The built-in configuration is actually used in two cases: first, when
the server is loaded (or reloaded), and second when
'rndc showconf -builtin' is called.
Considering the parsing of the builtin configuration is quick and does
not occur during exclusive mode, but the configuration tree takes
considerable memory space, the built-in configuration is no longer kept
in memory once it has been used; instead it is re-parsed on demand.
once the user configuration has been merged into the effective
configuration, it no longer needs to be accessed as a configuration
tree, but we still want to be able to show it with 'rndc showconf -user'.
because the recursive strucure of cfg_obj objects is fairly large, the
canonical text form is a fraction of the size of the configuration
tree, so we now save it in that form instead.
add a new rndc command to dump server configuration info:
- "rndc showconf -user" dumps the contents of named.conf
- "rndc showconf -builtin" dumps named_g_defaltconfig
- "rndc showconf -effective" dumps the effective configuration,
i.e., the merger of the builtin and the user configurations.
instead of using an opaque ns_cfgctx pointer to store the configuration
data to be used by addzone and modzone, there are now fields in the
dns_view object to store the view configuration and LMDB database
environment. the global configuration is now stored in the named_server
object, along with the ACL context.
instead of parsing the default configuration to extract the
geoip-directory value, we can just construct the value the same way
it's done for the defaults.
Since the effective configuration tree is a "merged" configuration tree
from the user and the default configurations, the effective configuration
provides a unique configuration tree used by apply_confiuration() to
configure the server.
However, there is one specific case where the configuration code needs
to differentiate whether the configuration originally came from the
default or the user configuration: the trust-anchors. This is because
the default trust-anchors _have_ to be those for the root zone, and the
one provided by the user can be for any zone. A check enforces this.
In order to keep this difference visible from the configuration code,
with a unique configuration tree, we now introduce a default-only
`builtin-trust-anchors` statement which holds the builtin root
trust-anchors. It can't be used from the user configuration (this would
raise an error), hence it is not documented.
The effective configuration now contains the actual named configuration
as a unified user/global single tree instance. We now provide this to
apply_configuration() instead of the user configuration.
The prefetch statement can be overriden by the user, but the user might
specify the prefetch without the trigger value, which needs to be
pulled from the default configuration. Handle this case by directly
getting the default value if needed from the default configuration when
building the effective configuration tree.
Also take care of keeping the values inside their bounds, and simplify
the server configuration code which then just have to read effective
configuration values.
Default dnssec-policies are not overridden by user-provided ones. Add
this specific case to make sure those are kept, and also ensure that the
default dnssec-policy is always in the first position (which is an
implicit requirement in the existing implementation).
Also simplify the server configuration code, as it only needs to build
the list of dnssec-policy based on the effective config list.
User specified views don't override default views. In particular, the
_bind/CH view is still active. However, the order is important: if the
user defines a foo/CH view, it must be able to override _bind/CH by
matching clients first (this is how the view is documented).
The server configuration code is now simpler; it only has to build the
views based on the effective view list, and only creates the _default
view if there are no explicit views created by the user.
Implement the specific rules of ACL inheritance when buiding the
effective configuration. As those rules are directly implemented in the
configuration tree, they are removed from `apply_configuation`.
Two programs need to be able to provide the effective configuration:
named (through a future rndc command) and named-checkconf (though a
future command line switch). In order to calculate the effective
configuration from named-checkconf, the default configuration
needs to be accessible.
The default configuration has now been moved from bin/named/config.c
into a common include directory for binaries: bin/include/defaultconfig.h.
Add a default "response-padding" option in the global defaults,
to disable the option, and simplify the configuration code so that
looks at the global defaults if the option is not set in named.conf.
This enables us to remove the 'optionmaps' variable in configure_view(),
which was used for options that only look in named.conf.
In the past, "dnssec-validation" was not looked up in the global
defaults unless "dnssec-enable" was true. "dnssec-enable" has been
obsolete for several years, but dnssec-validation was still being
configured in two steps. This commit removes the vestigial bits of
the old logic.
