htcache.c

Open VXI. This is a open source.

    	    if (!anchor) anchor = HTRequest_anchor(request);
	    HTCache_touch(request, response, anchor);
	} else {
	    HTParentAnchor * anchor = HTRequest_anchor(request);
	    char * default_name = HTRequest_defaultPutName(request);
	    HTCache * cache = HTCache_find(anchor, default_name);
	    if (cache) {
		/* 
		** If we receive a 204 and the method is unsafe then we have
		** to delete the cache body but not the header information
		*/
		/*
		** @@ JK: see how to add all the methods here (201, etc.) when
		** it's a PUT
		*/
		if (status == 204) {
		    HTCache_updateMeta(cache, request, response);
		    cache->size = 0;
		    cache->range = YES;
		    /* @@ JK: update the cache meta data on disk */
		    HTCache_writeMeta (cache, request, response);
		    /* @@ JK: and we remove the file name as it's obsolete 
		       now */
		    REMOVE(cache->cachename);
		} else
		    HTCache_remove(cache);
	    } 
	    /* @@  this operation will clear the cache->size and cache_range */
	    HTCache_touch(request, response, anchor);
	}
    }
    return HT_OK;
}

/*
**  Set the size of a cached object. We don't consider the metainformation as
**  part of the size which is the the reason for why the min cache size should
**  not be less than 5M. When we set the cache size we also check whether we 
**  should run the gc or not.
*/
PRIVATE BOOL HTCache_setSize (HTCache * cache, long written, BOOL append)
{
    if (cache) {
	/*
	**  First look to see if we already have registered this cache entry
	**  with a certain size. This size may be a subpart of the total entity
	**  (in case the download was interrupted)
	*/
	if (cache->size > 0 && !append) HTCacheContentSize -= cache->size;
	cache->size = written;
	HTCacheContentSize += written;

	/*
	**  Now add the new size to the total cache size. If the new size is
	**  bigger than the legal cache size then start the gc.
	*/
	HTTRACE(CACHE_TRACE, "Cache....... Total size %ld\n" _ HTCacheContentSize);
	if (startGC()) HTCacheGarbage();
	return YES;
    }
    return NO;
}

/*
**  Verifies if a cache object exists for this URL and if so returns a URL
**  for the cached object. It does not verify whether the object is valid or
**  not, for example it might have expired.
**
**  Returns: file name	If OK (must be freed by caller)
**	     NULL	If no cache object found
*/
PUBLIC HTCache * HTCache_find (HTParentAnchor * anchor, char * default_name)
{
    HTList * list = NULL;
    HTCache * pres = NULL;

    /* Find a hash entry for this URL */
    if (HTCacheMode_enabled() && anchor && CacheTable) {
	char * url = NULL;
	int hash = 0;
	char * ptr;

	if (default_name)
	    StrAllocCopy (url, default_name);
	  else
	    url = HTAnchor_address((HTAnchor *) anchor);
	ptr = url;

	for (; *ptr; ptr++)
	    hash = (int) ((hash * 3 + (*(unsigned char *) ptr)) % HT_XL_HASH_SIZE);
	if (!CacheTable[hash]) {
	    HT_FREE(url);
	    return NULL;
	}
	list = CacheTable[hash];

	/* Search the cache */
	{
	    HTList * cur = list;
	    while ((pres = (HTCache *) HTList_nextObject(cur))) {
		if (!strcmp(pres->url, url)) {
		    HTTRACE(CACHE_TRACE, "Cache....... Found %p hits %d\n" _ 
					     pres _ pres->hits);
		    break;
		}
	    }
	}
	HT_FREE(url);
    }
    return pres;
}

/*	HTCache_delete
**	--------------
**	Deletes a cache entry
*/
PRIVATE BOOL HTCache_delete (HTCache * cache)
{
    if (cache && CacheTable) {
	HTList * cur = CacheTable[cache->hash];
	return cur && delete_object(cur, cache);
    }
    return NO;
}

/*	HTCache_deleteAll
**	-----------------
**	Destroys all cache entried in memory but does not write anything to
**	disk
*/
PUBLIC BOOL HTCache_deleteAll (void)
{
    if (CacheTable) {
	HTList * cur;
	int cnt;

	/* Delete the rest */
	for (cnt=0; cnt<HT_XL_HASH_SIZE; cnt++) {
	    if ((cur = CacheTable[cnt])) { 
		HTCache * pres;
		while ((pres = (HTCache *) HTList_nextObject(cur)) != NULL)
		    free_object(pres);
	    }
	    HTList_delete(CacheTable[cnt]);
	}
	HT_FREE(CacheTable);
	HTCacheContentSize = 0L;
	return YES;
    }
    return NO;
}

