Adding upstream version 2.4.38.upstream/2.4.38

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 1460 insertions, 0 deletions

diff --git a/modules/md/mod_md.c b/modules/md/mod_md.c
new file mode 100644
index 0000000..249a0f0
--- /dev/null
+++ b/modules/md/mod_md.c
@@ -0,0 +1,1460 @@
+/* Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You 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.
+ */
+ 
+#include <assert.h>
+#include <apr_optional.h>
+#include <apr_strings.h>
+
+#include <ap_release.h>
+#ifndef AP_ENABLE_EXCEPTION_HOOK
+#define AP_ENABLE_EXCEPTION_HOOK 0
+#endif
+#include <mpm_common.h>
+#include <httpd.h>
+#include <http_core.h>
+#include <http_protocol.h>
+#include <http_request.h>
+#include <http_log.h>
+#include <http_vhost.h>
+#include <ap_listen.h>
+
+#include "md.h"
+#include "md_curl.h"
+#include "md_crypt.h"
+#include "md_http.h"
+#include "md_json.h"
+#include "md_store.h"
+#include "md_store_fs.h"
+#include "md_log.h"
+#include "md_reg.h"
+#include "md_util.h"
+#include "md_version.h"
+#include "md_acme.h"
+#include "md_acme_authz.h"
+
+#include "mod_md.h"
+#include "mod_md_config.h"
+#include "mod_md_os.h"
+#include "mod_ssl.h"
+#include "mod_watchdog.h"
+
+static void md_hooks(apr_pool_t *pool);
+
+AP_DECLARE_MODULE(md) = {
+    STANDARD20_MODULE_STUFF,
+    NULL,                 /* func to create per dir config */
+    NULL,                 /* func to merge per dir config */
+    md_config_create_svr, /* func to create per server config */
+    md_config_merge_svr,  /* func to merge per server config */
+    md_cmds,              /* command handlers */
+    md_hooks,
+#if defined(AP_MODULE_FLAG_NONE)
+    AP_MODULE_FLAG_ALWAYS_MERGE
+#endif
+};
+
+static void md_merge_srv(md_t *md, md_srv_conf_t *base_sc, apr_pool_t *p)
+{
+    if (!md->sc) {
+        md->sc = base_sc;
+    }
+
+    if (!md->ca_url) {
+        md->ca_url = md_config_gets(md->sc, MD_CONFIG_CA_URL);
+    }
+    if (!md->ca_proto) {
+        md->ca_proto = md_config_gets(md->sc, MD_CONFIG_CA_PROTO);
+    }
+    if (!md->ca_agreement) {
+        md->ca_agreement = md_config_gets(md->sc, MD_CONFIG_CA_AGREEMENT);
+    }
+    if (md->sc->s->server_admin && strcmp(DEFAULT_ADMIN, md->sc->s->server_admin)) {
+        apr_array_clear(md->contacts);
+        APR_ARRAY_PUSH(md->contacts, const char *) = 
+        md_util_schemify(p, md->sc->s->server_admin, "mailto");
+    }
+    if (md->drive_mode == MD_DRIVE_DEFAULT) {
+        md->drive_mode = md_config_geti(md->sc, MD_CONFIG_DRIVE_MODE);
+    }
+    if (md->renew_norm <= 0 && md->renew_window <= 0) {
+        md->renew_norm = md_config_get_interval(md->sc, MD_CONFIG_RENEW_NORM);
+        md->renew_window = md_config_get_interval(md->sc, MD_CONFIG_RENEW_WINDOW);
+    }
+    if (md->transitive < 0) {
+        md->transitive = md_config_geti(md->sc, MD_CONFIG_TRANSITIVE);
+    }
+    if (!md->ca_challenges && md->sc->ca_challenges) {
+        md->ca_challenges = apr_array_copy(p, md->sc->ca_challenges);
+    }        
+    if (!md->pkey_spec) {
+        md->pkey_spec = md->sc->pkey_spec;
+        
+    }
+    if (md->require_https < 0) {
+        md->require_https = md_config_geti(md->sc, MD_CONFIG_REQUIRE_HTTPS);
+    }
+    if (md->must_staple < 0) {
+        md->must_staple = md_config_geti(md->sc, MD_CONFIG_MUST_STAPLE);
+    }
+}
+
+static apr_status_t check_coverage(md_t *md, const char *domain, server_rec *s, apr_pool_t *p)
+{
+    if (md_contains(md, domain, 0)) {
+        return APR_SUCCESS;
+    }
+    else if (md->transitive) {
+        APR_ARRAY_PUSH(md->domains, const char*) = apr_pstrdup(p, domain);
+        return APR_SUCCESS;
+    }
+    else {
+        ap_log_error(APLOG_MARK, APLOG_ERR, 0, s, APLOGNO(10040)
+                     "Virtual Host %s:%d matches Managed Domain '%s', but the "
+                     "name/alias %s itself is not managed. A requested MD certificate "
+                     "will not match ServerName.",
+                     s->server_hostname, s->port, md->name, domain);
+        return APR_EINVAL;
+    }
+}
+
+static apr_status_t md_covers_server(md_t *md, server_rec *s, apr_pool_t *p)
+{
+    apr_status_t rv;
+    const char *name;
+    int i;
+    
+    if (APR_SUCCESS == (rv = check_coverage(md, s->server_hostname, s, p)) && s->names) {
+        for (i = 0; i < s->names->nelts; ++i) {
+            name = APR_ARRAY_IDX(s->names, i, const char*);
+            if (APR_SUCCESS != (rv = check_coverage(md, name, s, p))) {
+                break;
+            }
+        }
+    }
+    return rv;
+}
+
+static int matches_port_somewhere(server_rec *s, int port)
+{
+    server_addr_rec *sa;
+    
+    for (sa = s->addrs; sa; sa = sa->next) {
+        if (sa->host_port == port) {
+            /* host_addr might be general (0.0.0.0) or specific, we count this as match */
+            return 1;
+        }
+        if (sa->host_port == 0) {
+            /* wildcard port, answers to all ports. Rare, but may work. */
+            return 1;
+        }
+    }
+    return 0;
+}
+
+static int uses_port_only(server_rec *s, int port)
+{
+    server_addr_rec *sa;
+    int match = 0;
+    for (sa = s->addrs; sa; sa = sa->next) {
+        if (sa->host_port == port) {
+            /* host_addr might be general (0.0.0.0) or specific, we count this as match */
+            match = 1;
+        }
+        else {
+            /* uses other port/wildcard */
+            return 0;
+        }
+    }
+    return match;
+}
+
+static apr_status_t assign_to_servers(md_t *md, server_rec *base_server, 
+                                     apr_pool_t *p, apr_pool_t *ptemp)
+{
+    server_rec *s, *s_https;
+    request_rec r;
+    md_srv_conf_t *sc;
+    md_mod_conf_t *mc;
+    apr_status_t rv = APR_SUCCESS;
+    int i;
+    const char *domain;
+    apr_array_header_t *servers;
+    
+    sc = md_config_get(base_server);
+    mc = sc->mc;
+
+    /* Assign the MD to all server_rec configs that it matches. If there already
+     * is an assigned MD not equal this one, the configuration is in error.
