_control[HAST_ADDRSIZE]; static char depth0_listen[HAST_ADDRSIZE]; static int depth0_replication; +static int depth0_timeout; static char depth1_provname[PATH_MAX]; static char depth1_localpath[PATH_MAX]; @@ -115,6 +116,7 @@ yy_config_parse(const char *config) curres = NULL; mynode = false; + depth0_timeout = HAST_TIMEOUT; depth0_replication = HAST_REPLICATION_MEMSYNC; strlcpy(depth0_control, HAST_CONTROL, sizeof(depth0_control)); strlcpy(depth0_listen, HASTD_LISTEN, sizeof(depth0_listen)); @@ -154,6 +156,13 @@ yy_config_parse(const char *config) */ curres->hr_replication = depth0_replication; } + if (curres->hr_timeout == -1) { + /* + * Timeout is not set at resource-level. + * Use global or default setting. + */ + curres->hr_timeout = depth0_timeout; + } } return (&lconfig); @@ -171,7 +180,7 @@ yy_config_free(struct hastd_config *conf } %} -%token CONTROL LISTEN PORT REPLICATION EXTENTSIZE RESOURCE NAME LOCAL REMOTE ON +%token CONTROL LISTEN PORT REPLICATION TIMEOUT EXTENTSIZE RESOURCE NAME LOCAL REMOTE ON %token FULLSYNC MEMSYNC ASYNC %token NUM STR OB CB @@ -200,6 +209,8 @@ statement: | replication_statement | + timeout_statement + | node_statement | resource_statement @@ -281,6 +292,22 @@ replication_type: ASYNC { $$ = HAST_REPLICATION_ASYNC; } ; +timeout_statement: TIMEOUT NUM + { + switch (depth) { + case 0: + depth0_timeout = $2; + break; + case 1: + if (curres != NULL) + curres->hr_timeout = $2; + break; + default: + assert(!"timeout at wrong depth level"); + } + } + ; + node_statement: ON node_start OB node_entries CB { mynode = false; @@ -389,6 +416,7 @@ resource_start: STR curres->hr_role = HAST_ROLE_INIT; curres->hr_previous_role = HAST_ROLE_INIT; curres->hr_replication = -1; + curres->hr_timeout = -1; curres->hr_provname[0] = '\0'; curres->hr_localpath[0] = '\0'; curres->hr_localfd = -1; @@ -405,6 +433,8 @@ resource_entries: resource_entry: replication_statement | + timeout_statement + | name_statement | local_statement Modified: head/sbin/hastd/primary.c ============================================================================== --- head/sbin/hastd/primary.c Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/primary.c Thu Apr 29 15:36:32 2010 (r207371) @@ -489,6 +489,9 @@ init_remote(struct hast_resource *res, s res->hr_remoteaddr); goto close; } + /* Error in setting timeout is not critical, but why should it fail? */ + if (proto_timeout(out, res->hr_timeout) < 0) + pjdlog_errno(LOG_WARNING, "Unable to set connection timeout"); /* * First handshake step. * Setup outgoing connection with remote node. @@ -552,6 +555,9 @@ init_remote(struct hast_resource *res, s res->hr_remoteaddr); goto close; } + /* Error in setting timeout is not critical, but why should it fail? */ + if (proto_timeout(in, res->hr_timeout) < 0) + pjdlog_errno(LOG_WARNING, "Unable to set connection timeout"); nvout = nv_alloc(); nv_add_string(nvout, res->hr_name, "resource"); nv_add_uint8_array(nvout, res->hr_token, sizeof(res->hr_token), Modified: head/sbin/hastd/proto.c ============================================================================== --- head/sbin/hastd/proto.c Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/proto.c Thu Apr 29 15:36:32 2010 (r207371) @@ -30,7 +30,9 @@ #include <sys/cdefs.h> __FBSDID("$FreeBSD$"); +#include <sys/types.h> #include <sys/queue.h> +#include <sys/socket.h> #include <assert.h> #include <errno.h> @@ -247,6 +249,30 @@ proto_remote_address(const struct proto_ conn->pc_proto->hp_remote_address(conn->pc_ctx, addr, size); } +int +proto_timeout(const struct proto_conn *conn, int timeout) +{ + struct timeval tv; + int fd; + + assert(conn != NULL); + assert(conn->pc_magic == PROTO_CONN_MAGIC); + assert(conn->pc_proto != NULL); + + fd = proto_descriptor(conn); + if (fd < 0) + return (-1); + + tv.tv_sec = timeout; + tv.tv_usec = 0; + if (setsockopt(fd, SOL_SOCKET, SO_SNDTIMEO, &tv, sizeof(tv)) < 0) + return (-1); + if (setsockopt(fd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv)) < 0) + return (-1); + + return (0); +} + void proto_close(struct proto_conn *conn) { Modified: head/sbin/hastd/proto.h ============================================================================== --- head/sbin/hastd/proto.h Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/proto.h Thu Apr 29 15:36:32 2010 (r207371) @@ -49,6 +49,7 @@ void proto_local_address(const struct pr size_t size); void proto_remote_address(const struct proto_conn *conn, char *addr, size_t size); +int proto_timeout(const struct proto_conn *conn, int timeout); void proto_close(struct proto_conn *conn); #endif /* !_PROTO_H_ */ Modified: head/sbin/hastd/proto_common.c ============================================================================== --- head/sbin/hastd/proto_common.c Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/proto_common.c Thu Apr 29 15:36:32 2010 (r207371) @@ -58,7 +58,7 @@ proto_common_send(int fd, const unsigned if (done == 0) return (ENOTCONN); else if (done < 0) { - if (errno == EAGAIN) + if (errno == EINTR) continue; return (errno); } @@ -76,7 +76,7 @@ proto_common_recv(int fd, unsigned char do { done = recv(fd, data, size, MSG_WAITALL); - } while (done == -1 && errno == EAGAIN); + } while (done == -1 && errno == EINTR); if (done == 0) return (ENOTCONN); else if (done < 0) Modified: head/sbin/hastd/secondary.c ============================================================================== --- head/sbin/hastd/secondary.c Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/secondary.c Thu Apr 29 15:36:32 2010 (r207371) @@ -337,6 +337,12 @@ hastd_secondary(struct hast_resource *re setproctitle("%s (secondary)", res->hr_name); + /* Error in setting timeout is not critical, but why should it fail? */ + if (proto_timeout(res->hr_remotein, 0) < 0) + pjdlog_errno(LOG_WARNING, "Unable to set connection timeout"); + if (proto_timeout(res->hr_remoteout, res->hr_timeout) < 0) + pjdlog_errno(LOG_WARNING, "Unable to set connection timeout"); + init_local(res); init_remote(res, nvin); init_environment(); Modified: head/sbin/hastd/token.l ============================================================================== --- head/sbin/hastd/token.l Thu Apr 29 15:19:11 2010 (r207370) +++ head/sbin/hastd/token.l Thu Apr 29 15:36:32 2010 (r207371) @@ -48,6 +48,7 @@ control { DP; return CONTROL; } listen { DP; return LISTEN; } port { DP; return PORT; } replication { DP; return REPLICATION; } +timeout { DP; return TIMEOUT; } resource { DP; return RESOURCE; } name { DP; return NAME; } local { DP; return LOCAL; }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201004291536.o3TFaW19085308>