Three small cleanups:
1. Remove an unused keystr/dst_key_format.
2. Initialize a dst_key_state_t state with NA.
3. Update false comment about local policy (local policy only adds
barrier on transitions to the RUMOURED state, not the UNRETENTIVE
state).
There was a bug in function 'keymgr_ds_hidden_or_chained()'.
The funcion 'keymgr_ds_hidden_or_chained()' implements (3e) of rule2
as defined in the "Flexible and Robust Key Rollover" paper. The rules
says: All DS records need to be in the HIDDEN state, or if it is not
there must be a key with its DNSKEY and KRRSIG in OMNIPRESENT, and
its DS in the same state as the key in question. In human langauge,
if all keys have their DS in HIDDEN state you can do what you want,
but if a DS record is available to some validators, there must be
a chain of trust for it.
Note that the barriers on transitions first check if the current
state is valid, and then if the next state is valid too. But
here we falsely updated the 'dnskey_omnipresent' (now 'dnskey_chained')
with the next state. The next state applies to 'key' not to the state
to be checked. Updating the state here leads to (true) always, because
the key that will move its state will match the falsely updated
expected state. This could lead to the assumption that Key 2 would be
a valid chain of trust for Key 1, while clearly the presence of any
DS is uncertain.
The fix here is to check if the DNSKEY and KRRSIG are in OMNIPRESENT
state for the key that does not have its DS in the HIDDEN state, and
only if that is not the case, ensure that there is a key with the same
algorithm, that provides a valid chain of trust, that is, has its
DNSKEY, KRRSIG, and DS in OMNIPRESENT state.
The changes in 'keymgr_dnskey_hidden_or_chained()' are only cosmetical,
renaming 'rrsig_omnipresent' to 'rrsig_chained' and removing the
redundant initialization of the DST_KEY_DNSKEY expected state to NA.
The previous commit changed the function definition of
'keymgr_key_is_successor()', this commit updates the code where
this function is called.
In 'keymgr_key_exists_with_state()' the logic is also updated slightly
to become more readable. First handle the easy cases:
- If the key does not match the state, continue with the next key.
- If we found a key with matching state, and there is no need to
check the successor relationship, return (true).
- Otherwise check the successor relationship.
In 'keymgr_key_has_successor()' it is enough to check if a key has
a direct successor, so instead of calling 'keymgr_key_is_successor()',
we can just check 'keymgr_direct_dep()'.
In 'dns_keymgr_run()', we want to make sure that there is no
dependency on the keys before retiring excess keys, so replace
'keymgr_key_is_successor()' with 'keymgr_dep()'.
So far the key manager could only deal with two keys in a rollover,
because it used a simplified version of the successor relationship
equation from "Flexible and Robust Key Rollover" paper. The simplified
version assumes only two keys take part in the key rollover and it
for that it is enough to check the direct relationship between two
keys (is key x the direct predecessor of key z and is key z the direct
successor of key x?).
But when a third key (or more keys) comes into the equation, the key
manager would assume that one key (or more) is redundant and removed
it from the zone prematurely.
Fix by implementing Equation(2) correctly, where we check for
dependencies on keys:
z ->T x: Dep(x, T) = ∅ ∧
(x ∈ Dep(z, T) ∨
∃ y ∈ Dep(z, T)(y != z ∧ y ->T x ∧ DyKyRySy = DzKzRzSz))
This says: key z is a successor of key x if:
- key x depends on key z if z is a direct successor of x,
- or if there is another key y that depends on key z that has identical
key states as key z and key y is a successor of key x.
- Also, key x may not have any other keys depending on it.
This is still a simplified version of Equation(2) (but at least much
better), because the paper allows for a set of keys to depend on a
key. This is defined as the set Dep(x, T). Keys in the set Dep(x, T)
have a dependency on key x for record type T. The BIND implementation
can only have one key in the set Dep(x, T). The function
'keymgr_dep()' stores this key in 'uint32_t *dep' if there is a
dependency.