+     */
+    memset(&r, 0, sizeof(r));
+    servers = apr_array_make(ptemp, 5, sizeof(server_rec*));
+    
+    for (s = base_server; s; s = s->next) {
+        if (!mc->manage_base_server && s == base_server) {
+            /* we shall not assign ourselves to the base server */
+            continue;
+        }
+        
+        r.server = s;
+        for (i = 0; i < md->domains->nelts; ++i) {
+            domain = APR_ARRAY_IDX(md->domains, i, const char*);
+            
+            if (ap_matches_request_vhost(&r, domain, s->port)) {
+                /* Create a unique md_srv_conf_t record for this server, if there is none yet */
+                sc = md_config_get_unique(s, p);
+                
+                ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, base_server, APLOGNO(10041)
+                             "Server %s:%d matches md %s (config %s)", 
+                             s->server_hostname, s->port, md->name, sc->name);
+                
+                if (sc->assigned == md) {
+                    /* already matched via another domain name */
+                    goto next_server;
+                }
+                else if (sc->assigned) {
+                    ap_log_error(APLOG_MARK, APLOG_ERR, 0, base_server, APLOGNO(10042)
+                                 "conflict: MD %s matches server %s, but MD %s also matches.",
+                                 md->name, s->server_hostname, sc->assigned->name);
+                    return APR_EINVAL;
+                }
+                
+                /* If this server_rec is only for http: requests. Defined
+                 * alias names to not matter for this MD.
+                 * (see gh issue https://github.com/icing/mod_md/issues/57)
+                 * Otherwise, if server has name or an alias not covered,
+                 * it is by default auto-added (config transitive).
+                 * If mode is "manual", a generated certificate will not match
+                 * all necessary names. */
+                if ((!mc->local_80 || !uses_port_only(s, mc->local_80))
+                    && APR_SUCCESS != (rv = md_covers_server(md, s, p))) {
+                    return rv;
+                }
+
+                sc->assigned = md;
+                APR_ARRAY_PUSH(servers, server_rec*) = s;
+                
+                ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, base_server, APLOGNO(10043)
+                             "Managed Domain %s applies to vhost %s:%d", md->name,
+                             s->server_hostname, s->port);
+                
+                goto next_server;
+            }
+        }
+    next_server:
+        continue;
+    }
+
+    if (APR_SUCCESS == rv) {
+        if (apr_is_empty_array(servers)) {
+            if (md->drive_mode != MD_DRIVE_ALWAYS) {
+                /* Not an error, but looks suspicious */
+                ap_log_error(APLOG_MARK, APLOG_WARNING, 0, base_server, APLOGNO(10045)
+                             "No VirtualHost matches Managed Domain %s", md->name);
+                APR_ARRAY_PUSH(mc->unused_names, const char*)  = md->name;
+            }
+        }
+        else {
+            const char *uri;
+            
+            /* Found matching server_rec's. Collect all 'ServerAdmin's into MD's contact list */
+            apr_array_clear(md->contacts);
+            for (i = 0; i < servers->nelts; ++i) {
+                s = APR_ARRAY_IDX(servers, i, server_rec*);
+                if (s->server_admin && strcmp(DEFAULT_ADMIN, s->server_admin)) {
+                    uri = md_util_schemify(p, s->server_admin, "mailto");
+                    if (md_array_str_index(md->contacts, uri, 0, 0) < 0) {
+                        APR_ARRAY_PUSH(md->contacts, const char *) = uri; 
+                        ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, base_server, APLOGNO(10044)
+                                     "%s: added contact %s", md->name, uri);
+                    }
+                }
+            }
+            
+            if (md->require_https > MD_REQUIRE_OFF) {
+                /* We require https for this MD, but do we have port 443 (or a mapped one)
+                 * available? */
+                if (mc->local_443 <= 0) {
+                    ap_log_error(APLOG_MARK, APLOG_ERR, 0, base_server, APLOGNO(10105)
+                                 "MDPortMap says there is no port for https (443), "
+                                 "but MD %s is configured to require https. This "
+                                 "only works when a 443 port is available.", md->name);
+                    return APR_EINVAL;
+                    
+                }
+                
+                /* Ok, we know which local port represents 443, do we have a server_rec
+                 * for MD that has addresses with port 443? */
+                s_https = NULL;
+                for (i = 0; i < servers->nelts; ++i) {
+                    s = APR_ARRAY_IDX(servers, i, server_rec*);
+                    if (matches_port_somewhere(s, mc->local_443)) {
+                        s_https = s;
+                        break;
+                    }
+                }
+                
+                if (!s_https) {
+                    /* Did not find any server_rec that matches this MD *and* has an
+                     * s->addrs match for the https port. Suspicious. */
+                    ap_log_error(APLOG_MARK, APLOG_WARNING, 0, base_server, APLOGNO(10106)
+                                 "MD %s is configured to require https, but there seems to be "
+                                 "no VirtualHost for it that has port %d in its address list. "
+                                 "This looks as if it will not work.", 
+                                 md->name, mc->local_443);
+                }
+            }
+        }
+        
+    }
+    return rv;
+}
+
+static apr_status_t md_calc_md_list(apr_pool_t *p, apr_pool_t *plog,
+                                    apr_pool_t *ptemp, server_rec *base_server)
+{
+    md_srv_conf_t *sc;
+    md_mod_conf_t *mc;
+    md_t *md, *omd;
+    const char *domain;
+    apr_status_t rv = APR_SUCCESS;
+    ap_listen_rec *lr;
+    apr_sockaddr_t *sa;
+    int i, j;
+
+    (void)plog;
+    sc = md_config_get(base_server);
+    mc = sc->mc;
+    
+    mc->can_http = 0;
+    mc->can_https = 0;
+
+    for (lr = ap_listeners; lr; lr = lr->next) {
+        for (sa = lr->bind_addr; sa; sa = sa->next) {
+            if  (sa->port == mc->local_80 
+                 && (!lr->protocol || !strncmp("http", lr->protocol, 4))) {
+                mc->can_http = 1;
+            }
+            else if (sa->port == mc->local_443
+                     && (!lr->protocol || !strncmp("http", lr->protocol, 4))) {
+                mc->can_https = 1;
+            }
+        }
+    }
+    
+    ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, base_server, APLOGNO(10037)
+                 "server seems%s reachable via http: (port 80->%d) "
+                 "and%s reachable via https: (port 443->%d) ",
+                 mc->can_http? "" : " not", mc->local_80,
+                 mc->can_https? "" : " not", mc->local_443);
+    
+    /* Complete the properties of the MDs, now that we have the complete, merged
+     * server configurations. 
