Converting Modules from Apache 1.3 to Apache 2.0

This is a first attempt at writing the lessons I learned when trying to convert the mod_mmap_static module to Apache 2.0. It's by no means definitive and probably won't even be correct in some ways, but it's a start.

The easier changes ...

Cleanup Routines

These now need to be of type apr_status_t and return a value of that type. Normally the return value will be APR_SUCCESS unless there is some need to signal an error in the cleanup. Be aware that even though you signal an error not all code yet checks and acts upon the error.

Initialisation Routines

These should now be renamed to better signify where they sit in the overall process. So the name gets a small change from mmap_init to mmap_post_config. The arguments passed have undergone a radical change and now look like

  • apr_pool_t *p
  • apr_pool_t *plog
  • apr_pool_t *ptemp
  • server_rec *s

Data Types

A lot of the data types have been moved into the APR. This means that some have had a name change, such as the one shown above. The following is a brief list of some of the changes that you are likely to have to make.

  • pool becomes apr_pool_t
  • table becomes apr_table_t

The messier changes...

Register Hooks

The new architecture uses a series of hooks to provide for calling your functions. These you'll need to add to your module by way of a new function, static void register_hooks(void). The function is really reasonably straightforward once you understand what needs to be done. Each function that needs calling at some stage in the processing of a request needs to be registered, handlers do not. There are a number of phases where functions can be added, and for each you can specify with a high degree of control the relative order that the function will be called in.

This is the code that was added to mod_mmap_static:

static void register_hooks(void)
{
    static const char * const aszPre[]={ "http_core.c",NULL };
    ap_hook_post_config(mmap_post_config,NULL,NULL,HOOK_MIDDLE);
    ap_hook_translate_name(mmap_static_xlat,aszPre,NULL,HOOK_LAST);
};

This registers 2 functions that need to be called, one in the post_config stage (virtually every module will need this one) and one for the translate_name phase. note that while there are different function names the format of each is identical. So what is the format?

        ap_hook_<var class="calibre40">phase_name</var>(<var class="calibre40">function_name</var>,
        <var class="calibre40">predecessors</var>, <var class="calibre40">successors</var>, <var class="calibre40">position</var>);

There are 3 hook positions defined...

  • HOOK_FIRST
  • HOOK_MIDDLE
  • HOOK_LAST

To define the position you use the position and then modify it with the predecessors and successors. Each of the modifiers can be a list of functions that should be called, either before the function is run (predecessors) or after the function has run (successors).

In the mod_mmap_static case I didn't care about the post_config stage, but the mmap_static_xlat must be called after the core module had done it's name translation, hence the use of the aszPre to define a modifier to the position HOOK_LAST.

Module Definition

There are now a lot fewer stages to worry about when creating your module definition. The old defintion looked like

module MODULE_VAR_EXPORT <var class="calibre27">module_name</var>_module =
{
    STANDARD_MODULE_STUFF,
    /* initializer */
    /* dir config creater */
    /* dir merger --- default is to override */
    /* server config */
    /* merge server config */
    /* command handlers */
    /* handlers */
    /* filename translation */
    /* check_user_id */
    /* check auth */
    /* check access */
    /* type_checker */
    /* fixups */
    /* logger */
    /* header parser */
    /* child_init */
    /* child_exit */
    /* post read-request */
};

The new structure is a great deal simpler...

module MODULE_VAR_EXPORT <var class="calibre27">module_name</var>_module =
{
    STANDARD20_MODULE_STUFF,
    /* create per-directory config structures */
    /* merge per-directory config structures  */
    /* create per-server config structures    */
    /* merge per-server config structures     */
    /* command handlers */
    /* handlers */
    /* register hooks */
};

Some of these read directly across, some don't. I'll try to summarise what should be done below.

The stages that read directly across :

/* dir config creater */

/* create per-directory config structures */

/* server config */

/* create per-server config structures */

/* dir merger */

/* merge per-directory config structures */

/* merge server config */

/* merge per-server config structures */

/* command table */

/* command apr_table_t */

/* handlers */

/* handlers */

The remainder of the old functions should be registered as hooks. There are the following hook stages defined so far...

ap_hook_post_config

this is where the old _init routines get registered

ap_hook_http_method

retrieve the http method from a request. (legacy)

ap_hook_open_logs

open any specified logs

ap_hook_auth_checker

check if the resource requires authorization

ap_hook_access_checker

check for module-specific restrictions

ap_hook_check_user_id

check the user-id and password

ap_hook_default_port

retrieve the default port for the server

ap_hook_pre_connection

do any setup required just before processing, but after accepting

ap_hook_process_connection

run the correct protocol

ap_hook_child_init

call as soon as the child is started

ap_hook_create_request

??

ap_hook_fixups

last chance to modify things before generating content

ap_hook_handler

generate the content

ap_hook_header_parser

lets modules look at the headers, not used by most modules, because they use post_read_request for this

ap_hook_insert_filter

to insert filters into the filter chain

ap_hook_log_transaction

log information about the request

ap_hook_optional_fn_retrieve

retrieve any functions registered as optional

ap_hook_post_read_request

called after reading the request, before any other phase

ap_hook_quick_handler

called before any request processing, used by cache modules.

ap_hook_translate_name

translate the URI into a filename

ap_hook_type_checker

determine and/or set the doc type