"max-cache-size default;" is allowed, according to the documentation
and the parser, but when it's configured, named crashes due to an
INSIST that the only legal string value is "unlimited". this has
been fied.
the configuration has also been simplified. previously, we checked for
max-cache-size in view and options, then determined whether to look in
the global default options based on whether the view had recursion set.
the default value set there was only applicable to views with recursion.
now, the default is an explicit "default", which affects views with
and without recursion in different ways.
the cfg type for "max-cache-size" has been changed from
cfg_type_sizeorpercent to cfg_type_maxcachesize.
Passing the JSON_C_TO_STRING_PRETTY flag to
json_object_to_json_string_ext() makes the latter produce prettified
JSON output. This results in a huge amount of redundant whitespace
being inserted into each HTTP response (whitespace amounts to about 40%
of the entire JSON payload).
The bandwidth cost can be amortized by enabling HTTP compression on the
client side ("Accept-Encoding: deflate"), but that does not affect the
size of data at rest.
Use the JSON_C_TO_STRING_PLAIN flag instead of JSON_C_TO_STRING_PRETTY
to minimize the size of JSON responses sent via the statistics channel.
External tools should be used for prettifying JSON data.
change_directory() now lives in libisccfg. when it was moved,
the logging behavior changed: previously it had been logged
by named only, in the general logging category, and without the
named.conf filename and line number. it was not logged by
named-checkconf. this behavior has now been restored.
- several functions that can no longer fail have been changed to
type void, and unnecessary 'cleanup' sections were removed
- renamed cfg_create_obj() to cfg_obj_create(), and cfg_create_tuple()
to cfg_tuple_create(), to match typical nomenclature.
- fixed a memory leak bug, in which an element could be removed
from a list in delete_zoneconf() without being freed. this has
been addressed by adding a cfg_list_unlink() function.
list elements are now allocated based on the list they will
be stored in, using the same mctx.
- the cfg_parser_create() and cfg_parser_destroy() calls are no
longer used outside parser.c, so they are now static functions
- cfg_parser_attach(), cfg_parser_reset(), and cfg_parser_setflags()
are no longer used at all, and have been removed.
- cfg_parser_mapadd() has been renamed for clarity to cfg_map_add().
Instead of (1) allocating a parser, (2) parsing a file/buffer then (3)
freeing the parser, the parser is now internally created/destroyed from
within the `cfg_parse_*` functions. This simplifies a lot the use cases,
especially around the error cases where the parser needs to be freed in
a cleanup goto.
The only trick was the parser callback mechanism, which would previously
have been set up between steps 1 and 2. Since it's never been used for
any purpose other than the "directory" option, the chdir call has now
been moved inside the parser and the generic callback mechanism has been
removed, replacing CFG_CLAUSEFLAG_CALLBACK with CFG_CLAUSEFLAG_CHDIR.
when configuring a zone, we can now save the zone's configuration
object in the zone itself by calling dns_zone_setcfg(). this can
then be used by "rndc showzone" to print the zone's configuration,
which is simpler than searching for it using the new-zones
configuration, and allows it to work even if "allow-new-zones"
is disabled.
Remove all global cfg_parser objects as well as shared parsers between
views to dynamically add zones. Instead, parser are transirently created
whenever needed.
Remove the global named defaults parser. Instead, a parser is created
during the execution time of named_config_parsedefaults(). This
simplifies the API (no parser to pass around) and the life-cycle of the
default configuration tree (it doesn't depends on a parser instance).
cfg_obj_t doesn't store a pointer to its a parser context anymore,
and does not depend on the parser's lifecycle. Instead, it stores a
reference to its own memory context (and in principle, each node
could have different memory context). This also slightly simplifies
the _destroy API as there is no need to pass a context through it
anymore.
if a zone reload is already in progress when 'rndc reload <zone>' is
run, currently the message returned in "zone reload queued", which
is correct, but it's identical to the message returned when a reload
was *not* in progress, so the user can't easily tell what happened.
a user could reload a zone twice and not realize that only one
reload actually took place.
this has been addressed by changing the message returned to
"zone reload was already queued".
a new result code ISC_R_LOADING has been added to signal this
condition, taking the place of ISC_R_RELOAD, which was obsolete
and has been removed.
With the code handling the "tkey-gssapi-credential" statement removed,
the named_tkeyctx_fromconfig() function can no longer fail. Update its
return type to void and revise its only call site accordingly. Clean up
the function's documentation. Declare the 's' helper variable only in
the scope it is used in to improve readability.