There are two scenarios where multiple keys can depend on a single key:
1. Rolling keys is faster than the time required to finish the
rollover procedure. This scenario is covered by the recursive
implementation, and checking for a chain of direct dependencies
will suffice.
2. Changing the policy, when a zone is requested to be signed with
a different key length for example. BIND 9 will not mark successor
relationships in this case, but tries to move towards the new
policy. Since there is no successor relationship, the rules are
even more strict, and the DNSSEC reconfiguration is actually slower
than required.
Note: this commit breaks the build, because the function definition
of 'keymgr_key_is_successor' changed. This will be fixed in the
following commit.
*** CID 318094: Null pointer dereferences (REVERSE_INULL)
/lib/dns/rbtdb.c: 1389 in newversion()
1383 version->xfrsize = rbtdb->current_version->xfrsize;
1384 RWUNLOCK(&rbtdb->current_version->rwlock, isc_rwlocktype_read);
1385 rbtdb->next_serial++;
1386 rbtdb->future_version = version;
1387 RBTDB_UNLOCK(&rbtdb->lock, isc_rwlocktype_write);
1388
CID 318094: Null pointer dereferences (REVERSE_INULL)
Null-checking "version" suggests that it may be null, but it has already been dereferenced on all paths leading to the check.
1389 if (version == NULL) {
1390 return (result);
1391 }
1392
1393 *versionp = version;
1394
The 'key_init()' function is used to initialize a state file for keys
that don't have one yet. This can happen if you are migrating from a
'auto-dnssec' or 'inline-signing' to a 'dnssec-policy' configuration.
It did not look at the "Inactive" and "Delete" timing metadata and so
old keys left behind in the key directory would also be considered as
a possible active key. This commit fixes this and now explicitly sets
the key goal to OMNIPRESENT for keys that have their "Active/Publish"
timing metadata in the past, but their "Inactive/Delete" timing
metadata in the future. If the "Inactive/Delete" timing metadata is
also in the past, the key goal is set to HIDDEN.
If the "Inactive/Delete" timing metadata is in the past, also the
key states are adjusted to either UNRETENTIVE or HIDDEN, depending on
how far in the past the metadata is set.
Add support for a "tls" key/value pair for zone primaries, referencing
either a "tls" configuration statement or "ephemeral". If set to use
TLS, zones will send SOA and AXFR/IXFR queries over a TLS channel.
This commit fix a race that could happen when two or more threads have
failed to refresh the same RRset, the threads could simultaneously
attempt to update the header->last_refresh_fail_ts field in
check_stale_header, a field used to implement stale-refresh-time.
By making this field atomic we avoid such race.
If we did not attempt a fetch due to fetch-limits, we should not start
the stale-refresh-time window.
Introduce a new flag DNS_DBFIND_STALESTART to differentiate between
a resolver failure and unexpected error. If we are resuming, this
indicates a resolver failure, then start the stale-refresh-time window,
otherwise don't start the stale-refresh-time window, but still fall
back to using stale data.
(This commit also wraps some docstrings to 80 characters width)
When 'opensslecdsa_parse()' encounters a label tag in the private key
file, load the private key with 'opensslecdsa_fromlabel()'. Otherwise
load it from the private structure.
This was attempted before with 'load_privkey()' and 'uses_engine()',
but had the same flaw as 'opensslecdsa_fromlabel()' had previously,
that is getting the private and public key separately, juggling with
pointers between EC_KEY and EVP_PKEY, did not create a valid
cryptographic key that could be used for signing.
The 'opensslecdsa_fromlabel()' function does not need to get the
OpenSSL engine twice to load the private and public key. Also no need
to call 'dst_key_to_eckey()' as the EC_KEY can be derived from the
loaded EVP_PKEY's.
Add some extra checks to ensure the key has the same base id and curve
(group nid) as the dst key.