+     */
+    for (i = 0; i < mc->mds->nelts; ++i) {
+        md = APR_ARRAY_IDX(mc->mds, i, md_t*);
+        md_merge_srv(md, sc, p);
+
+        /* Check that we have no overlap with the MDs already completed */
+        for (j = 0; j < i; ++j) {
+            omd = APR_ARRAY_IDX(mc->mds, j, md_t*);
+            if ((domain = md_common_name(md, omd)) != NULL) {
+                ap_log_error(APLOG_MARK, APLOG_ERR, 0, base_server, APLOGNO(10038)
+                             "two Managed Domains have an overlap in domain '%s'"
+                             ", first definition in %s(line %d), second in %s(line %d)",
+                             domain, md->defn_name, md->defn_line_number,
+                             omd->defn_name, omd->defn_line_number);
+                return APR_EINVAL;
+            }
+        }
+
+        /* Assign MD to the server_rec configs that it matches. Perform some
+         * last finishing touches on the MD. */
+        if (APR_SUCCESS != (rv = assign_to_servers(md, base_server, p, ptemp))) {
+            return rv;
+        }
+
+        ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, base_server, APLOGNO(10039)
+                     "Completed MD[%s, CA=%s, Proto=%s, Agreement=%s, Drive=%d, renew=%ld]",
+                     md->name, md->ca_url, md->ca_proto, md->ca_agreement,
+                     md->drive_mode, (long)md->renew_window);
+    }
+    
+    return rv;
+}
+
+/**************************************************************************************************/
+/* store & registry setup */
+
+static apr_status_t store_file_ev(void *baton, struct md_store_t *store,
+                                    md_store_fs_ev_t ev, int group, 
+                                    const char *fname, apr_filetype_e ftype,  
+                                    apr_pool_t *p)
+{
+    server_rec *s = baton;
+    apr_status_t rv;
+    
+    (void)store;
+    ap_log_error(APLOG_MARK, APLOG_TRACE3, 0, s, "store event=%d on %s %s (group %d)", 
+                 ev, (ftype == APR_DIR)? "dir" : "file", fname, group);
+                 
+    /* Directories in group CHALLENGES and STAGING are written to by our watchdog,
+     * running on certain mpms in a child process under a different user. Give them
+     * ownership. 
+     */
+    if (ftype == APR_DIR) {
+        switch (group) {
+            case MD_SG_CHALLENGES:
+            case MD_SG_STAGING:
+                rv = md_make_worker_accessible(fname, p);
+                if (APR_ENOTIMPL != rv) {
+                    return rv;
+                }
+                break;
+            default: 
+                break;
+        }
+    }
+    return APR_SUCCESS;
+}
+
+static apr_status_t check_group_dir(md_store_t *store, md_store_group_t group, 
+                                    apr_pool_t *p, server_rec *s)
+{
+    const char *dir;
+    apr_status_t rv;
+    
+    if (APR_SUCCESS == (rv = md_store_get_fname(&dir, store, group, NULL, NULL, p))
+        && APR_SUCCESS == (rv = apr_dir_make_recursive(dir, MD_FPROT_D_UALL_GREAD, p))) {
+        rv = store_file_ev(s, store, MD_S_FS_EV_CREATED, group, dir, APR_DIR, p);
+    }
+    return rv;
+}
+
+static apr_status_t setup_store(md_store_t **pstore, md_mod_conf_t *mc, 
+                                apr_pool_t *p, server_rec *s)
+{
+    const char *base_dir;
+    apr_status_t rv;
+    MD_CHK_VARS;
+    
+    base_dir = ap_server_root_relative(p, mc->base_dir);
+    
+    if (!MD_OK(md_store_fs_init(pstore, p, base_dir))) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10046)"setup store for %s", base_dir);
+        goto out;
+    }
+
+    md_store_fs_set_event_cb(*pstore, store_file_ev, s);
+    if (   !MD_OK(check_group_dir(*pstore, MD_SG_CHALLENGES, p, s))
+        || !MD_OK(check_group_dir(*pstore, MD_SG_STAGING, p, s))
+        || !MD_OK(check_group_dir(*pstore, MD_SG_ACCOUNTS, p, s))) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10047) 
+                     "setup challenges directory, call %s", MD_LAST_CHK);
+    }
+    
+out:
+    return rv;
+}
+
+static apr_status_t setup_reg(md_reg_t **preg, apr_pool_t *p, server_rec *s, 
+                              int can_http, int can_https)
+{
+    md_srv_conf_t *sc;
+    md_mod_conf_t *mc;
+    md_store_t *store;
+    apr_status_t rv;
+    MD_CHK_VARS;
+    
+    sc = md_config_get(s);
+    mc = sc->mc;
+    
+    if (   MD_OK(setup_store(&store, mc, p, s))
+        && MD_OK(md_reg_init(preg, p, store, mc->proxy_url))) {
+        mc->reg = *preg;
+        return md_reg_set_props(*preg, p, can_http, can_https); 
+    }
+    return rv;
+}
+
+/**************************************************************************************************/
+/* logging setup */
+
+static server_rec *log_server;
+
+static int log_is_level(void *baton, apr_pool_t *p, md_log_level_t level)
+{
+    (void)baton;
+    (void)p;
+    if (log_server) {
+        return APLOG_IS_LEVEL(log_server, (int)level);
+    }
+    return level <= MD_LOG_INFO;
+}
+
+#define LOG_BUF_LEN 16*1024
+
+static void log_print(const char *file, int line, md_log_level_t level, 
+                      apr_status_t rv, void *baton, apr_pool_t *p, const char *fmt, va_list ap)
+{
+    if (log_is_level(baton, p, level)) {
+        char buffer[LOG_BUF_LEN];
+        
+        memset(buffer, 0, sizeof(buffer));
+        apr_vsnprintf(buffer, LOG_BUF_LEN-1, fmt, ap);
+        buffer[LOG_BUF_LEN-1] = '\0';
+
+        if (log_server) {
+            ap_log_error(file, line, APLOG_MODULE_INDEX, level, rv, log_server, "%s",buffer);
+        }
+        else {
+            ap_log_perror(file, line, APLOG_MODULE_INDEX, level, rv, p, "%s", buffer);
+        }
+    }
+}
+
+/**************************************************************************************************/
+/* lifecycle */
+
+static apr_status_t cleanup_setups(void *dummy)
+{
+    (void)dummy;
+    log_server = NULL;
+    return APR_SUCCESS;
+}
+
+static void init_setups(apr_pool_t *p, server_rec *base_server) 
+{
+    log_server = base_server;
+    apr_pool_cleanup_register(p, NULL, cleanup_setups, apr_pool_cleanup_null);
+}
+
+/**************************************************************************************************/
+/* mod_ssl interface */
+
+static APR_OPTIONAL_FN_TYPE(ssl_is_https) *opt_ssl_is_https;
+
+static void init_ssl(void)
+{
+    opt_ssl_is_https = APR_RETRIEVE_OPTIONAL_FN(ssl_is_https);
+}
+
+/**************************************************************************************************/
+/* watchdog based impl. */
+
+#define MD_WATCHDOG_NAME   "_md_"
+
+static APR_OPTIONAL_FN_TYPE(ap_watchdog_get_instance) *wd_get_instance;
+static APR_OPTIONAL_FN_TYPE(ap_watchdog_register_callback) *wd_register_callback;
+static APR_OPTIONAL_FN_TYPE(ap_watchdog_set_callback_interval) *wd_set_interval;
+
+typedef struct {
+    md_t *md;
+
+    int stalled;
+    int renewed;
+    int renewal_notified;
+    apr_time_t restart_at;
+    int need_restart;
+    int restart_processed;
+
+    apr_status_t last_rv;
+    apr_time_t next_check;
+    int error_runs;
+} md_job_t;
+
+typedef struct {
+    apr_pool_t *p;
+    server_rec *s;
+    md_mod_conf_t *mc;
+    ap_watchdog_t *watchdog;
+    
+    apr_time_t next_change;
+    
+    apr_array_header_t *jobs;
+    md_reg_t *reg;
+} md_watchdog;
+
+static void assess_renewal(md_watchdog *wd, md_job_t *job, apr_pool_t *ptemp) 
+{
+    apr_time_t now = apr_time_now();
+    if (now >= job->restart_at) {
+        job->need_restart = 1;
+        ap_log_error( APLOG_MARK, APLOG_TRACE1, 0, wd->s, 
+                     "md(%s): has been renewed, needs restart now", job->md->name);
+    }
+    else {
+        job->next_check = job->restart_at;
+        
+        if (job->renewal_notified) {
+            ap_log_error(APLOG_MARK, APLOG_TRACE1, 0, wd->s, 