Since the "tkey-gssapi-credential" statement has been previously
deprecated, mark it as ancient and remove all code related to it:
- The code processing the "tkey-gssapi-credential" statement in the
configuration is the only user of the dst_gssapi_acquirecred() and
dst_gssapi_releasecred() functions, so remove them along with their
static helper functions and a backup definition of the
GSS_KRB5_MECHANISM macro.
- When calling gss_accept_sec_context(), pass GSS_C_NO_CREDENTIAL
instead of the credential acquired by gss_acquire_cred().
(Previously, NULL was passed when "tkey-gssapi-credential" was not
specified. Kerberos headers define GSS_C_NO_CREDENTIAL as
(gss_cred_id_t) 0, so the logic was effectively the same, but using
the GSS_C_NO_CREDENTIAL macro is more appropriate.) This renders
the 'cred' parameter for dst_gssapi_acceptctx() redundant, so remove
it from the prototype of the latter. (Contrary to what the
documentation for dst_gssapi_acceptctx() claims,
dst_gssapi_releasecred() does not need to subsequently be called to
free the GSS-API context; a dst_gssapi_deletectx() call in
gssapi_destroy() takes care of that when the dynamically generated
TSIG key is destroyed.)
- Remove the 'gsscred' member from struct dns_tkeyctx, along with its
related dns_gss_cred_id_t typedef.
Update the relevant sections of the ARM and code comments accordingly.
This makes the "tkey-gssapi-keytab" statement the only way to set up
GSS-TSIG in named.
Remove redundant code from bin/named/tkeyconf.c while at it.
The type `ns_pluginregister_ctx_t` was initially added to pass plugin
contextual data when the plugin is registered, but this is also now
passed into `plugin_check`. Furthermore, those various data are not
specific to the registration in particular. Rename the type into
`ns_pluginctx_t` for clarity.
This commit introduces a new type, ns_pluginregister_ctx_t,
which is passed to plugin_check() and plugin_register() in place of the
'source' parameter. The source value is now just part of the structure,
which also holds a pointer to the zone origin if the plugin is loaded at
a zone level.
This provides more contextual information, enabling the plugin to make
specific configuration decisions based on the name of the zone for which
it is loaded.
It's also flexible if more contextual data are needed in the future:
add a new field to ns_pluginregister_ctx_t, and new plugins can use
it without affecting compatibility with existing plugins.
When a re-configuration fails, `apply_configuration` flows jump to a
cleanup label and, at some point, leave the exclusive mode and cleanup
the viewlist. It looks fine as the viewlist is at this point only
locally known (if this is a configuration failure, this is the new view
list, if this is a success, this is the old list which has been swapped
out from the production list during the exclusive mode).
However, the view and zone initialization code enqueues job callbacks,
for instance from `dns_zone_setsigninginterval` (but there are others
cases) which will be called for the new views and zones after the
exclusive mode is over.
Depending where the configuration fails, those views and zones can be
half-configured, for instance a view might have an unfrozen resolver.
Hence, leaving the exclusive mode before cleaning up those views ans
zones will immediately called the previously enqueued callbacks and lead
to this reconfiguration-failure crash stack:
```
isc_assertion_failed
dns_resolver_createfetch
do_keyfetch
isc__async_cb
...
uv_run
loop_thread
thread_body
thread_run
start_thread
...
```
To avoid the problem, the views are now cleaned up before leaving the
exclusive mode (which also clean up the zones and enqueued callbacks).
As context, the bug was introduced by !10910 which moved the creation
(not configuration) of the view outsides of the exclusive mode. This is
a safe move (as at this point, the newly view are only known locally by
`apply_configuration`) but the re-order was wrong regarding the point
where the exclusive mode was ended (before the change, the exclusive
mode as always ended before the new view are detached).
ACL configuration context variables are inconsistently named as `actx`,
`ac`, or `aclconfctx`, which caused confusion during code reviews. This
commit renames all `cfg_aclconfctx_t` variables to `aclctx`, which is
short, consistent, and unambiguous.