Since we already have the EVP_PKEY, no need to call 'finalize_eckey()',
instead just set the right values in the key structure.
The 'ecdsa_check()' function tries to correctly set the public key
on the eckey, but this should be skipped if the public key is
retrieved via the private key.
The functions 'load_pubkey_from_engine()' and
'load_privkey_from_engine()' did not correctly store the pointers.
Update both functions to add 'EC_KEY_set_public_key()' and
'EC_KEY_set_private_key()' respectively, so that the pointers to
the public and private keys survive the "load from engine" functions.
The 'function load_pubkey_from_engine()' made a call to the libssl
function 'ENGINE_load_private_key'. This is a copy paste error and
should be 'ENGINE_load_public_key'.
First of all, there was a flaw in the code related to the
'stale-refresh-time' option. If stale answers are enabled, and we
returned stale data, then it was assumed that it was because we were
in the 'stale-refresh-time' window. But now we could also have returned
stale data because of a 'stale-answer-client-timeout'. To fix this,
introduce a rdataset attribute DNS_RDATASETATTR_STALE_WINDOW to
indicate whether the stale cache entry was returned because the
'stale-refresh-time' window is active.
Second, remove the special case handling when the result is
DNS_R_NCACHENXRRSET. This can be done more generic in the code block
when dealing with stale data.
Putting all stale case handling in the code block when dealing with
stale data makes the code more easy to follow.
Update documentation to be more verbose and to match then new code
flow.
The general logic behind the addition of this new feature works as
folows:
When a client query arrives, the basic path (query.c / ns_query_recurse)
was to create a fetch, waiting for completion in fetch_callback.
With the introduction of stale-answer-client-timeout, a new event of
type DNS_EVENT_TRYSTALE may invoke fetch_callback, whenever stale
answers are enabled and the fetch took longer than
stale-answer-client-timeout to complete.
When an event of type DNS_EVENT_TRYSTALE triggers fetch_callback, we
must ensure that the folowing happens:
1. Setup a new query context with the sole purpose of looking up for
stale RRset only data, for that matters a new flag was added
'DNS_DBFIND_STALEONLY' used in database lookups.
. If a stale RRset is found, mark the original client query as
answered (with a new query attribute named NS_QUERYATTR_ANSWERED),
so when the fetch completion event is received later, we avoid
answering the client twice.
. If a stale RRset is not found, cleanup and wait for the normal
fetch completion event.
2. In ns_query_done, we must change this part:
/*
* If we're recursing then just return; the query will
* resume when recursion ends.
*/
if (RECURSING(qctx->client)) {
return (qctx->result);
}
To this:
if (RECURSING(qctx->client) && !QUERY_STALEONLY(qctx->client)) {
return (qctx->result);
}
Otherwise we would not proceed to answer the client if it happened
that a stale answer was found when looking up for stale only data.
When an event of type DNS_EVENT_FETCHDONE triggers fetch_callback, we
proceed as before, resuming query, updating stats, etc, but a few
exceptions had to be added, most important of which are two:
1. Before answering the client (ns_client_send), check if the query
wasn't already answered before.
2. Before detaching a client, e.g.
isc_nmhandle_detach(&client->reqhandle), ensure that this is the
fetch completion event, and not the one triggered due to
stale-answer-client-timeout, so a correct call would be:
if (!QUERY_STALEONLY(client)) {
isc_nmhandle_detach(&client->reqhandle);
}
Other than these notes, comments were added in code in attempt to make
these updates easier to follow.
This is a minor performance improvement, we store the result of the
first call to strlcat to use as an offset in the next call when
constructing fctx->info string.
The BIND 9 libraries are considered to be internal only and hence the
API and ABI changes a lot. Keeping track of the API/ABI changes takes
time and it's a complicated matter as the safest way to make everything
stable would be to bump any library in the dependency chain as in theory
if libns links with libdns, and a binary links with both, and we bump
the libdns SOVERSION, but not the libns SOVERSION, the old libns might
be loaded by binary pulling old libdns together with new libdns loaded
by the binary. The situation gets even more complicated with loading
the plugins that have been compiled with few versions old BIND 9
libraries and then dynamically loaded into the named.