+                         "%s: renewed cert valid in %s", 
+                         job->md->name, md_print_duration(ptemp, job->restart_at - now));
+        }
+        else {
+            char ts[APR_RFC822_DATE_LEN];
+
+            apr_rfc822_date(ts, job->restart_at);
+            ap_log_error(APLOG_MARK, APLOG_NOTICE, 0, wd->s, APLOGNO(10051) 
+                         "%s: has been renewed successfully and should be activated at %s"
+                         " (this requires a server restart latest in %s)", 
+                         job->md->name, ts, md_print_duration(ptemp, job->restart_at - now));
+            job->renewal_notified = 1;
+        }
+    }
+}
+
+static apr_status_t load_job_props(md_reg_t *reg, md_job_t *job, apr_pool_t *p)
+{
+    md_store_t *store = md_reg_store_get(reg);
+    md_json_t *jprops;
+    apr_status_t rv;
+    
+    rv = md_store_load_json(store, MD_SG_STAGING, job->md->name,
+                            MD_FN_JOB, &jprops, p);
+    if (APR_SUCCESS == rv) {
+        job->restart_processed = md_json_getb(jprops, MD_KEY_PROCESSED, NULL);
+        job->error_runs = (int)md_json_getl(jprops, MD_KEY_ERRORS, NULL);
+    }
+    return rv;
+}
+
+static apr_status_t save_job_props(md_reg_t *reg, md_job_t *job, apr_pool_t *p)
+{
+    md_store_t *store = md_reg_store_get(reg);
+    md_json_t *jprops;
+    apr_status_t rv;
+    
+    rv = md_store_load_json(store, MD_SG_STAGING, job->md->name, MD_FN_JOB, &jprops, p);
+    if (APR_STATUS_IS_ENOENT(rv)) {
+        jprops = md_json_create(p);
+        rv = APR_SUCCESS;
+    }
+    if (APR_SUCCESS == rv) {
+        md_json_setb(job->restart_processed, jprops, MD_KEY_PROCESSED, NULL);
+        md_json_setl(job->error_runs, jprops, MD_KEY_ERRORS, NULL);
+        rv = md_store_save_json(store, p, MD_SG_STAGING, job->md->name,
+                                MD_FN_JOB, jprops, 0);
+    }
+    return rv;
+}
+
+static apr_status_t check_job(md_watchdog *wd, md_job_t *job, apr_pool_t *ptemp)
+{
+    apr_status_t rv = APR_SUCCESS;
+    apr_time_t valid_from, delay;
+    int errored, renew, error_runs;
+    char ts[APR_RFC822_DATE_LEN];
+    
+    if (apr_time_now() < job->next_check) {
+        /* Job needs to wait */
+        return APR_EAGAIN;
+    }
+    
+    job->next_check = 0;
+    error_runs = job->error_runs;
+
+    if (job->md->state == MD_S_MISSING) {
+        job->stalled = 1;
+    }
+    
+    if (job->stalled) {
+        /* Missing information, this will not change until configuration
+         * is changed and server restarted */
+        rv = APR_INCOMPLETE;
+        ++job->error_runs;
+        goto out;
+    }
+    else if (job->renewed) {
+        assess_renewal(wd, job, ptemp);
+    }
+    else if (APR_SUCCESS == (rv = md_reg_assess(wd->reg, job->md, &errored, &renew, wd->p))) {
+        if (errored) {
+            ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10050) 
+                         "md(%s): in error state", job->md->name);
+        }
+        else if (renew) {
+            ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10052) 
+                         "md(%s): state=%d, driving", job->md->name, job->md->state);
+                         
+            rv = md_reg_stage(wd->reg, job->md, NULL, 0, &valid_from, ptemp);
+            
+            if (APR_SUCCESS == rv) {
+                job->renewed = 1;
+                job->restart_at = valid_from;
+                assess_renewal(wd, job, ptemp);
+            }
+        }
+        else {
+            /* Renew is not necessary yet, leave job->next_check as 0 since 
+             * that keeps the default schedule of running twice a day. */
+            apr_rfc822_date(ts, job->md->expires);
+            ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10053) 
+                         "md(%s): no need to renew yet, cert expires %s", job->md->name, ts);
+        }
+    }
+    
+    if (APR_SUCCESS == rv) {
+        job->error_runs = 0;
+    }
+    else {
+        ap_log_error( APLOG_MARK, APLOG_ERR, rv, wd->s, APLOGNO(10056) 
+                     "processing %s", job->md->name);
+        ++job->error_runs;
+        /* back off duration, depending on the errors we encounter in a row */
+        delay = apr_time_from_sec(5 << (job->error_runs - 1));
+        if (delay > apr_time_from_sec(60*60)) {
+            delay = apr_time_from_sec(60*60);
+        }
+        job->next_check = apr_time_now() + delay;
+        ap_log_error(APLOG_MARK, APLOG_INFO, 0, wd->s, APLOGNO(10057) 
+                     "%s: encountered error for the %d. time, next run in %s",
+                     job->md->name, job->error_runs, md_print_duration(ptemp, delay));
+    }
+    
+out:
+    if (error_runs != job->error_runs) {
+        apr_status_t rv2 = save_job_props(wd->reg, job, ptemp);
+        ap_log_error(APLOG_MARK, APLOG_TRACE1, rv2, wd->s, "%s: saving job props", job->md->name);
+    }
+
+    job->last_rv = rv;
+    return rv;
+}
+
+static apr_status_t run_watchdog(int state, void *baton, apr_pool_t *ptemp)
+{
+    md_watchdog *wd = baton;
+    apr_status_t rv = APR_SUCCESS;
+    md_job_t *job;
+    apr_time_t next_run, now;
+    int restart = 0;
+    int i;
+    
+    switch (state) {
+        case AP_WATCHDOG_STATE_STARTING:
+            ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10054)
+                         "md watchdog start, auto drive %d mds", wd->jobs->nelts);
+            assert(wd->reg);
+        
+            for (i = 0; i < wd->jobs->nelts; ++i) {
+                job = APR_ARRAY_IDX(wd->jobs, i, md_job_t *);
+                load_job_props(wd->reg, job, ptemp);
+            }
+            break;
+        case AP_WATCHDOG_STATE_RUNNING:
+        
+            wd->next_change = 0;
+            ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10055)
+                         "md watchdog run, auto drive %d mds", wd->jobs->nelts);
+                         
+            /* normally, we'd like to run at least twice a day */
+            next_run = apr_time_now() + apr_time_from_sec(MD_SECS_PER_DAY / 2);
+
+            /* Check on all the jobs we have */
+            for (i = 0; i < wd->jobs->nelts; ++i) {
+                job = APR_ARRAY_IDX(wd->jobs, i, md_job_t *);
+                
+                rv = check_job(wd, job, ptemp);
+
+                if (job->need_restart && !job->restart_processed) {
+                    restart = 1;
+                }
+                if (job->next_check && job->next_check < next_run) {
+                    next_run = job->next_check;
+                }
+            }
+
+            now = apr_time_now();
+            if (APLOGdebug(wd->s)) {
+                ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10107)
+                             "next run in %s", md_print_duration(ptemp, next_run - now));
+            }
+            wd_set_interval(wd->watchdog, next_run - now, wd, run_watchdog);
+            break;
+            
+        case AP_WATCHDOG_STATE_STOPPING:
+            ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10058)
+                         "md watchdog stopping");
+            break;
+    }
+
+    if (restart) {
+        const char *action, *names = "";
+        int n;
+        
+        for (i = 0, n = 0; i < wd->jobs->nelts; ++i) {
+            job = APR_ARRAY_IDX(wd->jobs, i, md_job_t *);
+            if (job->need_restart && !job->restart_processed) {
+                names = apr_psprintf(ptemp, "%s%s%s", names, n? " " : "", job->md->name);
+                ++n;
+            }
+        }
+
+        if (n > 0) {
+            int notified = 1;
+
+            /* Run notify command for ready MDs (if configured) and persist that
+             * we have done so. This process might be reaped after n requests or die
+             * of another cause. The one taking over the watchdog need to notify again.