/*
**  Is we have a valid entry in the cache then we also need a location
**  where we can get it. Hopefully, we may be able to access it
**  thourgh one of our protocol modules, for example the local file
**  module. The name returned is in URL syntax and must be freed by
**  the caller
*/
PRIVATE char * HTCache_metaLocation (HTCache * cache)
{
    char * local = NULL;
    if (cache && cache->cachename && *cache->cachename) {
	if ((local = (char *) HT_MALLOC(strlen(cache->cachename) +
					strlen(HT_CACHE_META) + 5)) == NULL)
	    HT_OUTOFMEM("HTCache_metaLocation");
	sprintf(local, "%s%s", cache->cachename, HT_CACHE_META);
    }
    return local;
}

/*
**  Walk through the set of headers and write those out that we are allowed
**  to store in the cache. We look into the connection header to see what the 
**  hop-by-hop headers are and also into the cache-control directive to see what
**  headers should not be cached.
*/
PRIVATE BOOL meta_write (FILE * fp, HTRequest * request, HTResponse * response)
{
    if (fp && request && response) {
	HTAssocList * headers = HTAnchor_header(HTRequest_anchor(request));
	HTAssocList * connection = HTResponse_connection(response);
	char * nocache = HTResponse_noCache(response);

	/*
	**  If we don't have any headers then just return now.
	*/
	if (!headers) return NO;

	/*
	**  Check whether either the connection header or the cache control
	**  header includes header names that we should not cache
	*/
	if (connection || nocache) {

	    /*
	    **  Walk though the cache control no-cache directive
	    */
	    if (nocache) {
		char * line = NULL;
		char * ptr = NULL;
		char * field = NULL;
		StrAllocCopy(line, nocache);		 /* Get our own copy */
		ptr = line;
		while ((field = HTNextField(&ptr)) != NULL)
		    HTAssocList_removeObject(headers, field);
		HT_FREE(line);
	    }

	    /*
	    **  Walk though the connection header
	    */
	    if (connection) {
		HTAssoc * pres;
		while ((pres=(HTAssoc *) HTAssocList_nextObject(connection))) {
		    char * field = HTAssoc_name(pres);
		    HTAssocList_removeObject(headers, field);
		}
	    }
	}

	/*
	**  Write out the remaining list of headers that we not already store
	**  in the index file.
	*/
	{
	    HTAssocList * cur = headers;
	    HTAssoc * pres;
	    while ((pres = (HTAssoc *) HTAssocList_nextObject(cur))) {
		char * name = HTAssoc_name(pres);

		/* Don't write the headers that are already hop-by-hop */
		if (strcasecomp(name, "connection") &&
		    strcasecomp(name, "keep-alive") &&
		    strcasecomp(name, "proxy-authenticate") &&
		    strcasecomp(name, "proxy-authorization") &&
		    strcasecomp(name, "transfer-encoding") &&
		    strcasecomp(name, "upgrade")) {
		    if (fprintf(fp, "%s: %s\n", name, HTAssoc_value(pres)) < 0) {
			HTTRACE(CACHE_TRACE, "Cache....... Error writing metainfo\n");
			return NO;
		    }
		}
	    }
	}

	/*
	**  Terminate the header with a newline
	*/
	if (fprintf(fp, "\n") < 0) {
	    HTTRACE(CACHE_TRACE, "Cache....... Error writing metainfo\n");
	    return NO;
	}
	return YES;
    }
    return NO;
}

/*
**  Save the metainformation for the data object. If no headers
**  are available then the meta file is empty
*/
PUBLIC BOOL HTCache_writeMeta (HTCache * cache, HTRequest * request,
			       HTResponse * response)
{
    if (cache && request && response) {
	BOOL status;
	FILE * fp;
	char * name = HTCache_metaLocation(cache);
	if (!name) {
	    HTTRACE(CACHE_TRACE, "Cache....... Invalid cache entry\n");
	    HTCache_remove(cache);
	    return NO;
	}
	if ((fp = fopen(name, "wb")) == NULL) {
	    HTTRACE(CACHE_TRACE, "Cache....... Can't open `%s\' for writing\n" _ name);
	    HTCache_remove(cache);
	    HT_FREE(name);	    
	    return NO;
	}
	status = meta_write(fp, request, response);
	fclose(fp);
	HT_FREE(name);
	return status;
    }
    return NO;
}

PRIVATE BOOL meta_read (FILE * fp, HTRequest * request, HTStream * target)
{
    if (fp && request && target) {
	int status;
	char buffer[512];
	while (1) {
	    if ((status = fread(buffer, 1, 512, fp)) <= 0) {
		HTTRACE(PROT_TRACE, "Cache....... Meta information loaded\n");
		return YES;
	    }
	
	    /* Send the data down the pipe */
	    status = (*target->isa->put_block)(target, buffer, status);

	    if (status == HT_LOADED) {
		(*target->isa->flush)(target);
		return YES;
	    }
	    if (status < 0) {
		HTTRACE(PROT_TRACE, "Cache....... Target ERROR %d\n" _ status);
		break;
	    }
	}
    }
    return NO;
}

