mod_cache_disk.c

            if (rv != APR_SUCCESS) {
                ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00721)
                        "could not create vary file %s",
                        dobj->vary.tempfile);
                return rv;
            }

            amt = sizeof(format);
            rv = apr_file_write_full(dobj->vary.tempfd, &format, amt, NULL);
            if (rv != APR_SUCCESS) {
                ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00722)
                        "could not write to vary file %s",
                        dobj->vary.tempfile);
                apr_file_close(dobj->vary.tempfd);
                apr_pool_destroy(dobj->vary.pool);
                return rv;
            }

            amt = sizeof(h->cache_obj->info.expire);
            rv = apr_file_write_full(dobj->vary.tempfd,
                                     &h->cache_obj->info.expire, amt, NULL);
            if (rv != APR_SUCCESS) {
                ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00723)
                        "could not write to vary file %s",
                        dobj->vary.tempfile);
                apr_file_close(dobj->vary.tempfd);
                apr_pool_destroy(dobj->vary.pool);
                return rv;
            }

            varray = apr_array_make(r->pool, 6, sizeof(char*));
            tokens_to_array(r->pool, tmp, varray);

            store_array(dobj->vary.tempfd, varray);

            rv = apr_file_close(dobj->vary.tempfd);
            if (rv != APR_SUCCESS) {
                ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00724)
                        "could not close vary file %s",
                        dobj->vary.tempfile);
                apr_pool_destroy(dobj->vary.pool);
                return rv;
            }

            tmp = regen_key(r->pool, dobj->headers_in, varray, dobj->name);
            dobj->prefix = dobj->hdrs.file;
            dobj->hashfile = NULL;
            dobj->data.file = data_file(r->pool, conf, dobj, tmp);
            dobj->hdrs.file = header_file(r->pool, conf, dobj, tmp);
        }
    }


    rv = apr_file_mktemp(&dobj->hdrs.tempfd, dobj->hdrs.tempfile,
                         APR_CREATE | APR_WRITE | APR_BINARY |
                         APR_BUFFERED | APR_EXCL, dobj->hdrs.pool);

    if (rv != APR_SUCCESS) {
        ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00725)
                "could not create header file %s",
                dobj->hdrs.tempfile);
        return rv;
    }

    disk_info.format = DISK_FORMAT_VERSION;
    disk_info.date = h->cache_obj->info.date;
    disk_info.expire = h->cache_obj->info.expire;
    disk_info.entity_version = dobj->disk_info.entity_version++;
    disk_info.request_time = h->cache_obj->info.request_time;
    disk_info.response_time = h->cache_obj->info.response_time;
    disk_info.status = h->cache_obj->info.status;
    disk_info.inode = dobj->disk_info.inode;
    disk_info.device = dobj->disk_info.device;
    disk_info.has_body = dobj->disk_info.has_body;
    disk_info.header_only = dobj->disk_info.header_only;

    disk_info.name_len = strlen(dobj->name);

    memcpy(&disk_info.control, &h->cache_obj->info.control, sizeof(cache_control_t));

    iov[0].iov_base = (void*)&disk_info;
    iov[0].iov_len = sizeof(disk_cache_info_t);
    iov[1].iov_base = (void*)dobj->name;
    iov[1].iov_len = disk_info.name_len;

    rv = apr_file_writev_full(dobj->hdrs.tempfd, (const struct iovec *) &iov,
                              2, &amt);
    if (rv != APR_SUCCESS) {
        ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00726)
                "could not write info to header file %s",
                dobj->hdrs.tempfile);
        apr_file_close(dobj->hdrs.tempfd);
        apr_pool_destroy(dobj->hdrs.pool);
        return rv;
    }

    if (dobj->headers_out) {
        rv = store_table(dobj->hdrs.tempfd, dobj->headers_out);
        if (rv != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00727)
                    "could not write out-headers to header file %s",
                    dobj->hdrs.tempfile);
            apr_file_close(dobj->hdrs.tempfd);
            apr_pool_destroy(dobj->hdrs.pool);
            return rv;
        }
    }