+             */
+            if (wd->mc->notify_cmd) {
+                const char * const *argv;
+                const char *cmdline;
+                int exit_code;
+                
+                cmdline = apr_psprintf(ptemp, "%s %s", wd->mc->notify_cmd, names); 
+                apr_tokenize_to_argv(cmdline, (char***)&argv, ptemp);
+                if (APR_SUCCESS == (rv = md_util_exec(ptemp, argv[0], argv, &exit_code))) {
+                    ap_log_error(APLOG_MARK, APLOG_DEBUG, rv, wd->s, APLOGNO(10108) 
+                                 "notify command '%s' returned %d", 
+                                 wd->mc->notify_cmd, exit_code);
+                }
+                else {
+                    if (APR_EINCOMPLETE == rv && exit_code) {
+                        rv = 0;
+                    }
+                    ap_log_error(APLOG_MARK, APLOG_ERR, rv, wd->s, APLOGNO(10109) 
+                                 "executing MDNotifyCmd %s returned %d", 
+                                  wd->mc->notify_cmd, exit_code);
+                    notified = 0;
+                } 
+            }
+            
+            if (notified) {
+                /* persist the jobs that were notified */
+                for (i = 0, n = 0; i < wd->jobs->nelts; ++i) {
+                    job = APR_ARRAY_IDX(wd->jobs, i, md_job_t *);
+                    if (job->need_restart && !job->restart_processed) {
+                        job->restart_processed = 1;
+                        save_job_props(wd->reg, job, ptemp);
+                    }
+                }
+            }
+            
+            /* FIXME: the server needs to start gracefully to take the new certificate in.
+             * This poses a variety of problems to solve satisfactory for everyone:
+             * - I myself, have no implementation for Windows 
+             * - on *NIX, child processes run with less privileges, preventing
+             *   the signal based restart trigger to work
+             * - admins want better control of timing windows for restarts, e.g.
+             *   during less busy hours/days.
+             */
+            rv = md_server_graceful(ptemp, wd->s);
+            if (APR_ENOTIMPL == rv) {
+                /* self-graceful restart not supported in this setup */
+                action = " and changes will be activated on next (graceful) server restart.";
+            }
+            else {
+                action = " and server has been asked to restart now.";
+            }
+            ap_log_error(APLOG_MARK, APLOG_NOTICE, 0, wd->s, APLOGNO(10059) 
+                         "The Managed Domain%s %s %s been setup%s",
+                         (n > 1)? "s" : "", names, (n > 1)? "have" : "has", action);
+        }
+    }
+    
+    return APR_SUCCESS;
+}
+
+static apr_status_t start_watchdog(apr_array_header_t *names, apr_pool_t *p, 
+                                   md_reg_t *reg, server_rec *s, md_mod_conf_t *mc)
+{
+    apr_allocator_t *allocator;
+    md_watchdog *wd;
+    apr_pool_t *wdp;
+    apr_status_t rv;
+    const char *name;
+    md_t *md;
+    md_job_t *job;
+    int i, errored, renew;
+    
+    wd_get_instance = APR_RETRIEVE_OPTIONAL_FN(ap_watchdog_get_instance);
+    wd_register_callback = APR_RETRIEVE_OPTIONAL_FN(ap_watchdog_register_callback);
+    wd_set_interval = APR_RETRIEVE_OPTIONAL_FN(ap_watchdog_set_callback_interval);
+    
+    if (!wd_get_instance || !wd_register_callback || !wd_set_interval) {
+        ap_log_error(APLOG_MARK, APLOG_CRIT, 0, s, APLOGNO(10061) "mod_watchdog is required");
+        return !OK;
+    }
+    
+    /* We want our own pool with own allocator to keep data across watchdog invocations */
+    apr_allocator_create(&allocator);
+    apr_allocator_max_free_set(allocator, ap_max_mem_free);
+    rv = apr_pool_create_ex(&wdp, p, NULL, allocator);
+    if (rv != APR_SUCCESS) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10062) "md_watchdog: create pool");
+        return rv;
+    }
+    apr_allocator_owner_set(allocator, wdp);
+    apr_pool_tag(wdp, "md_watchdog");
+
+    wd = apr_pcalloc(wdp, sizeof(*wd));
+    wd->p = wdp;
+    wd->reg = reg;
+    wd->s = s;
+    wd->mc = mc;
+    
+    wd->jobs = apr_array_make(wd->p, 10, sizeof(md_job_t *));
+    for (i = 0; i < names->nelts; ++i) {
+        name = APR_ARRAY_IDX(names, i, const char *);
+        md = md_reg_get(wd->reg, name, wd->p);
+        if (md) {
+            md_reg_assess(wd->reg, md, &errored, &renew, wd->p);
+            if (errored) {
+                ap_log_error( APLOG_MARK, APLOG_WARNING, 0, wd->s, APLOGNO(10063) 
+                             "md(%s): seems errored. Will not process this any further.", name);
+            }
+            else {
+                job = apr_pcalloc(wd->p, sizeof(*job));
+                
+                job->md = md;
+                APR_ARRAY_PUSH(wd->jobs, md_job_t*) = job;
+
+                ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, wd->s, APLOGNO(10064) 
+                             "md(%s): state=%d, driving", name, md->state);
+                
+                load_job_props(reg, job, wd->p);
+                if (job->error_runs) {
+                    /* We are just restarting. If we encounter jobs that had errors
+                     * running the protocol on previous staging runs, we reset
+                     * the staging area for it, in case we persisted something that
+                     * causes a loop. */
+                    md_store_t *store = md_reg_store_get(wd->reg);
+                    
+                    md_store_purge(store, p, MD_SG_STAGING, job->md->name);
+                    md_store_purge(store, p, MD_SG_CHALLENGES, job->md->name);
+                }
+            }
+        }
+    }
+
+    if (!wd->jobs->nelts) {
+        ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, s, APLOGNO(10065)
+                     "no managed domain in state to drive, no watchdog needed, "
+                     "will check again on next server (graceful) restart");
+        apr_pool_destroy(wd->p);
+        return APR_SUCCESS;
+    }
+    
+    if (APR_SUCCESS != (rv = wd_get_instance(&wd->watchdog, MD_WATCHDOG_NAME, 0, 1, wd->p))) {
+        ap_log_error(APLOG_MARK, APLOG_CRIT, rv, s, APLOGNO(10066) 
+                     "create md watchdog(%s)", MD_WATCHDOG_NAME);
+        return rv;
+    }
+    rv = wd_register_callback(wd->watchdog, 0, wd, run_watchdog);
+    ap_log_error(APLOG_MARK, rv? APLOG_CRIT : APLOG_DEBUG, rv, s, APLOGNO(10067) 
+                 "register md watchdog(%s)", MD_WATCHDOG_NAME);
+    return rv;
+}
+ 
+static void load_stage_sets(apr_array_header_t *names, apr_pool_t *p, 
+                            md_reg_t *reg, server_rec *s)
+{
+    const char *name; 
+    apr_status_t rv;
+    int i;
+    
+    for (i = 0; i < names->nelts; ++i) {
+        name = APR_ARRAY_IDX(names, i, const char*);
+        if (APR_SUCCESS == (rv = md_reg_load(reg, name, p))) {
+            ap_log_error( APLOG_MARK, APLOG_INFO, rv, s, APLOGNO(10068) 
+                         "%s: staged set activated", name);
+        }
+        else if (!APR_STATUS_IS_ENOENT(rv)) {
+            ap_log_error( APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10069)
+                         "%s: error loading staged set", name);
+        }
+    }
+    return;
+}
+
+static apr_status_t md_post_config(apr_pool_t *p, apr_pool_t *plog,
+                                   apr_pool_t *ptemp, server_rec *s)
+{
+    void *data = NULL;
+    const char *mod_md_init_key = "mod_md_init_counter";
+    md_srv_conf_t *sc;
+    md_mod_conf_t *mc;
+    md_reg_t *reg;
+    const md_t *md;
+    apr_array_header_t *drive_names;
+    apr_status_t rv = APR_SUCCESS;
+    int i, dry_run = 0;
+
+    apr_pool_userdata_get(&data, mod_md_init_key, s->process->pool);
+    if (data == NULL) {
+        /* At the first start, httpd makes a config check dry run. It
+         * runs all config hooks to check if it can. If so, it does
+         * this all again and starts serving requests.
