>From ce294a5f179d743ea60281b34499fc334a258525 Mon Sep 17 00:00:00 2001 From: Jakub Hrozek Date: Mon, 5 Mar 2012 21:09:37 +0100 Subject: [PATCH 3/5] Only do one cycle when resolving a server https://fedorahosted.org/sssd/ticket/1214 --- src/providers/data_provider_fo.c | 71 +++++++++++++++++++++++----- src/providers/dp_backend.h | 5 ++ src/providers/fail_over.c | 6 +++ src/providers/fail_over.h | 3 ++ src/providers/krb5/krb5_auth.c | 19 +++++--- src/providers/ldap/ldap_auth.c | 5 +- src/providers/ldap/sdap_async_connection.c | 13 +++-- 7 files changed, 93 insertions(+), 29 deletions(-) diff --git a/src/providers/data_provider_fo.c b/src/providers/data_provider_fo.c index c3f8c3ba72a8696d6bffef08f0b88a3616843bd0..5d634cab948fdf7a8805653c7aa29cadba36e5d2 100644 --- a/src/providers/data_provider_fo.c +++ b/src/providers/data_provider_fo.c @@ -46,6 +46,7 @@ struct be_svc_data { bool run_callbacks; struct be_svc_callback *callbacks; + struct fo_server *first_resolved; }; struct be_failover_ctx { @@ -408,31 +409,31 @@ static void be_resolve_server_done(struct tevent_req *subreq) switch (ret) { case EOK: if (!state->srv) { - tevent_req_error(req, EFAULT); - return; + ret = EFAULT; + goto fail; } break; case ENOENT: /* all servers have been tried and none * was found good, go offline */ - tevent_req_error(req, EIO); - return; + ret = EIO; + goto fail; default: /* mark server as bad and retry */ if (!state->srv) { - tevent_req_error(req, EFAULT); - return; + ret = EFAULT; + goto fail; } DEBUG(6, ("Couldn't resolve server (%s), resolver returned (%d)\n", - fo_get_server_str_name(state->srv), ret)); + fo_get_server_str_name(state->srv), ret)); state->attempts++; if (state->attempts >= 10) { DEBUG(2, ("Failed to find a server after 10 attempts\n")); - tevent_req_error(req, EIO); - return; + ret = EIO; + goto fail; } /* now try next one */ @@ -442,8 +443,8 @@ static void be_resolve_server_done(struct tevent_req *subreq) state->ctx->be_fo->fo_ctx, state->svc->fo_service); if (!subreq) { - tevent_req_error(req, ENOMEM); - return; + ret = ENOMEM; + goto fail; } tevent_req_set_callback(subreq, be_resolve_server_done, req); @@ -451,6 +452,14 @@ static void be_resolve_server_done(struct tevent_req *subreq) } /* all fine we got the server */ + if (state->svc->first_resolved == NULL) { + DEBUG(7, ("Saving the first resolved server\n")); + state->svc->first_resolved = state->srv; + } else if (state->svc->first_resolved == state->srv) { + DEBUG(2, ("The fail over cycled through all available servers\n")); + ret = ENOENT; + goto fail; + } if (debug_level >= 4 && fo_get_server_name(state->srv)) { struct resolv_hostent *srvaddr; @@ -459,8 +468,8 @@ static void be_resolve_server_done(struct tevent_req *subreq) if (!srvaddr) { DEBUG(3, ("FATAL: No hostent available for server (%s)\n", fo_get_server_str_name(state->srv))); - tevent_req_error(req, EFAULT); - return; + ret = EFAULT; + goto fail; } inet_ntop(srvaddr->family, srvaddr->addr_list[0]->ipaddr, @@ -488,6 +497,12 @@ static void be_resolve_server_done(struct tevent_req *subreq) } tevent_req_done(req); + return; + +fail: + DEBUG(7, ("Server resolution failed: %d\n", ret)); + state->svc->first_resolved = NULL; + tevent_req_error(req, ret); } int be_resolve_server_recv(struct tevent_req *req, struct fo_server **srv) @@ -534,3 +549,33 @@ void reset_fo(struct be_ctx *be_ctx) fo_reset_services(be_ctx->be_fo->fo_ctx); } +void be_fo_set_port_status(struct be_ctx *ctx, + struct fo_server *server, + enum port_status status) +{ + struct be_svc_data *svc; + struct fo_service *fsvc; + + fo_set_port_status(server, status); + + fsvc = fo_get_server_service(server); + if (!fsvc) { + DEBUG(3, ("BUG: No service associated with server\n")); + return; + } + + DLIST_FOR_EACH(svc, ctx->be_fo->svcs) { + if (svc->fo_service == fsvc) break; + } + + if (!svc) { + DEBUG(3, ("BUG: Unknown service\n")); + return; + } + + if (status == PORT_WORKING) { + /* We were successful in connecting to the server. Cycle through all + * available servers next time */ + svc->first_resolved = NULL; + } +} diff --git a/src/providers/dp_backend.h b/src/providers/dp_backend.h index 3d5e40bae43b326c00ce9464abfca1e5918ced01..5d75287ea20a7bea04de0ac2f44aa60701c92553 100644 --- a/src/providers/dp_backend.h +++ b/src/providers/dp_backend.h @@ -185,6 +185,11 @@ struct tevent_req *be_resolve_server_send(TALLOC_CTX *memctx, struct be_ctx *ctx, const char *service_name); int be_resolve_server_recv(struct tevent_req *req, struct fo_server **srv); + +void be_fo_set_port_status(struct be_ctx *ctx, + struct fo_server *server, + enum port_status status); + /* * Instruct fail-over to try next server on the next connect attempt. * Should be used after connection to service was unexpectedly dropped diff --git a/src/providers/fail_over.c b/src/providers/fail_over.c index 3c46d8dd233b55306b64a84a6a16335b0c384d1e..2b47e55cbd4e5c703ec7c741625bbf6b690bec9f 100644 --- a/src/providers/fail_over.c +++ b/src/providers/fail_over.c @@ -1411,3 +1411,9 @@ void fo_reset_services(struct fo_ctx *fo_ctx) } } +struct fo_service * +fo_get_server_service(struct fo_server *server) +{ + if (!server) return NULL; + return server->service; +} diff --git a/src/providers/fail_over.h b/src/providers/fail_over.h index 3ca6096c45c4b73706f07b593448734a69272678..50c0dcf8d48df2bb6a3534e99c3de980f873f817 100644 --- a/src/providers/fail_over.h +++ b/src/providers/fail_over.h @@ -178,4 +178,7 @@ time_t fo_get_server_hostname_last_change(struct fo_server *server); int fo_is_srv_lookup(struct fo_server *s); void fo_reset_services(struct fo_ctx *fo_ctx); + +struct fo_service *fo_get_server_service(struct fo_server *server); + #endif /* !__FAIL_OVER_H__ */ diff --git a/src/providers/krb5/krb5_auth.c b/src/providers/krb5/krb5_auth.c index 7b72660adaaff35d20bfa5482a3d20a5774fd6cf..f2777f2f9f696dd82545c02f8c74ae1e57f15fbb 100644 --- a/src/providers/krb5/krb5_auth.c +++ b/src/providers/krb5/krb5_auth.c @@ -876,14 +876,16 @@ static void krb5_child_done(struct tevent_req *subreq) if (kr->kpasswd_srv != NULL) { /* ..which is unreachable by now.. */ if (msg_status == PAM_AUTHTOK_LOCK_BUSY) { - fo_set_port_status(kr->kpasswd_srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be_ctx, + kr->kpasswd_srv, PORT_NOT_WORKING); /* ..try to resolve next kpasswd server */ if (krb5_next_kpasswd(req) == NULL) { tevent_req_error(req, ENOMEM); } return; } else { - fo_set_port_status(kr->kpasswd_srv, PORT_WORKING); + be_fo_set_port_status(state->be_ctx, + kr->kpasswd_srv, PORT_WORKING); } } @@ -893,7 +895,7 @@ static void krb5_child_done(struct tevent_req *subreq) if (msg_status == PAM_AUTHINFO_UNAVAIL || (kr->kpasswd_srv == NULL && msg_status == PAM_AUTHTOK_LOCK_BUSY)) { if (kr->srv != NULL) { - fo_set_port_status(kr->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be_ctx, kr->srv, PORT_NOT_WORKING); /* ..try to resolve next KDC */ if (krb5_next_kdc(req) == NULL) { tevent_req_error(req, ENOMEM); @@ -901,7 +903,7 @@ static void krb5_child_done(struct tevent_req *subreq) return; } } else if (kr->srv != NULL) { - fo_set_port_status(kr->srv, PORT_WORKING); + be_fo_set_port_status(state->be_ctx, kr->srv, PORT_WORKING); } /* Now only a successful authentication or password change is left. @@ -972,17 +974,20 @@ static struct tevent_req *krb5_next_server(struct tevent_req *req) switch (pd->cmd) { case SSS_PAM_AUTHENTICATE: case SSS_CMD_RENEW: - fo_set_port_status(state->kr->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be_ctx, + state->kr->srv, PORT_NOT_WORKING); next_req = krb5_next_kdc(req); break; case SSS_PAM_CHAUTHTOK: case SSS_PAM_CHAUTHTOK_PRELIM: if (state->kr->kpasswd_srv) { - fo_set_port_status(state->kr->kpasswd_srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be_ctx, + state->kr->kpasswd_srv, PORT_NOT_WORKING); next_req = krb5_next_kpasswd(req); break; } else { - fo_set_port_status(state->kr->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be_ctx, + state->kr->srv, PORT_NOT_WORKING); next_req = krb5_next_kdc(req); break; } diff --git a/src/providers/ldap/ldap_auth.c b/src/providers/ldap/ldap_auth.c index f01c23d6e5de3d55d64f858803054d930e2577f5..a1b0acd1071809fdb5388dd25d5dd476ab7ba0ca 100644 --- a/src/providers/ldap/ldap_auth.c +++ b/src/providers/ldap/ldap_auth.c @@ -588,7 +588,8 @@ static void auth_connect_done(struct tevent_req *subreq) if (ret) { if (state->srv) { /* mark this server as bad if connection failed */ - fo_set_port_status(state->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->ctx->be, + state->srv, PORT_NOT_WORKING); } if (ret == ETIMEDOUT) { if (auth_get_server(req) == NULL) { @@ -600,7 +601,7 @@ static void auth_connect_done(struct tevent_req *subreq) tevent_req_error(req, ret); return; } else if (state->srv) { - fo_set_port_status(state->srv, PORT_WORKING); + be_fo_set_port_status(state->ctx->be, state->srv, PORT_WORKING); } ret = get_user_dn(state, state->ctx->be->sysdb, state->ctx->opts, diff --git a/src/providers/ldap/sdap_async_connection.c b/src/providers/ldap/sdap_async_connection.c index 570d6d4908ef057a7dd5fca4d36626c738bc10d5..3210784787f78a5bb58af06c81ae7f73c77e3867 100644 --- a/src/providers/ldap/sdap_async_connection.c +++ b/src/providers/ldap/sdap_async_connection.c @@ -930,7 +930,7 @@ static void sdap_kinit_done(struct tevent_req *subreq) return; } else { if (kerr == KRB5_KDC_UNREACH) { - fo_set_port_status(state->kdc_srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be, state->kdc_srv, PORT_NOT_WORKING); nextreq = sdap_kinit_next_kdc(req); if (!nextreq) { tevent_req_error(req, ENOMEM); @@ -1156,7 +1156,6 @@ struct tevent_req *sdap_cli_connect_send(TALLOC_CTX *memctx, state->be = be; state->srv = NULL; state->srv_opts = NULL; - state->be = be; state->use_rootdse = !skip_rootdse; ret = sdap_cli_resolve_next(req); @@ -1239,7 +1238,7 @@ static void sdap_cli_connect_done(struct tevent_req *subreq) talloc_zfree(subreq); if (ret) { /* retry another server */ - fo_set_port_status(state->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be, state->srv, PORT_NOT_WORKING); ret = sdap_cli_resolve_next(req); if (ret != EOK) { tevent_req_error(req, ret); @@ -1313,7 +1312,7 @@ static void sdap_cli_rootdse_done(struct tevent_req *subreq) talloc_zfree(subreq); if (ret) { if (ret == ETIMEDOUT) { /* retry another server */ - fo_set_port_status(state->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be, state->srv, PORT_NOT_WORKING); ret = sdap_cli_resolve_next(req); if (ret != EOK) { tevent_req_error(req, ret); @@ -1426,7 +1425,7 @@ static void sdap_cli_kinit_done(struct tevent_req *subreq) talloc_zfree(subreq); if (ret) { if (ret == ETIMEDOUT) { /* child timed out, retry another server */ - fo_set_port_status(state->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be, state->srv, PORT_NOT_WORKING); ret = sdap_cli_resolve_next(req); if (ret != EOK) { tevent_req_error(req, ret); @@ -1510,7 +1509,7 @@ int sdap_cli_connect_recv(struct tevent_req *req, if (tevent_req_is_error(req, &tstate, &err)) { /* mark the server as bad if connection failed */ if (state->srv) { - fo_set_port_status(state->srv, PORT_NOT_WORKING); + be_fo_set_port_status(state->be, state->srv, PORT_NOT_WORKING); } else { if (can_retry) { *can_retry = false; @@ -1522,7 +1521,7 @@ int sdap_cli_connect_recv(struct tevent_req *req, } return EIO; } else if (state->srv) { - fo_set_port_status(state->srv, PORT_WORKING); + be_fo_set_port_status(state->be, state->srv, PORT_WORKING); } if (gsh) { -- 1.7.10.2