The first working multi-threaded qp-trie was stuck with an unpleasant
trade-off:
* Use `isc_rwlock`, which has acceptable write performance, but
terrible read scalability because the qp-trie made all accesses
through a single lock.
* Use `liburcu`, which has great read scalability, but terrible
write performance, because I was relying on `rcu_synchronize()`
which is rather slow. And `liburcu` is LGPL.
To get the best of both worlds, we need our own scalable read side,
which we now have with `isc_qsbr`. And we need to modify the write
side so that it is not blocked by readers.
Better write performance requires an async cleanup function like
`call_rcu()`, instead of the blocking `rcu_synchronize()`. (There
is no blocking cleanup in `isc_qsbr`, because I have concluded
that it would be an attractive nuisance.)
Until now, all my multithreading qp-trie designs have been based
around two versions, read-only and mutable. This is too few to
work with asynchronous cleanup. The bare minimum (as in epoch
based reclamation) is three, but it makes more sense to support an
arbitrary number. Doing multi-version support "properly" makes
fewer assumptions about how safe memory reclamation works, and it
makes snapshots and rollbacks simpler.
To avoid making the memory management even more complicated, I
have introduced a new kind of "packed reader node" to anchor the
root of a version of the trie. This is simpler because it re-uses
the existing chunk lifetime logic - see the discussion under
"packed reader nodes" in `qp_p.h`.
I have also made the chunk lifetime logic simpler. The idea of a
"generation" is gone; instead, chunks are either mutable or
immutable. And the QSBR phase number is used to indicate when a
chunk can be reclaimed.
Instead of the `shared_base` flag (which was basically a one-bit
reference count, with a two version limit) the base array now has a
refcount, which replaces the confusing ad-hoc lifetime logic with
something more familiar and systematic.
Adjust the dns_qp_memusage() and dns_qp_compact() functions
to be more informative and flexible about handling fragmentation.
Avoid wasting space in runt chunks.
Switch from twigs_mutable() to cells_immutable() because that is the
sense we usually want.
Drop the redundant evacuate() function and rename evacuate_twigs() to
evacuate(). Move some chunk test functions closer to their point of
use.
Clarify compact_recursive(). Some small cleanups to comments.
Use isc_time_monotonic() for qp-trie timing stats.
Use #define constants to control debug logging.
Set up DNS name label offsets in dns_qpkey_fromname() so it is easier
to use in cases where the name is not fully hydrated.
The main benchmark is `qpmulti`, which exercizes the qp-trie
transactional API with differing numbers of threads and differing data
sizes, to get some idea of how its performance scales.
The `load-names` benchmark compares the times to populate and query
and the memory used by various BIND data structures: qp-trie, hash
table (chained), hash map (closed), and red-black tree.
The `qp-dump` program is a test utility rather than a benchmark. It
populates a qp-trie and prints it out, either in an ad-hoc text
format, or as input to the graphviz `dot` program.
Ensure dns_qpkey_fromname() and dns_qpkey_toname() are inverses.
Excercise a single-threaded dns_qp_t with a fixed set of random keys
and a small chunk size. Use the table of names to ensure that the trie
is behaving as expected. This is (in effect) randomized testing like
the `qpmulti` unit test, but making use of coverage-guided fuzzing
and (in principle) test case minimization.
Randomized testing with intensive consistency and correctness checks
make it much easier to get good coverage and to shake out bugs than
hand-written unit tests for specific cases.
These tests only run in a single thread, but each test transaction
uses both a write/update and a query/snapshot, to ensure that
modifications are not visible to concurrent readers.
This change adds a number of support routines for the unit tests, and
for benchmarks and fuzz tests to be added later. It isn't necessary to
include the support routines in libdns, since they are not needed by
BIND's installed programs. So `libtest` seems like the best place for
them.
The tests themselves verify that dns_qpkey_fromname() behaves as
expected.
The error occurred when:
* The bump chunk was re-used across multiple write transactions.
In this situation the bump chunk is marked immutable, but the
immutable flag is disregarded for cells after the fender, which
were allocated in the current transaction.
* The bump chunk fills up during an insert operation, so that the
enlarged twigs vector is allocated from a new bump chunk.
* Before this happened, we should have (but didn't) made the twigs
vector mutable. This would have adjusted its refcounts as necessary.
* However, moving to a new bump chunk has a side effect: twigs that
were previously considered mutable because they are after the
fender become immutable.
* Because of this, the old twigs vector was not destroyed as expected.