    /* Parse the vary header and dump those fields from the headers_in. */
    /* FIXME: Make call to the same thing cache_select calls to crack Vary. */
    if (dobj->headers_in) {
        rv = store_table(dobj->hdrs.tempfd, dobj->headers_in);
        if (rv != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00728)
                    "could not write in-headers to header file %s",
                    dobj->hdrs.tempfile);
            apr_file_close(dobj->hdrs.tempfd);
            apr_pool_destroy(dobj->hdrs.pool);
            return rv;
        }
    }

    rv = apr_file_close(dobj->hdrs.tempfd); /* flush and close */
    if (rv != APR_SUCCESS) {
        ap_log_rerror(APLOG_MARK, APLOG_WARNING, rv, r, APLOGNO(00729)
                "could not close header file %s",
                dobj->hdrs.tempfile);
        apr_pool_destroy(dobj->hdrs.pool);
        return rv;
    }

    return APR_SUCCESS;
}

static apr_status_t store_body(cache_handle_t *h, request_rec *r,
                               apr_bucket_brigade *in, apr_bucket_brigade *out)
{
    apr_bucket *e;
    apr_status_t rv = APR_SUCCESS;
    disk_cache_object_t *dobj = (disk_cache_object_t *) h->cache_obj->vobj;
    disk_cache_dir_conf *dconf = ap_get_module_config(r->per_dir_config, &cache_disk_module);
    int seen_eos = 0;

    if (!dobj->offset) {
        dobj->offset = dconf->readsize;
    }
    if (!dobj->timeout && dconf->readtime) {
        dobj->timeout = apr_time_now() + dconf->readtime;
    }

    if (dobj->offset) {
        apr_brigade_partition(in, dobj->offset, &e);
    }

    while (APR_SUCCESS == rv && !APR_BRIGADE_EMPTY(in)) {
        const char *str;
        apr_size_t length, written;

        e = APR_BRIGADE_FIRST(in);

        /* are we done completely? if so, pass any trailing buckets right through */
        if (dobj->done || !dobj->data.pool) {
            APR_BUCKET_REMOVE(e);
            APR_BRIGADE_INSERT_TAIL(out, e);
            continue;
        }

        /* have we seen eos yet? */
        if (APR_BUCKET_IS_EOS(e)) {
            seen_eos = 1;
            dobj->done = 1;
            APR_BUCKET_REMOVE(e);
            APR_BRIGADE_INSERT_TAIL(out, e);
            break;
        }

        /* honour flush buckets, we'll get called again */
        if (APR_BUCKET_IS_FLUSH(e)) {
            APR_BUCKET_REMOVE(e);
            APR_BRIGADE_INSERT_TAIL(out, e);
            break;
        }

        /* metadata buckets are preserved as is */
        if (APR_BUCKET_IS_METADATA(e)) {
            APR_BUCKET_REMOVE(e);
            APR_BRIGADE_INSERT_TAIL(out, e);
            continue;
        }

        /* read the bucket, write to the cache */
        rv = apr_bucket_read(e, &str, &length, APR_BLOCK_READ);
        APR_BUCKET_REMOVE(e);
        APR_BRIGADE_INSERT_TAIL(out, e);
        if (rv != APR_SUCCESS) {
            ap_log_rerror(APLOG_MARK, APLOG_ERR, 0, r, APLOGNO(00730)
                    "Error when reading bucket for URL %s",
                    h->cache_obj->key);
            /* Remove the intermediate cache file and return non-APR_SUCCESS */
            apr_pool_destroy(dobj->data.pool);
            return rv;
        }

        /* don't write empty buckets to the cache */
        if (!length) {
            continue;
        }

        if (!dobj->disk_info.header_only) {

            /* Attempt to create the data file at the last possible moment, if
             * the body is empty, we don't write a file at all, and save an inode.
             */
            if (!dobj->data.tempfd) {
                apr_finfo_t finfo;
                rv = apr_file_mktemp(&dobj->data.tempfd, dobj->data.tempfile,
                        APR_CREATE | APR_WRITE | APR_BINARY | APR_BUFFERED
                                | APR_EXCL, dobj->data.pool);
                if (rv != APR_SUCCESS) {
                    apr_pool_destroy(dobj->data.pool);
                    return rv;
                }
                dobj->file_size = 0;
                rv = apr_file_info_get(&finfo, APR_FINFO_IDENT,
                        dobj->data.tempfd);
                if (rv != APR_SUCCESS) {
                    apr_pool_destroy(dobj->data.pool);
                    return rv;
                }
                dobj->disk_info.device = finfo.device;
                dobj->disk_info.inode = finfo.inode;
                dobj->disk_info.has_body = 1;
            }