We are picking the safest option possible and usable for internal
libraries - instead of using -version-info that has only a weak link to
BIND 9 version number, we are using -release libtool option that will
embed the corresponding BIND 9 version number into the library name.
That means that instead of libisc.so.1701 (as an example) the library
will now be named libisc-9.17.10.so.
* Following the example set in 634bdfb16d8, the tlsdns netmgr
module now uses libuv and SSL primitives directly, rather than
opening a TLS socket which opens a TCP socket, as the previous
model was difficult to debug. Closes#2335.
* Remove the netmgr tls layer (we will have to re-add it for DoH)
* Add isc_tls API to wrap the OpenSSL SSL_CTX object into libisc
library; move the OpenSSL initialization/deinitialization from dstapi
needed for OpenSSL 1.0.x to the isc_tls_{initialize,destroy}()
* Add couple of new shims needed for OpenSSL 1.0.x
* When LibreSSL is used, require at least version 2.7.0 that
has the best OpenSSL 1.1.x compatibility and auto init/deinit
* Enforce OpenSSL 1.1.x usage on Windows
* Added a TLSDNS unit test and implemented a simple TLSDNS echo
server and client.
It is possible to have two threads destroying an rbtdb at the same
time when detachnode() executes and removes the last reference to
a node between exiting being set to true for the node and testing
if the references are zero in maybe_free_rbtdb(). Move NODE_UNLOCK()
to after checking if references is zero to prevent detachnode()
changing the reference count too early.
While fixing #2359, 'report()' was changed so that it would print the
newline.
Newlines were missing from the output of 'dnssec-signzone'
and 'dnssec-verify' because change
664b8f04f5f2322086138f5eda5899a62bcc019b moved the printing from
newlines to the library.
This had to be reverted because this also would print redundant
newlines in logfiles.
While doing the revert, some newlines in 'lib/dns/zoneverify.c'
were left in place, now making 'dnssec-signzone' and 'dnssec-verify'
print too many newlines.
This commit removes those newlines, so that the output looks nice
again.
The keymgr prevented zones from going to insecure mode. If we
have a policy with an empty key list this is a signal that the zone
wants to go back to insecure mode. In this case allow one extra state
transition to be valid when checking for DNSSEC safety.
Configure "none" as a builtin policy. Change the 'cfg_kasp_fromconfig'
api so that the 'name' will determine what policy needs to be
configured.
When transitioning a zone from secure to insecure, there will be
cases when a zone with no DNSSEC policy (dnssec-policy none) should
be using KASP. When there are key state files available, this is an
indication that the zone once was DNSSEC signed but is reconfigured
to become insecure.
If we would not run the keymgr, named would abruptly remove the
DNSSEC records from the zone, making the zone bogus. Therefore,
change the code such that a zone will use kasp if there is a valid
dnssec-policy configured, or if there are state files available.
When using the `unixtime` or `date` method to update the SOA serial,
`named` and `dnssec-signzone` would silently fallback to `increment`
method to prevent the new serial number to be smaller than the old
serial number (using the serial number arithmetics). Add a warning
message when such fallback happens.
This is a part of the works that intends to make the netmgr stable,
testable, maintainable and tested. It contains a numerous changes to
the netmgr code and unfortunately, it was not possible to split this
into smaller chunks as the work here needs to be committed as a complete
works.
NOTE: There's a quite a lot of duplicated code between udp.c, tcp.c and
tcpdns.c and it should be a subject to refactoring in the future.
The changes that are included in this commit are listed here
(extensively, but not exclusively):
* The netmgr_test unit test was split into individual tests (udp_test,
tcp_test, tcpdns_test and newly added tcp_quota_test)
* The udp_test and tcp_test has been extended to allow programatic
failures from the libuv API. Unfortunately, we can't use cmocka
mock() and will_return(), so we emulate the behaviour with #define and
including the netmgr/{udp,tcp}.c source file directly.
