This commit ensures that write callbacks are getting called only after
the data has been sent via the network.
Without this fix, a situation could appear when a write callback could
get called before the actual encrypted data would have been sent to
the network. Instead, it would get called right after it would have
been passed to the OpenSSL (i.e. encrypted).
Most likely, the issue does not reveal itself often because the
callback call was asynchronous, so in most cases it should have been
called after the data has been sent, but that was not guaranteed by
the code logic.
Also, this commit removes one memory allocation (netievent) from a hot
path, as there is no need to call this callback asynchronously
anymore.
In dns_adb_cancelfind(), we need to release the find lock and
then acquire the bucket and find locks in that order, for
consistency with locking hierarchy elsehwere. Previously we
were only acquiring the bucket lock.
Also rewrote the function for better readability.
The connect()ed UDP socket provides feedback on a variety of ICMP
errors (eg port unreachable) which bind can then use to decide what to
do with errors (report them to the client, try again with a different
nameserver etc). However, Linux's implementation does not report what
it considers "transient" conditions, which is defined as Destination
host Unreachable, Destination network unreachable, Source Route Failed
and Message Too Big.
Explicitly enable IP_RECVERR / IPV6_RECVERR (via libuv uv_udp_bind()
flag) to learn about ICMP destination network/host unreachable.
When we compile with libuv that has some capabilities via flags passed
to f.e. uv_udp_listen() or uv_udp_bind(), the call with such flags would
fail with invalid arguments when older libuv version is linked at the
runtime that doesn't understand the flag that was available at the
compile time.
Enforce minimal libuv version when flags have been available at the
compile time, but are not available at the runtime. This check is less
strict than enforcing the runtime libuv version to be same or higher
than compile time libuv version.
The interfacemgr and the .route was being detached while the network
manager had pending read from the socket. Instead of detaching from the
socket, we need to cancel the read which in turn will detach the route
socket and the associated interfacemgr.
Instead of checking if we need to re-seed for every isc_random call,
seed the random number generator in the libisc global initializer
and the per-thread initializer.
It used to require two 32-bit integer divisions to get a random number
less than some limit. Now we use Daniel Lemire's "nearly-divisionless"
algorithm for unbiased bounded random numbers, which requires one
64-bit integer multiply in the usual case, and one 32-bit integer
division in rare slow cases. Even the slow cases are faster than
before; there are also fewer branches.
I think this algorithm is exceptionally beautiful. It also has more
clever tricks than lines of code, so I have done my best to explain
how it works.
The dns__adb_attach() had an assertion failure that prevented to attach
to dns_adb if the dns_adb was shutting down. There was a race between
checking for .exiting in dns_adb_createfind and creating new_adbfind() -
other thread could have set the .exiting to true between the check.
Remove the assertion failure and allow attaching to dns_adb even while
shutting down. The process of dns_adb shutting down would be noticed
only a moments later when any other callback is called.
The rctx_chaseds() function calls dns_resolver_createfetch(), passing
fctx->task as the target task to run resume_dslookup() from. This
breaks task-based serialization of events as fctx->task is the task that
the dns_resolver_createfetch() caller wants to receive its fetch
completion event in; meanwhile, intermediate fetches started by the
resolver itself (e.g. related to QNAME minimization) must use
res->buckets[bucketnum].task instead. This discrepancy may cause
trouble if the resume_dslookup() callback happens to be run concurrently
with e.g. fctx_doshutdown().
Fix by passing the correct task to dns_resolver_createfetch() in
rctx_chaseds().
BIND 9 plugins are installed using Automake's pkglib_LTLIBRARIES stanza,
which causes the relevant shared objects to be placed in the
$(libdir)/@PACKAGE@/ directory, where @PACKAGE@ is expanded to the
lowercase form of the first argument passed to AC_INIT(), i.e. "bind".