            /* write to the cache, leave if we fail */
            rv = apr_file_write_full(dobj->data.tempfd, str, length, &written);
            if (rv != APR_SUCCESS) {
                ap_log_rerror(
                        APLOG_MARK, APLOG_ERR, 0, r, APLOGNO(00731) "Error when writing cache file for URL %s", h->cache_obj->key);
                /* Remove the intermediate cache file and return non-APR_SUCCESS */
                apr_pool_destroy(dobj->data.pool);
                return rv;
            }
            dobj->file_size += written;
            if (dobj->file_size > dconf->maxfs) {
                ap_log_rerror(
                        APLOG_MARK, APLOG_DEBUG, 0, r, APLOGNO(00732) "URL %s failed the size check "
                        "(%" APR_OFF_T_FMT ">%" APR_OFF_T_FMT ")", h->cache_obj->key, dobj->file_size, dconf->maxfs);
                /* Remove the intermediate cache file and return non-APR_SUCCESS */
                apr_pool_destroy(dobj->data.pool);
                return APR_EGENERAL;
            }

        }

        /* have we reached the limit of how much we're prepared to write in one
         * go? If so, leave, we'll get called again. This prevents us from trying
         * to swallow too much data at once, or taking so long to write the data
         * the client times out.
         */
        dobj->offset -= length;
        if (dobj->offset <= 0) {
            dobj->offset = 0;
            break;
        }
        if ((dconf->readtime && apr_time_now() > dobj->timeout)) {
            dobj->timeout = 0;
            break;
        }

    }

    /* Was this the final bucket? If yes, close the temp file and perform
     * sanity checks.
     */
    if (seen_eos) {
        const char *cl_header = apr_table_get(r->headers_out, "Content-Length");

        if (!dobj->disk_info.header_only) {

            if (dobj->data.tempfd) {
                rv = apr_file_close(dobj->data.tempfd);
                if (rv != APR_SUCCESS) {
                    /* Buffered write failed, abandon attempt to write */
                    apr_pool_destroy(dobj->data.pool);
                    return rv;
                }
            }

            if (r->connection->aborted || r->no_cache) {
                ap_log_rerror(
                        APLOG_MARK, APLOG_INFO, 0, r, APLOGNO(00733) "Discarding body for URL %s "
                        "because connection has been aborted.", h->cache_obj->key);
                /* Remove the intermediate cache file and return non-APR_SUCCESS */
                apr_pool_destroy(dobj->data.pool);
                return APR_EGENERAL;
            }
            if (dobj->file_size < dconf->minfs) {
                ap_log_rerror(
                        APLOG_MARK, APLOG_DEBUG, 0, r, APLOGNO(00734) "URL %s failed the size check "
                        "(%" APR_OFF_T_FMT "<%" APR_OFF_T_FMT ")", h->cache_obj->key, dobj->file_size, dconf->minfs);
                /* Remove the intermediate cache file and return non-APR_SUCCESS */
                apr_pool_destroy(dobj->data.pool);
                return APR_EGENERAL;
            }
            if (cl_header) {
                apr_int64_t cl = apr_atoi64(cl_header);
                if ((errno == 0) && (dobj->file_size != cl)) {
                    ap_log_rerror(
                            APLOG_MARK, APLOG_DEBUG, 0, r, APLOGNO(00735) "URL %s didn't receive complete response, not caching", h->cache_obj->key);
                    /* Remove the intermediate cache file and return non-APR_SUCCESS */
                    apr_pool_destroy(dobj->data.pool);
                    return APR_EGENERAL;
                }
            }

        }

        /* All checks were fine, we're good to go when the commit comes */
    }

    return APR_SUCCESS;
}

static apr_status_t commit_entity(cache_handle_t *h, request_rec *r)
{
    disk_cache_conf *conf = ap_get_module_config(r->server->module_config,
                                                 &cache_disk_module);
    disk_cache_object_t *dobj = (disk_cache_object_t *) h->cache_obj->vobj;
    apr_status_t rv;

    /* write the headers to disk at the last possible moment */
    rv = write_headers(h, r);

    /* move header and data tempfiles to the final destination */
    if (APR_SUCCESS == rv) {
        rv = file_cache_el_final(conf, &dobj->hdrs, r);
    }
    if (APR_SUCCESS == rv) {
        rv = file_cache_el_final(conf, &dobj->vary, r);
    }
    if (APR_SUCCESS == rv) {
        if (!dobj->disk_info.header_only) {
            rv = file_cache_el_final(conf, &dobj->data, r);
        }
        else if (dobj->data.file) {
            rv = apr_file_remove(dobj->data.file, dobj->data.pool);
        }
    }