+         * 
+         * This is known.
+         *
+         * On a dry run, we therefore do all the cheap config things we
+         * need to do. Because otherwise mod_ssl fails because it calls
+         * us unprepared.
+         * But synching our configuration with the md store
+         * and determining which domains to drive and start a watchdog
+         * and all that, we do not.
+         */
+        ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, s, APLOGNO(10070)
+                     "initializing post config dry run");
+        apr_pool_userdata_set((const void *)1, mod_md_init_key,
+                              apr_pool_cleanup_null, s->process->pool);
+        dry_run = 1;
+    }
+    else {
+        ap_log_error( APLOG_MARK, APLOG_INFO, 0, s, APLOGNO(10071)
+                     "mod_md (v%s), initializing...", MOD_MD_VERSION);
+    }
+
+    (void)plog;
+    init_setups(p, s);
+    md_log_set(log_is_level, log_print, NULL);
+
+    /* Check uniqueness of MDs, calculate global, configured MD list.
+     * If successful, we have a list of MD definitions that do not overlap. */
+    /* We also need to find out if we can be reached on 80/443 from the outside (e.g. the CA) */
+    if (APR_SUCCESS != (rv =  md_calc_md_list(p, plog, ptemp, s))) {
+        return rv;
+    }
+
+    md_config_post_config(s, p);
+    sc = md_config_get(s);
+    mc = sc->mc;
+
+    /* Synchronize the definitions we now have with the store via a registry (reg). */
+    if (APR_SUCCESS != (rv = setup_reg(&reg, p, s, mc->can_http, mc->can_https))) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10072)
+                     "setup md registry");
+        goto out;
+    }
+    
+    if (APR_SUCCESS != (rv = md_reg_sync(reg, p, ptemp, mc->mds))) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10073)
+                     "synching %d mds to registry", mc->mds->nelts);
+    }
+    
+    /* Determine the managed domains that are in auto drive_mode. For those,
+     * determine in which state they are:
+     *  - UNKNOWN:            should not happen, report, don't drive
+     *  - ERROR:              something we do not know how to fix, report, don't drive
+     *  - INCOMPLETE/EXPIRED: need to drive them right away
+     *  - COMPLETE:           determine when cert expires, drive when the time comes
+     *
+     * Start the watchdog if we have anything, now or in the future.
+     */
+    drive_names = apr_array_make(ptemp, mc->mds->nelts+1, sizeof(const char *));
+    for (i = 0; i < mc->mds->nelts; ++i) {
+        md = APR_ARRAY_IDX(mc->mds, i, const md_t *);
+        switch (md->drive_mode) {
+            case MD_DRIVE_AUTO:
+                if (md_array_str_index(mc->unused_names, md->name, 0, 0) >= 0) {
+                    break;
+                }
+                /* fall through */
+            case MD_DRIVE_ALWAYS:
+                APR_ARRAY_PUSH(drive_names, const char *) = md->name; 
+                break;
+            default:
+                /* leave out */
+                break;
+        }
+    }
+    
+    init_ssl();
+    
+    if (dry_run) {
+        goto out;
+    }
+    
+    /* If there are MDs to drive, start a watchdog to check on them regularly */
+    if (drive_names->nelts > 0) {
+        ap_log_error(APLOG_MARK, APLOG_DEBUG, rv, s, APLOGNO(10074)
+                     "%d out of %d mds are configured for auto-drive", 
+                     drive_names->nelts, mc->mds->nelts);
+    
+        load_stage_sets(drive_names, p, reg, s);
+        md_http_use_implementation(md_curl_get_impl(p));
+        rv = start_watchdog(drive_names, p, reg, s, mc);
+    }
+    else {
+        ap_log_error( APLOG_MARK, APLOG_DEBUG, 0, s, APLOGNO(10075)
+                     "no mds to auto drive, no watchdog needed");
+    }
+out:
+    return rv;
+}
+
+/**************************************************************************************************/
+/* Access API to other httpd components */
+
+static int md_is_managed(server_rec *s)
+{
+    md_srv_conf_t *conf = md_config_get(s);
+
+    if (conf && conf->assigned) {
+        ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, s, APLOGNO(10076) 
+                     "%s: manages server %s", conf->assigned->name, s->server_hostname);
+        return 1;
+    }
+    ap_log_error(APLOG_MARK, APLOG_TRACE1, 0, s,  
+                 "server %s is not managed", s->server_hostname);
+    return 0;
+}
+
+static apr_status_t setup_fallback_cert(md_store_t *store, const md_t *md, 
+                                        server_rec *s, apr_pool_t *p)
+{
+    md_pkey_t *pkey;
+    md_cert_t *cert;
+    md_pkey_spec_t spec;
+    apr_status_t rv;
+    MD_CHK_VARS;
+    
+    spec.type = MD_PKEY_TYPE_RSA;
+    spec.params.rsa.bits = MD_PKEY_RSA_BITS_DEF;
+    
+    if (   !MD_OK(md_pkey_gen(&pkey, p, &spec))
+        || !MD_OK(md_store_save(store, p, MD_SG_DOMAINS, md->name, 
+                                MD_FN_FALLBACK_PKEY, MD_SV_PKEY, (void*)pkey, 0))
+        || !MD_OK(md_cert_self_sign(&cert, "Apache Managed Domain Fallback", 
+                                    md->domains, pkey, apr_time_from_sec(14 * MD_SECS_PER_DAY), p))
+        || !MD_OK(md_store_save(store, p, MD_SG_DOMAINS, md->name, 
+                                MD_FN_FALLBACK_CERT, MD_SV_CERT, (void*)cert, 0))) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s,  
+                     "%s: setup fallback certificate, call %s", md->name, MD_LAST_CHK);
+    }
+    return rv;
+}
+
+static int fexists(const char *fname, apr_pool_t *p)
+{
+    return (*fname && APR_SUCCESS == md_util_is_file(fname, p));
+}
+
+static apr_status_t md_get_certificate(server_rec *s, apr_pool_t *p,
+                                       const char **pkeyfile, const char **pcertfile)
+{
+    apr_status_t rv = APR_ENOENT;    
+    md_srv_conf_t *sc;
+    md_reg_t *reg;
+    md_store_t *store;
+    const md_t *md;
+    MD_CHK_VARS;
+    
+    *pkeyfile = NULL;
+    *pcertfile = NULL;
+
+    ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, s, APLOGNO(10113)
+                 "md_get_certificate called for vhost %s.", s->server_hostname);
+
+    sc = md_config_get(s);
+    if (!sc) {
+        ap_log_error(APLOG_MARK, APLOG_TRACE1, 0, s,  
+                     "asked for certificate of server %s which has no md config", 
+                     s->server_hostname);
+        return APR_ENOENT;
+    }
+    
+    if (!sc->assigned) {
+        /* Hmm, mod_ssl (or someone like it) asks for certificates for a server
+         * where we did not assign a MD to. Either the user forgot to configure
+         * that server with SSL certs, has misspelled a server name or we have
+         * a bug that prevented us from taking responsibility for this server.