Meanwhile, NAMED_PLUGINDIR - the preprocessor macro that the
ns_plugin_expandpath() function uses for determining the absolute path
to a plugin for which only a filename has been provided (rather than a
path) - is set to $(libdir)/named. This discrepancy breaks loading
plugins using just their filenames. Fix the issue (and also prevent it
from reoccurring) by setting NAMED_PLUGINDIR to $(pkglibdir).
Since version 5.0.0, decay-based purging is the only available dirty
page cleanup mechanism in jemalloc. It relies on so-called tickers,
which are simple data structures used for ensuring that certain actions
are taken "once every N times". Ticker data (state) is stored in a
thread-specific data structure called tsd in jemalloc parlance. Ticks
are triggered when extents are allocated and deallocated. Once every
1000 ticks, jemalloc attempts to release some of the dirty pages hanging
around (if any). This allows memory use to be kept in check over time.
This dirty page cleanup mechanism has a quirk. If the first
allocator-related action for a given thread is a free(), a
minimally-initialized tsd is set up which does not include ticker data.
When that thread subsequently calls *alloc(), the tsd transitions to its
nominal state, but due to a certain flag being set during minimal tsd
initialization, ticker data remains unallocated. This prevents
decay-based dirty page purging from working, effectively enabling memory
exhaustion over time. [1]
The quirk described above has been addressed (by moving ticker state to
a different structure) in jemalloc's development branch [2], but not in
any numbered jemalloc version released to date (the latest one being
5.2.1 as of this writing).
Work around the problem by ensuring that every thread spawned by
isc_thread_create() starts with a malloc() call. Avoid immediately
calling free() for the dummy allocation to prevent an optimizing
compiler from stripping away the malloc() + free() pair altogether.
An alternative implementation of this workaround was considered that
used a pair of isc_mem_create() + isc_mem_destroy() calls instead of
malloc() + free(), enabling the change to be fully contained within
isc__trampoline_run() (i.e. to not touch struct isc__trampoline), as the
compiler is not allowed to strip away arbitrary function calls.
However, that solution was eventually dismissed as it triggered
ThreadSanitizer reports when tools like dig, nsupdate, or rndc exited
abruptly without waiting for all worker threads to finish their work.
[1] https://github.com/jemalloc/jemalloc/issues/2251
[2] c259323ab3
dns_message_findname and dns_message_sectiontotext incorrectly accepted
DNS_SECTION_ANY. If DNS_SECTION_ANY was passed the section array could
be incorrectly accessed at (-1).
dns_message_pseudosectiontotext and dns_message_pseudosectiontoyaml
incorrectly accepted DNS_PSEUDOSECTION_ANY. These functions are
designed to process a single section.
When TASKMGR_TRACE=1 is defined, the task and event objects have
detailed tracing information about function, file, line, and
backtrace (to the extent tracked by gcc) where it was created.
At exit, when there are unfinished tasks, they will be printed along
with the detailed information.
The only place where isc_task_sendto() was used was in dns_resolver
unit, where the "sendto" part was actually no-op, because dns_resolver
uses bound tasks. Remove the isc_task_sendto() and
isc_task_sendtoanddetach() functions in favor of using bound tasks
create with isc_task_create_bound().
Additionally, cache the number of running netmgr threads (nworkers)
locally to reduce the number of function calls.
For some applications, it's useful to not listen on full battery of
threads. Add workers argument to all isc_nm_listen*() functions and
convenience ISC_NM_LISTEN_ONE and ISC_NM_LISTEN_ALL macros.
dns_rdata_fromtext and dns_rdata_fromwire now checks that there is
a valid name or oid at the start of the keydata when the key algorithm
is PRIVATEDNS and PRIVATEOID respectively.
dns_rdata_totext now prints out the oid if the algorithm is PRIVATEOID.
Catalog zones change of ownership is special mechanism to facilitate
controlled migration of a member zone from one catalog to another.
It is implemented using catalog zones property named "coo" and is
documented in DNS catalog zones draft version 5 document.