    /* remove the cached items completely on any failure */
    if (APR_SUCCESS != rv) {
        remove_url(h, r);
        ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r, APLOGNO(00736)
                "commit_entity: URL '%s' not cached due to earlier disk error.",
                dobj->name);
    }
    else {
        ap_log_rerror(APLOG_MARK, APLOG_DEBUG, 0, r, APLOGNO(00737)
                "commit_entity: Headers and body for URL %s cached.",
                dobj->name);
    }

    apr_pool_destroy(dobj->data.pool);

    return APR_SUCCESS;
}

static apr_status_t invalidate_entity(cache_handle_t *h, request_rec *r)
{
    apr_status_t rv;

    rv = recall_headers(h, r);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    /* mark the entity as invalidated */
    h->cache_obj->info.control.invalidated = 1;

    return commit_entity(h, r);
}

static void *create_dir_config(apr_pool_t *p, char *dummy)
{
    disk_cache_dir_conf *dconf = apr_pcalloc(p, sizeof(disk_cache_dir_conf));

    dconf->maxfs = DEFAULT_MAX_FILE_SIZE;
    dconf->minfs = DEFAULT_MIN_FILE_SIZE;
    dconf->readsize = DEFAULT_READSIZE;
    dconf->readtime = DEFAULT_READTIME;

    return dconf;
}

static void *merge_dir_config(apr_pool_t *p, void *basev, void *addv)
{
    disk_cache_dir_conf *new = (disk_cache_dir_conf *) apr_pcalloc(p, sizeof(disk_cache_dir_conf));
    disk_cache_dir_conf *add = (disk_cache_dir_conf *) addv;
    disk_cache_dir_conf *base = (disk_cache_dir_conf *) basev;

    new->maxfs = (add->maxfs_set == 0) ? base->maxfs : add->maxfs;
    new->maxfs_set = add->maxfs_set || base->maxfs_set;
    new->minfs = (add->minfs_set == 0) ? base->minfs : add->minfs;
    new->minfs_set = add->minfs_set || base->minfs_set;
    new->readsize = (add->readsize_set == 0) ? base->readsize : add->readsize;
    new->readsize_set = add->readsize_set || base->readsize_set;
    new->readtime = (add->readtime_set == 0) ? base->readtime : add->readtime;
    new->readtime_set = add->readtime_set || base->readtime_set;

    return new;
}

static void *create_config(apr_pool_t *p, server_rec *s)
{
    disk_cache_conf *conf = apr_pcalloc(p, sizeof(disk_cache_conf));

    /* XXX: Set default values */
    conf->dirlevels = DEFAULT_DIRLEVELS;
    conf->dirlength = DEFAULT_DIRLENGTH;

    conf->cache_root = NULL;
    conf->cache_root_len = 0;

    return conf;
}

/*
 * mod_cache_disk configuration directives handlers.
 */
static const char
*set_cache_root(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_conf *conf = ap_get_module_config(parms->server->module_config,
                                                 &cache_disk_module);
    conf->cache_root = arg;
    conf->cache_root_len = strlen(arg);
    /* TODO: canonicalize cache_root and strip off any trailing slashes */

    return NULL;
}

/*
 * Consider eliminating the next two directives in favor of
 * Ian's prime number hash...
 * key = hash_fn( r->uri)
 * filename = "/key % prime1 /key %prime2/key %prime3"
 */
static const char
*set_cache_dirlevels(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_conf *conf = ap_get_module_config(parms->server->module_config,
                                                 &cache_disk_module);
    int val = atoi(arg);
    if (val < 1)
        return "CacheDirLevels value must be an integer greater than 0";
    if (val * conf->dirlength > CACHEFILE_LEN)
        return "CacheDirLevels*CacheDirLength value must not be higher than 20";
    conf->dirlevels = val;
    return NULL;
}
static const char
*set_cache_dirlength(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_conf *conf = ap_get_module_config(parms->server->module_config,
                                                 &cache_disk_module);
    int val = atoi(arg);
    if (val < 1)
        return "CacheDirLength value must be an integer greater than 0";
    if (val * conf->dirlevels > CACHEFILE_LEN)
        return "CacheDirLevels*CacheDirLength value must not be higher than 20";

