make builtin a standalone dns_db implementation

instead of using the SDB API as a wrapper to register and
unregister and provide a call framework for the builtin databases,
this commit flattens things so that the builtin databases implement
dns_db directly.
This commit is contained in:
Evan Hunt 2023-02-09 23:21:57 -08:00
parent 603cdb6332
commit aea4f10817
2 changed files with 490 additions and 993 deletions

File diff suppressed because it is too large Load diff

View file

@ -23,146 +23,10 @@
#include <dns/clientinfo.h>
#include <dns/types.h>
/*
* Simple database implementation. Originally sdb.
* XXX: this will be moved to its own file later.
*/
/* Opaque types: */
/* A simple database. */
typedef struct sdb sdb_t;
typedef struct sdbimplementation sdbimplementation_t;
/* A simple database lookup in progress. */
typedef struct sdblookup sdblookup_t;
/* A simple database traversal in progress. */
typedef struct sdballnodes sdballnodes_t;
/* Callback function types */
typedef isc_result_t (*sdblookupfunc_t)(const dns_name_t *zone,
const dns_name_t *name, void *dbdata,
sdblookup_t *lookup,
dns_clientinfomethods_t *methods,
dns_clientinfo_t *clientinfo);
typedef isc_result_t (*sdbauthorityfunc_t)(const char *zone, void *dbdata,
sdblookup_t *);
typedef isc_result_t (*sdballnodesfunc_t)(const char *zone, void *dbdata,
sdballnodes_t *allnodes);
typedef isc_result_t (*sdbcreatefunc_t)(const char *zone, int argc, char **argv,
void *driverdata, void **dbdata);
typedef void (*sdbdestroyfunc_t)(const char *zone, void *driverdata,
void **dbdata);
typedef struct sdbmethods {
sdblookupfunc_t lookup;
sdbauthorityfunc_t authority;
sdbcreatefunc_t create;
sdbdestroyfunc_t destroy;
} sdbmethods_t;
/***
*** Functions
***/
#define DNS_SDBFLAG_DNS64 0x00000001U
isc_result_t
sdb_register(const char *drivername, const sdbmethods_t *methods,
void *driverdata, unsigned int flags, isc_mem_t *mctx,
sdbimplementation_t **sdbimp);
/*%<
* Register a simple database driver for the database type 'drivername',
* implemented by the functions in '*methods'.
*
* sdbimp must point to a NULL sdbimplementation_t pointer. That is,
* sdbimp != NULL && *sdbimp == NULL. It will be assigned a value that
* will later be used to identify the driver when deregistering it.
*
* The name server will perform lookups in the database by calling the
* function 'lookup', passing it a printable zone name 'zone', a printable
* domain name 'name', and a copy of the argument 'dbdata' that
* was potentially returned by the create function. The 'sdblookup_t'
* argument to 'lookup' and 'authority' is an opaque pointer to be passed to
* ns_sdb_putrr().
*
* The lookup function returns the lookup results to the name server
* by calling ns_sdb_putrr() once for each record found. On success,
* the return value of the lookup function should be ISC_R_SUCCESS.
* If the domain name 'name' does not exist, the lookup function should
* ISC_R_NOTFOUND. Any other return value is treated as an error.
*
* Lookups at the zone apex will cause the server to also call the
* function 'authority' (if non-NULL), which must provide an SOA record
* and NS records for the zone by calling ns_sdb_putrr() once for each of
* these records. The 'authority' function may be NULL if invoking
* the 'lookup' function on the zone apex will return SOA and NS records.
*
* The allnodes function, if non-NULL, fills in an opaque structure to be
* used by a database iterator. This allows the zone to be transferred.
* This may use a considerable amount of memory for large zones, and the
* zone transfer may not be fully RFC1035 compliant if the zone is
* frequently changed.
*
* The create function will be called for each zone configured
* into the name server using this database type. It can be used
* to create a "database object" containing zone specific data,
* which can make use of the database arguments specified in the
* name server configuration.
*
* The destroy function will be called to free the database object
* when its zone is destroyed.
*
* The create and destroy functions may be NULL.
*
* The lookup and authority functions are called with relative names
* rather than absolute names. The string "@" represents the zone apex.
*
* Rdata strings may include relative names. Be aware that absolute names
* must contain a trailing dot.
*/
void
sdb_unregister(sdbimplementation_t **sdbimp);
/*%<
* Removes the simple database driver from the list of registered database
* types. There must be no active databases of this type when this function
* is called.
*/
/*% See sdb_putradata() */
isc_result_t
sdb_putrr(sdblookup_t *lookup, const char *type, dns_ttl_t ttl,
const char *data);
/*
* Add a single resource record to the lookup structure to be
* returned in the query response. sdb_putrr() takes the
* resource record in master file text format as a null-terminated
* string, and sdb_putrdata() takes the raw RDATA in
* uncompressed wire format.
*/
isc_result_t
sdb_putrdata(sdblookup_t *lookup, dns_rdatatype_t type, dns_ttl_t ttl,
const unsigned char *rdata, unsigned int rdlen);
/*
* May be called from the 'authority' callback to simplify construction of
* an SOA record for 'zone'. It will provide a SOA listing 'mname' as as
* the primary server and 'rname' as the responsible person mailbox. It is
* the responsibility of the driver to increment the serial number between
* responses if necessary. All other SOA fields will have reasonable
* default values.
*/
isc_result_t
sdb_putsoa(sdblookup_t *lookup, const char *mname, const char *rname,
uint32_t serial);
/* Initialization functions for builtin zone databases */
isc_result_t
named_builtin_init(void);