+         * Either way, make some polite noise */
+        ap_log_error(APLOG_MARK, APLOG_NOTICE, 0, s, APLOGNO(10114)  
+                     "asked for certificate of server %s which has no MD assigned. This "
+                     "could be ok, but most likely it is either a misconfiguration or "
+                     "a bug. Please check server names and MD names carefully and if "
+                     "everything checks open, please open an issue.", 
+                     s->server_hostname);
+        return APR_ENOENT;
+    }
+    
+    assert(sc->mc);
+    reg = sc->mc->reg;
+    assert(reg);
+    
+    md = md_reg_get(reg, sc->assigned->name, p);
+    if (!md) {
+        ap_log_error(APLOG_MARK, APLOG_WARNING, 0, s, APLOGNO(10115) 
+                     "unable to hand out certificates, as registry can no longer "
+                     "find MD '%s'.", sc->assigned->name);
+        return APR_ENOENT;
+    }
+    
+    if (!MD_OK(md_reg_get_cred_files(reg, md, p, pkeyfile, pcertfile))) {
+        ap_log_error(APLOG_MARK, APLOG_ERR, rv, s, APLOGNO(10110) 
+                     "retrieving credentials for MD %s", md->name);
+        return rv;
+    }
+    
+    if (!fexists(*pkeyfile, p) || !fexists(*pcertfile, p)) { 
+        /* Provide temporary, self-signed certificate as fallback, so that
+         * clients do not get obscure TLS handshake errors or will see a fallback
+         * virtual host that is not intended to be served here. */
+        store = md_reg_store_get(reg);
+        assert(store);    
+        
+        md_store_get_fname(pkeyfile, store, MD_SG_DOMAINS, 
+                           md->name, MD_FN_FALLBACK_PKEY, p);
+        md_store_get_fname(pcertfile, store, MD_SG_DOMAINS, 
+                           md->name, MD_FN_FALLBACK_CERT, p);
+        if (!fexists(*pkeyfile, p) || !fexists(*pcertfile, p)) { 
+            if (!MD_OK(setup_fallback_cert(store, md, s, p))) {
+                return rv;
+            }
+        }
+        
+        ap_log_error(APLOG_MARK, APLOG_DEBUG, 0, s, APLOGNO(10116)  
+                     "%s: providing fallback certificate for server %s", 
+                     md->name, s->server_hostname);
+        return APR_EAGAIN;
+    }
+    
+    /* We have key and cert files, but they might no longer be valid or not
+     * match all domain names. Still use these files for now, but indicate that 
+     * resources should no longer be served until we have a new certificate again. */
+    if (md->state != MD_S_COMPLETE) {
+        rv = APR_EAGAIN;
+    }
+    ap_log_error(APLOG_MARK, APLOG_DEBUG, rv, s, APLOGNO(10077) 
+                 "%s: providing certificate for server %s", md->name, s->server_hostname);
+    return rv;
+}
+
+static int compat_warned;
+static apr_status_t md_get_credentials(server_rec *s, apr_pool_t *p,
+                                       const char **pkeyfile, 
+                                       const char **pcertfile, 
+                                       const char **pchainfile)
+{
+    *pchainfile = NULL;
+    if (!compat_warned) {
+        compat_warned = 1;
+        ap_log_error(APLOG_MARK, APLOG_WARNING, 0, s, /* no APLOGNO */
+                     "You are using mod_md with an old patch to mod_ssl. This will "
+                     " work for now, but support will be dropped in a future release.");
+    }
+    return md_get_certificate(s, p, pkeyfile, pcertfile);
+}
+
+static int md_is_challenge(conn_rec *c, const char *servername,
+                           X509 **pcert, EVP_PKEY **pkey)
+{
+    md_srv_conf_t *sc;
+    apr_size_t slen, sufflen = sizeof(MD_TLSSNI01_DNS_SUFFIX) - 1;
+    apr_status_t rv;
+
+    slen = strlen(servername);
+    if (slen <= sufflen 
+        || apr_strnatcasecmp(MD_TLSSNI01_DNS_SUFFIX, servername + slen - sufflen)) {
+        return 0;
+    }
+    
+    sc = md_config_get(c->base_server);
+    if (sc && sc->mc->reg) {
+        md_store_t *store = md_reg_store_get(sc->mc->reg);
+        md_cert_t *mdcert;
+        md_pkey_t *mdpkey;
+        
+        rv = md_store_load(store, MD_SG_CHALLENGES, servername, 
+                           MD_FN_TLSSNI01_CERT, MD_SV_CERT, (void**)&mdcert, c->pool);
+        if (APR_SUCCESS == rv && (*pcert = md_cert_get_X509(mdcert))) {
+            rv = md_store_load(store, MD_SG_CHALLENGES, servername, 
+                               MD_FN_TLSSNI01_PKEY, MD_SV_PKEY, (void**)&mdpkey, c->pool);
+            if (APR_SUCCESS == rv && (*pkey = md_pkey_get_EVP_PKEY(mdpkey))) {
+                ap_log_cerror(APLOG_MARK, APLOG_INFO, 0, c, APLOGNO(10078)
+                              "%s: is a tls-sni-01 challenge host", servername);
+                return 1;
+            }
+            ap_log_cerror(APLOG_MARK, APLOG_WARNING, rv, c, APLOGNO(10079)
+                          "%s: challenge data not complete, key unavailable", servername);
+        }
+        else {
+            ap_log_cerror(APLOG_MARK, APLOG_INFO, rv, c, APLOGNO(10080)
+                          "%s: unknown TLS SNI challenge host", servername);
+        }
+    }
+    *pcert = NULL;
+    *pkey = NULL;
+    return 0;
+}
+
+/**************************************************************************************************/
+/* ACME challenge responses */
+
+#define WELL_KNOWN_PREFIX           "/.well-known/"
+#define ACME_CHALLENGE_PREFIX       WELL_KNOWN_PREFIX"acme-challenge/"
+
+static int md_http_challenge_pr(request_rec *r)
+{
+    apr_bucket_brigade *bb;
+    const md_srv_conf_t *sc;
+    const char *name, *data;
+    md_reg_t *reg;
+    int configured;
+    apr_status_t rv;
+    
+    if (r->parsed_uri.path 
+        && !strncmp(ACME_CHALLENGE_PREFIX, r->parsed_uri.path, sizeof(ACME_CHALLENGE_PREFIX)-1)) {
+        sc = ap_get_module_config(r->server->module_config, &md_module);
+        if (sc && sc->mc) {
+            ap_log_rerror(APLOG_MARK, APLOG_TRACE1, 0, r, 
+                          "access inside /.well-known/acme-challenge for %s%s", 
+                          r->hostname, r->parsed_uri.path);