    conf->dirlength = val;
    return NULL;
}

static const char
*set_cache_minfs(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_dir_conf *dconf = (disk_cache_dir_conf *)in_struct_ptr;

    if (apr_strtoff(&dconf->minfs, arg, NULL, 10) != APR_SUCCESS ||
            dconf->minfs < 0)
    {
        return "CacheMinFileSize argument must be a non-negative integer representing the min size of a file to cache in bytes.";
    }
    dconf->minfs_set = 1;
    return NULL;
}

static const char
*set_cache_maxfs(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_dir_conf *dconf = (disk_cache_dir_conf *)in_struct_ptr;

    if (apr_strtoff(&dconf->maxfs, arg, NULL, 10) != APR_SUCCESS ||
            dconf->maxfs < 0)
    {
        return "CacheMaxFileSize argument must be a non-negative integer representing the max size of a file to cache in bytes.";
    }
    dconf->maxfs_set = 1;
    return NULL;
}

static const char
*set_cache_readsize(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_dir_conf *dconf = (disk_cache_dir_conf *)in_struct_ptr;

    if (apr_strtoff(&dconf->readsize, arg, NULL, 10) != APR_SUCCESS ||
            dconf->readsize < 0)
    {
        return "CacheReadSize argument must be a non-negative integer representing the max amount of data to cache in go.";
    }
    dconf->readsize_set = 1;
    return NULL;
}

static const char
*set_cache_readtime(cmd_parms *parms, void *in_struct_ptr, const char *arg)
{
    disk_cache_dir_conf *dconf = (disk_cache_dir_conf *)in_struct_ptr;
    apr_off_t milliseconds;

    if (apr_strtoff(&milliseconds, arg, NULL, 10) != APR_SUCCESS ||
            milliseconds < 0)
    {
        return "CacheReadTime argument must be a non-negative integer representing the max amount of time taken to cache in go.";
    }
    dconf->readtime = apr_time_from_msec(milliseconds);
    dconf->readtime_set = 1;
    return NULL;
}

static const command_rec disk_cache_cmds[] =
{
    AP_INIT_TAKE1("CacheRoot", set_cache_root, NULL, RSRC_CONF,
                 "The directory to store cache files"),
    AP_INIT_TAKE1("CacheDirLevels", set_cache_dirlevels, NULL, RSRC_CONF,
                  "The number of levels of subdirectories in the cache"),
    AP_INIT_TAKE1("CacheDirLength", set_cache_dirlength, NULL, RSRC_CONF,
                  "The number of characters in subdirectory names"),
    AP_INIT_TAKE1("CacheMinFileSize", set_cache_minfs, NULL, RSRC_CONF | ACCESS_CONF,
                  "The minimum file size to cache a document"),
    AP_INIT_TAKE1("CacheMaxFileSize", set_cache_maxfs, NULL, RSRC_CONF | ACCESS_CONF,
                  "The maximum file size to cache a document"),
    AP_INIT_TAKE1("CacheReadSize", set_cache_readsize, NULL, RSRC_CONF | ACCESS_CONF,
                  "The maximum quantity of data to attempt to read and cache in one go"),
    AP_INIT_TAKE1("CacheReadTime", set_cache_readtime, NULL, RSRC_CONF | ACCESS_CONF,
                  "The maximum time taken to attempt to read and cache in go"),
    {NULL}
};

static const cache_provider cache_disk_provider =
{
    &remove_entity,
    &store_headers,
    &store_body,
    &recall_headers,
    &recall_body,
    &create_entity,
    &open_entity,
    &remove_url,
    &commit_entity,
    &invalidate_entity
};

static void disk_cache_register_hook(apr_pool_t *p)
{
    /* cache initializer */
    ap_register_provider(p, CACHE_PROVIDER_GROUP, "disk", "0",
                         &cache_disk_provider);
}

AP_DECLARE_MODULE(cache_disk) = {
    STANDARD20_MODULE_STUFF,
    create_dir_config,          /* create per-directory config structure */
    merge_dir_config,           /* merge per-directory config structures */
    create_config,              /* create per-server config structure */
    NULL,                       /* merge per-server config structures */
    disk_cache_cmds,            /* command apr_table_t */
    disk_cache_register_hook    /* register hooks */
};