Implement the feature using a new hash table in the catalog zone
structure, which holds the added "coo" properties for the catalog zone
(containing the target catalog zone's name), and the key for the hash
table being the member zone's name for which the "coo" property is being
created.
Change some log messages to have consistent zone name quoting types.
Update the ARM with change of ownership documentation and usage
examples.
Add tests which check newly the added features.
When there are multiple record datasets in a database node of a catalog
zone, and BIND encounters a soft error during processing of a dataset,
it breaks from the loop and doesn't process the other datasets in the
node.
There are cases when this is not desired. For example, the catalog zones
draft version 5 states that there must be a TXT RRset named
`version.$CATZ` with exactly one RR, but it doesn't set a limitation
on possible non-TXT RRsets named `version.$CATZ` existing alongside
with the TXT one. In case when one exists, we will get a processing
error and will not continue the loop to process the TXT RRset coming
next.
Remove the "break" statement to continue processing all record datasets.
When processing a new or updated catalog zone, the record datasets
from the database are being processed in order. This creates a
problem because we need to know the version of the catalog zone
schema to process some of the records differently, but we do not
know the version until the 'version' record gets processed.
Find the 'version' record and process it first, only then iterate over
the database to process the rest, making sure not to process the
'version' record twice.
According to DNS catalog zones draft version 5 document, catalog
zone custom properties must be placed under the "ext" label.
Make necessary changes to support the new custom properties syntax in
catalog zones with version "2" of the schema.
Change the default catalog zones schema version from "1" to "2" in
ARM to prepare for the new features and changes which come starting
from this commit in order to support the latest DNS catalog zones draft
document.
Make some restructuring in ARM and rename the term catalog zone "option"
to "custom property" to better reflect the terms used in the draft.
Change the version of 'catalog1.zone.' catalog zone in the "catz" system
test to "2", and leave the version of 'catalog2.zone.' catalog zone at
version "1" to test both versions.
Add tests to check that the new syntax works only with the new schema
version, and that the old syntax works only with the legacy schema
version catalog zones.
Update the function that synchronizes the CDS and CDNSKEY DELETE
records. It now allows for the possibility that the CDS DELETE record
is published and the CDNSKEY DELETE record is not, and vice versa.
Also update the code in zone.c how 'dns_dnssec_syncdelete()' is called.
With KASP, we still maintain the DELETE records our self. Otherwise,
we publish the CDS and CDNSKEY DELETE record only if they are added
to the zone. We do still check if these records can be signed by a KSK.
This change will allow users to add a CDS and/or CDNSKEY DELETE record
manually, without BIND removing them on the next zone sign.
Note that this commit removes the check whether the key is a KSK, this
check is redundant because this check is also made in
'dst_key_is_signing()' when the role is set to DST_BOOL_KSK.
due to a typo in the code, ADB entries were unlinked from their entry
buckets during shutdown if they had a nonzero reference count. they
were only supposed to be unlinked if the reference count was exactly
one (that being the reference held by the bucket itself).
The mctx, zonetask and loadtask pools were being destroyed in the
shutdown function where in theory a dangling zone could be still
attached to it.
Move the isc_mem_put() on the pools to the destroy() function.
There are a couple of problems with dns_request_createvia(): a UDP
retry count of zero means unlimited retries (it should mean no
retries), and the overall request timeout is not enforced. The
combination of these bugs means that requests can be retried forever.
This change alters calls to dns_request_createvia() to avoid the
infinite retry bug by providing an explicit retry count. Previously,
the calls specified infinite retries and relied on the limit implied
by the overall request timeout and the UDP timeout (which did not work
because the overall timeout is not enforced). The `udpretries`
argument is also changed to be the number of retries; previously, zero
was interpreted as infinity because of an underflow to UINT_MAX, which
appeared to be a mistake. And `mdig` is updated to match the change in
retry accounting.
The bug could be triggered by zone maintenance queries, including
NOTIFY messages, DS parental checks, refresh SOA queries and stub zone
nameserver lookups. It could also occur with `nsupdate -r 0`.