+            configured = (NULL != md_get_by_domain(sc->mc->mds, r->hostname));
+            name = r->parsed_uri.path + sizeof(ACME_CHALLENGE_PREFIX)-1;
+            reg = sc && sc->mc? sc->mc->reg : NULL;
+            
+            if (strlen(name) && !ap_strchr_c(name, '/') && reg) {
+                md_store_t *store = md_reg_store_get(reg);
+                
+                rv = md_store_load(store, MD_SG_CHALLENGES, r->hostname, 
+                                   MD_FN_HTTP01, MD_SV_TEXT, (void**)&data, r->pool);
+                ap_log_rerror(APLOG_MARK, APLOG_DEBUG, rv, r, 
+                              "loading challenge for %s (%s)", r->hostname, r->uri);
+                if (APR_SUCCESS == rv) {
+                    apr_size_t len = strlen(data);
+                    
+                    if (r->method_number != M_GET) {
+                        return HTTP_NOT_IMPLEMENTED;
+                    }
+                    /* A GET on a challenge resource for a hostname we are
+                     * configured for. Let's send the content back */
+                    r->status = HTTP_OK;
+                    apr_table_setn(r->headers_out, "Content-Length", apr_ltoa(r->pool, (long)len));
+                    
+                    bb = apr_brigade_create(r->pool, r->connection->bucket_alloc);
+                    apr_brigade_write(bb, NULL, NULL, data, len);
+                    ap_pass_brigade(r->output_filters, bb);
+                    apr_brigade_cleanup(bb);
+                    
+                    return DONE;
+                }
+                else if (!configured) {
+                    /* The request hostname is not for a configured domain. We are not
+                     * the sole authority here for /.well-known/acme-challenge (see PR62189).
+                     * So, we decline to handle this and let others step in.
+                     */
+                    return DECLINED;
+                }
+                else if (APR_STATUS_IS_ENOENT(rv)) {
+                    return HTTP_NOT_FOUND;
+                }
+                else {
+                    ap_log_rerror(APLOG_MARK, APLOG_ERR, rv, r, APLOGNO(10081)
+                                  "loading challenge %s from store", name);
+                    return HTTP_INTERNAL_SERVER_ERROR;
+                }
+            }
+        }
+    }
+    return DECLINED;
+}
+
+/**************************************************************************************************/
+/* Require Https hook */
+
+static int md_require_https_maybe(request_rec *r)
+{
+    const md_srv_conf_t *sc;
+    apr_uri_t uri;
+    const char *s;
+    int status;
+    
+    if (opt_ssl_is_https && r->parsed_uri.path
+        && strncmp(WELL_KNOWN_PREFIX, r->parsed_uri.path, sizeof(WELL_KNOWN_PREFIX)-1)) {
+        
+        sc = ap_get_module_config(r->server->module_config, &md_module);
+        if (sc && sc->assigned && sc->assigned->require_https > MD_REQUIRE_OFF) {
+            if (opt_ssl_is_https(r->connection)) {
+                /* Using https:
+                 * if 'permanent' and no one else set a HSTS header already, do it */
+                if (sc->assigned->require_https == MD_REQUIRE_PERMANENT 
+                    && sc->mc->hsts_header && !apr_table_get(r->headers_out, MD_HSTS_HEADER)) {
+                    apr_table_setn(r->headers_out, MD_HSTS_HEADER, sc->mc->hsts_header);
+                }
+            }
+            else {
+                /* Not using https:, but require it. Redirect. */
+                if (r->method_number == M_GET) {
+                    /* safe to use the old-fashioned codes */
+                    status = ((MD_REQUIRE_PERMANENT == sc->assigned->require_https)? 
+                              HTTP_MOVED_PERMANENTLY : HTTP_MOVED_TEMPORARILY);
+                }
+                else {
+                    /* these should keep the method unchanged on retry */
+                    status = ((MD_REQUIRE_PERMANENT == sc->assigned->require_https)? 
+                              HTTP_PERMANENT_REDIRECT : HTTP_TEMPORARY_REDIRECT);
+                }
+                
+                s = ap_construct_url(r->pool, r->uri, r);
+                if (APR_SUCCESS == apr_uri_parse(r->pool, s, &uri)) {
+                    uri.scheme = (char*)"https";
+                    uri.port = 443;
+                    uri.port_str = (char*)"443";
+                    uri.query = r->parsed_uri.query;
+                    uri.fragment = r->parsed_uri.fragment;
+                    s = apr_uri_unparse(r->pool, &uri, APR_URI_UNP_OMITUSERINFO);
+                    if (s && *s) {
+                        apr_table_setn(r->headers_out, "Location", s);
+                        return status;
+                    }
+                }
+            }
+        }
+    }
+    return DECLINED;
+}
+
+/* Runs once per created child process. Perform any process 
+ * related initialization here.
+ */
+static void md_child_init(apr_pool_t *pool, server_rec *s)
+{
+    (void)pool;
+    (void)s;
+}
+
+/* Install this module into the apache2 infrastructure.
+ */
+static void md_hooks(apr_pool_t *pool)
+{
+    static const char *const mod_ssl[] = { "mod_ssl.c", NULL};
+
+    md_acme_init(pool, AP_SERVER_BASEVERSION);
+        
+    ap_log_perror(APLOG_MARK, APLOG_TRACE1, 0, pool, "installing hooks");
+    
+    /* Run once after configuration is set, before mod_ssl.
+     */
+    ap_hook_post_config(md_post_config, NULL, mod_ssl, APR_HOOK_MIDDLE);
+    
+    /* Run once after a child process has been created.
+     */
+    ap_hook_child_init(md_child_init, NULL, mod_ssl, APR_HOOK_MIDDLE);
+
+    /* answer challenges *very* early, before any configured authentication may strike */
+    ap_hook_post_read_request(md_require_https_maybe, NULL, NULL, APR_HOOK_FIRST);
+    ap_hook_post_read_request(md_http_challenge_pr, NULL, NULL, APR_HOOK_MIDDLE);
+
+    APR_REGISTER_OPTIONAL_FN(md_is_managed);
+    APR_REGISTER_OPTIONAL_FN(md_get_certificate);
+    APR_REGISTER_OPTIONAL_FN(md_is_challenge);
+    APR_REGISTER_OPTIONAL_FN(md_get_credentials);
+}
+