/*
**  Read the metainformation for the data object. If no headers are
**  available then the meta file is empty
*/
PRIVATE BOOL HTCache_readMeta (HTCache * cache, HTRequest * request)
{
    HTParentAnchor * anchor = HTRequest_anchor(request);
    if (cache && request && anchor) {
	BOOL status;
	FILE * fp;
	char * name = HTCache_metaLocation(cache);
	if (!name) {
	    HTTRACE(CACHE_TRACE, "Cache....... Invalid meta name\n" _ name);
	    HTCache_remove(cache);
	    return NO;
	}
	HTTRACE(CACHE_TRACE, "Cache....... Looking for `%s\'\n" _ name);
	if ((fp = fopen(name, "rb")) == NULL) {
	    HTTRACE(CACHE_TRACE, "Cache....... Can't open `%s\' for reading\n" _ name);
	    HTCache_remove(cache);
	    HT_FREE(name);	    
	} else {
	    HTStream * target = HTStreamStack(WWW_MIME_HEAD, WWW_DEBUG,
					      HTBlackHole(), request, NO);
	    /*
	    **  Make sure that we save the reponse information in the anchor
	    */
	    HTResponse_setCachable(HTRequest_response(request), HT_CACHE_ALL);
	    status = meta_read(fp, request, target);
	    (*target->isa->_free)(target);
	    /* JK: Moved the delete outside of meta_read, because it was being
	       deleted multiple times. 
	       Delete the response headers. In principle, they are
	       already available in the anchor */ 
	    HTRequest_setResponse(request, NULL);
	    fclose(fp);
	    HT_FREE(name);
	    return status;
	}
    }
    return NO;
}

/*
**  Merge metainformation with existing version. This means that we have had a
**  successful validation and hence a true cache hit. We only regard the
**  following headers: Date, content-location, expires, cache-control, and vary.
*/
PUBLIC BOOL HTCache_updateMeta (HTCache * cache, HTRequest * request,
				HTResponse * response)
{
    if (cache && request && response) {
	HTParentAnchor * anchor = HTRequest_anchor(request);
	cache->hits++;

	/* Calculate the various times */
	calculate_time(cache, request, response);

	/* Get the last-modified and etag values if any */
	{
	    char * etag = HTAnchor_etag(anchor);
	    if (etag) StrAllocCopy(cache->etag, etag);
	    cache->lm = HTAnchor_lastModified(anchor);
	}

	/* Must we revalidate this every time? */
	cache->must_revalidate = HTResponse_mustRevalidate(response);

	return YES;
    }
    return NO;
}

/*
**  Remove from disk. You must explicitly remove a lock
**  before this operation can succeed
*/
PRIVATE BOOL flush_object (HTCache * cache)
{
    if (cache && !HTCache_hasLock(cache)) {
	char * head = HTCache_metaLocation(cache);
	REMOVE(head);
	HT_FREE(head);
	REMOVE(cache->cachename);
	return YES;
    }
    return NO;
}

/*	HTCache_flushAll
**	----------------
**	Destroys all cache entried in memory and on disk. Resets the cache
**	to empty but the cache does not have to be reinitialized before we
**	can use it again.
*/
PUBLIC BOOL HTCache_flushAll (void)
{
    if (CacheTable) {
	HTList * cur;
	int cnt;

	/* Delete the rest */
	for (cnt=0; cnt<HT_XL_HASH_SIZE; cnt++) {
	    if ((cur = CacheTable[cnt])) { 
		HTCache * pres;
		while ((pres = (HTCache *) HTList_nextObject(cur)) != NULL) {
		    flush_object(pres);
		    free_object(pres);
		}
	    }
	    HTList_delete(CacheTable[cnt]);
	    CacheTable[cnt] = NULL;
	}

	/* Write the new empty index to disk */
	HTCacheIndex_write(HTCacheRoot);
	HTCacheContentSize = 0L;
	return YES;
    }
    return NO;
}

/*
**  This function checks whether a document has expired or not.
**  The check is based on the metainformation passed in the anchor object
**  The function returns the level of validation needed for getting a fresh
**  version. We also check the cache control directives in the request to
**  see if they change the freshness discission. 
*/
PUBLIC HTReload HTCache_isFresh (HTCache * cache, HTRequest * request)
{
    HTAssocList * cc = HTRequest_cacheControl(request);
    if (cache) {
	time_t max_age = -1;
	time_t max_stale = -1;
	time_t min_fresh = -1;

	/*
	**  Make sure that we have the metainformation loaded from the
	**  persistent cache
	*/
	HTParentAnchor * anchor = HTRequest_anchor(request);
	if (!HTAnchor_headerParsed(anchor)) {
	    if (HTCache_readMeta(cache, request) != YES)
		return HT_CACHE_ERROR;
	    HTAnchor_setHeaderParsed(anchor);
	}