(But `mdig` had its own code to avoid the bug.)
This commit makes use of isc_nmsocket_set_tlsctx(). Now, instead of
recreating TLS-enabled listeners (including the underlying TCP
listener sockets), only the TLS context in use is replaced.
This commit adds isc_nmsocket_set_tlsctx() - an asynchronous function
that replaces the TLS context within a given TLS-enabled listener
socket object. It is based on the newly added reference counting
functionality.
The intention of adding this function is to add functionality to
replace a TLS context without recreating the whole socket object,
including the underlying TCP listener socket, as a BIND process might
not have enough permissions to re-create it fully on reconfiguration.
The implementation is done on top of the reference counting
functionality found in OpenSSL/LibreSSL, which allows for avoiding
wrapping the object.
Adding this function allows using reference counting for TLS contexts
in BIND 9's codebase.
After some back and forth, it was decidede to match the configuration
option with unbound ("so-reuseport"), PowerDNS ("reuseport") and/or
nginx ("reuseport").
as far as I can determine the order of operations is not important.
*** CID 351372: Concurrent data access violations (ATOMICITY)
/lib/isc/timer.c: 227 in timer_purge()
221 LOCK(&timer->lock);
222 if (!purged) {
223 /*
224 * The event has already been executed, but not
225 * yet destroyed.
226 */
>>> CID 351372: Concurrent data access violations (ATOMICITY)
>>> Using an unreliable value of "event" inside the second locked section. If the data that "event" depends on was changed by another thread, this use might be incorrect.
227 timerevent_unlink(timer, event);
228 }
229 }
230 }
231
232 void
*** CID 351371: Null pointer dereferences (REVERSE_INULL)
/lib/dns/adb.c: 2615 in dns_adb_createfind()
2609 /*
2610 * Copy out error flags from the name structure into the find.
2611 */
2612 find->result_v4 = find_err_map[adbname->fetch_err];
2613 find->result_v6 = find_err_map[adbname->fetch6_err];
2614
>>> CID 351371: Null pointer dereferences (REVERSE_INULL)
>>> Null-checking "find" suggests that it may be null, but it has already been dereferenced on all paths leading to the check.
2615 if (find != NULL) {
2616 if (want_event) {
2617 INSIST((find->flags & DNS_ADBFIND_ADDRESSMASK) != 0);
2618 isc_task_attach(task, &(isc_task_t *){ NULL });
2619 find->event.ev_sender = task;
2620 find->event.ev_action = action;
Previously, HAVE_SO_REUSEPORT_LB has been defined only in the private
netmgr-int.h header file, making the configuration of load balanced
sockets inoperable.
Move the missing HAVE_SO_REUSEPORT_LB define the isc/netmgr.h and add
missing isc_nm_getloadbalancesockets() implementation.
Previously, the option to enable kernel load balancing of the sockets
was always enabled when supported by the operating system (SO_REUSEPORT
on Linux and SO_REUSEPORT_LB on FreeBSD).
It was reported that in scenarios where the networking threads are also
responsible for processing long-running tasks (like RPZ processing, CATZ
processing or large zone transfers), this could lead to intermitten
brownouts for some clients, because the thread assigned by the operating
system might be busy. In such scenarious, the overall performance would
be better served by threads competing over the sockets because the idle
threads can pick up the incoming traffic.
Add new configuration option (`load-balance-sockets`) to allow enabling
or disabling the load balancing of the sockets.
Previously, the RPZ updates ran quantized on the main nm_worker loops.
As the quantum was set to 1024, this might lead to service
interruptions when large RPZ update was processed.
Change the RPZ update process to run as the offloaded work. The update
and cleanup loops were refactored to do as little locking of the
maintenance lock as possible for the shortest periods of time and the db
iterator is being paused for every iteration, so we don't hold the rbtdb
tree lock for prolonged periods of time.