* The netievents that we put on the nm queue have variable number of
members, out of these the isc_nmsocket_t and isc_nmhandle_t always
needs to be attached before enqueueing the netievent_<foo> and
detached after we have called the isc_nm_async_<foo> to ensure that
the socket (handle) doesn't disappear between scheduling the event and
actually executing the event.
* Cancelling the in-flight TCP connection using libuv requires to call
uv_close() on the original uv_tcp_t handle which just breaks too many
assumptions we have in the netmgr code. Instead of using uv_timer for
TCP connection timeouts, we use platform specific socket option.
* Fix the synchronization between {nm,async}_{listentcp,tcpconnect}
When isc_nm_listentcp() or isc_nm_tcpconnect() is called it was
waiting for socket to either end up with error (that path was fine) or
to be listening or connected using condition variable and mutex.
Several things could happen:
0. everything is ok
1. the waiting thread would miss the SIGNAL() - because the enqueued
event would be processed faster than we could start WAIT()ing.
In case the operation would end up with error, it would be ok, as
the error variable would be unchanged.
2. the waiting thread miss the sock->{connected,listening} = `true`
would be set to `false` in the tcp_{listen,connect}close_cb() as
the connection would be so short lived that the socket would be
closed before we could even start WAIT()ing
* The tcpdns has been converted to using libuv directly. Previously,
the tcpdns protocol used tcp protocol from netmgr, this proved to be
very complicated to understand, fix and make changes to. The new
tcpdns protocol is modeled in a similar way how tcp netmgr protocol.
Closes: #2194, #2283, #2318, #2266, #2034, #1920
* The tcp and tcpdns is now not using isc_uv_import/isc_uv_export to
pass accepted TCP sockets between netthreads, but instead (similar to
UDP) uses per netthread uv_loop listener. This greatly reduces the
complexity as the socket is always run in the associated nm and uv
loops, and we are also not touching the libuv internals.
There's an unfortunate side effect though, the new code requires
support for load-balanced sockets from the operating system for both
UDP and TCP (see #2137). If the operating system doesn't support the
load balanced sockets (either SO_REUSEPORT on Linux or SO_REUSEPORT_LB
on FreeBSD 12+), the number of netthreads is limited to 1.
* The netmgr has now two debugging #ifdefs:
1. Already existing NETMGR_TRACE prints any dangling nmsockets and
nmhandles before triggering assertion failure. This options would
reduce performance when enabled, but in theory, it could be enabled
on low-performance systems.
2. New NETMGR_TRACE_VERBOSE option has been added that enables
extensive netmgr logging that allows the software engineer to
precisely track any attach/detach operations on the nmsockets and
nmhandles. This is not suitable for any kind of production
machine, only for debugging.
* The tlsdns netmgr protocol has been split from the tcpdns and it still
uses the old method of stacking the netmgr boxes on top of each other.
We will have to refactor the tlsdns netmgr protocol to use the same
approach - build the stack using only libuv and openssl.
* Limit but not assert the tcp buffer size in tcp_alloc_cb
Closes: #2061
Since the queries sent towards root and TLD servers are now included in
the count (as a result of the fix for CVE-2020-8616),
"max-recursion-queries" has a higher chance of being exceeded by
non-attack queries. Increase its default value from 75 to 100.
Fallback to TCP when we have already seen a DNS COOKIE response
from the given address and don't have one in this UDP response. This
could be a server that has turned off DNS COOKIE support, a
misconfigured anycast server with partial DNS COOKIE support, or a
spoofed response. Falling back to TCP is the correct behaviour in
all 3 cases.
Make sure pointer checks in unit tests use cmocka assertion macros
dedicated for use with pointers instead of those dedicated for use with
integers or booleans.