With named-checkconf -k you can check your configuration including
checking the dnssec-policy keys against the configured keystores. If
there is a mismatch in the key files versus the policy, named-checkconf
will fail. This is useful for running before migrating to dnssec-policy.
For logging purposes, introduce a function that writes the identifying
information about a policy key into a string.
Allow a dnssec key to be initialized outside the keymgr code.
Add 'log_errors' to 'cfg_kasp_fromconfig' to avoid duplicate error
logs.
In order to have a (minimal) test ensuring we don't move back
`apply_configuration` subroutines which can be done before the exclusive
lock is taken, `APPLY_CONFIGURATION_SUBROUTINE_LOG` macro is added and
used for the few subroutines already extracted from the exclusive mode.
Those expected logs are added in `configloading` system test checks.
When the server is configured (inside `apply_configuration`) a client
TLS context cache is created and attached to the global server object.
It is then used by `configure_view` flow (and also during runtime though
the zone manager).
It is now created before the exclusive mode, and the swap of the
previous TLS cache ctx is done at the end of the exclusive mode, if
everything went well.
This allows us (among other follow-up changes) to move the
`configure_views` function outside of the exclusive mode.
The keystores initialization, the KASP list initialization as well as
the initialization of the view no longer depends of any data shared by
running "production" objects during re-configuration of the server. This
allows us to move those outside (before) the exclusive mode is taken.
`named_g_actconfctx` is a global variable holding the ACL configuration
context alive (in particular, to dynamically load zones). However, this
object is build once per configuration (early) and is used only inside
server.c `apply_configuration` flow. (Two exceptions: the shutdown flow,
still in server.c and plugin check flow, which doesn't need it, so it's
NULL in such case).
Instead of leaving this global publicly exposed, it is now part of the
`named_server_t` object. This allows us to clearly see that, when
reconfigureing the server, the new instance of the ACL context is known
only by the newly built object and not currently used by "production"
object; and will help to move move logic before the exclusive mode is
taken.
The other advantage is that the ACL configuration context can now be
built before the exclusive lock as well.
The kasplist (dnssec-policy defined in the builtin and global
configuration options) was built inside apply_configuration. This
commit extracts this logic into its separate function.
In order to make the view configuration independent of the global
`server` object, the newly built kasplist is now passed as parameter.
(This eventually will help to be able to configure the views outside of
the exclusive mode by limiting its dependency to the global
`server`/`named_g_server`).
When creating/configuring the view, the user-defined views are built and
set into the viewlist, then builtin-view inside the builtin_viewlist.
But there is no seperate logic applied to those two lists, and they are
immediately merged into viewlist right after. This commit removes this
intermediate list and add builtin-views directly into the main viewlist
instead.
In order to help splitting apply_configuration, the inline loops and bit
of logic around it for views creation and configuration, each of those
are now in a dedicatated function.
If the server is reloaded, new views are created and preexisting cache
is attached to those _but_ something goes wrong later, the previous
views are restored but the previous cache list is destroyed. This makes
the subsequent reload to drop the existing cache. This fixes it by
avoiding a mutation of the old cache list.
As we removed the ability to count nodes in the auxiliary trees (because
there are no auxiliary trees), we can also cleanup the API and
associated enum type (dns_dbtree_t).
if the dns_db_getoriginnode() call is not implemented, we can
fall back to running dns_db_findnode() on the database origin.
we now only implement getoriginnode directly in databases where
it's clearly faster than the fallback implementation would be.
the dns_db_findext and _findnodeext calls are extended versions
of dns_db_find and _findnode, which take additional arguments for
client information in order to support ECS. previously, database
implementations could support either API call, with cross-compatibility
so that, for example, dns_db_findext() could call a find implementation
if findext was not implemented, and dns_db_find() could call findext
if find was not implemented.
this has now been simplified. the find and findnodeext implementations
now support client info. all database implementations will now provide
these calls. implementations which do not support ECS will simply
ignore the clientinfo and clientinfomethods parameters.
this only affects the underlying implementation; callers will still
use the same interface. dns_db_find() and dns_db_findnode() are now
macros which pass NULL to the clientinfo parameters, so that callers
don't have to do so explicitly. dns_db_findext() and dns_db_findnodeext()
are still available for callers that do wish to pass clientinfo pointers.