* So leaves were duplicated without their refcounts being increased.
The effect is that the refcounts were lower than they should have
been, and underflowed. The tests failed to check for refcount
underflow, so this mistake was detected much later than it ideally
could have been.
After the fix, it is now correct not to ensure the twigs are mutable,
because they are about to be copied to a larger vector. Instead, we
need to find out whether `squash_twigs()` destroyed the old twigs, and
adjust the refcounts accordingly.
A qp-trie is a kind of radix tree that is particularly well-suited to
DNS servers. I invented the qp-trie in 2015, based on Dan Bernstein's
crit-bit trees and Phil Bagwell's HAMT. https://dotat.at/prog/qp/
This code incorporates some new ideas that I prototyped using
NLnet Labs NSD in 2020 (optimizations for DNS names as keys)
and 2021 (custom allocator and garbage collector).
https://dotat.at/cgi/git/nsd.git
The BIND version of my qp-trie code has a number of improvements
compared to the prototype developed for NSD.
* The main omission in the prototype was the very sketchy outline of
how locking might work. Now the locking has been implemented,
using a reader/writer lock and a mutex. However, it is designed to
benefit from liburcu if that is available.
* The prototype was designed for two-version concurrency, one
version for readers and one for the writer. The new code supports
multiversion concurrency, to provide a basis for BIND's dbversion
machinery, so that updates are not blocked by long-running zone
transfers.
* There are now two kinds of transaction that modify the trie: an
`update` aims to support many very small zones without wasting
memory; a `write` avoids unnecessary allocation to help the
performance of many small changes to the cache.
* There is also a single-threaded interface for situations where
concurrent access is not necessary.
* The API makes better use of types to make it more clear which
operations are permitted when.
* The lookup table used to convert a DNS name to a qp-trie key is
now initialized by a run-time constructor instead of a programmer
using copy-and-paste. Key conversion is more flexible, so the
qp-trie can be used with keys other than DNS names.
* There has been much refactoring and re-arranging things to improve
the terminology and order of presentation in the code, and the
internal documentation has been moved from a comment into a file
of its own.
Some of the required functionality has been stripped out, to be
brought back later after the basics are known to work.
* Garbage collector performance statistics are missing.
* Fancy searches are missing, such as longest match and
nearest match.
* Iteration is missing.
* Search for update is missing, for cases where the caller needs to
know if the value object is mutable or not.
Some qp-trie operations will need to know the maximum number of labels
in a name, so I wanted a standard macro definition with the right
value.
Replace DNS_MAX_LABELS from <dns/resolver.h with DNS_NAME_MAXLABELS in
<dns/name.h>, and add its counterpart DNS_NAME_LABELLEN.
Use these macros in `name.c` and `resolver.c`.
Fix an off-by-one error in an assertion in `dns_name_countlabels()`.
When detaching from the previous version of the database, make sure
that the update-notify callback is unregistered, otherwise there is
an INSIST check which can generate an assertion failure in free_rbtdb(),
which checks that there are no outstanding update listeners in the list.
There is a similar code already in place for RPZ.
The zone_postload() function can fail and unregister the callbacks.
Call dns_db_endload() only after calling zone_postload() to make
sure that the registered update-notify callbacks are not called
when the zone loading has failed during zone_postload().
Also, don't ignore the return value of zone_postload().
Add the 'ixfr-from-differences yes;' option to trigger a failed
zone postload operation when a zone is updated but the serial
number is not updated, then issue two successive 'rndc reload'
commands to trigger the bug, which causes an assertion failure.
The dbiterator read-locks the whole zone and it stayed locked during
whole processing time when catz is being read. Pause the iterator, so
the updates to catz zone are not being blocked while processing the catz
update.
Instead of holding the catzs->lock the whole time we process the catz
update, only hold it for hash table lookup and then release it. This
should unblock any other threads that might be processing updates to
catzs triggered by extra incoming transfer.
Offload catalog zone processing so that the network manager threads
are not interrupted by a large catalog zone update.
Introduce a new 'updaterunning' state alongside with 'updatepending',
like it is done in the RPZ module.
Note that the dns__catz_update_cb() function currently holds the
catzs->lock during the whole process, which is far from being optimal,
but the issue is going to be addressed separately.
This change should make sure that catalog zone update processing
doesn't happen when the catalog zone is being shut down. This
should help avoid races when offloading the catalog zone updates
in the follow-up commit.
The configure_catz() function creates the catalog zones structure
for the view even when it is not needed, in which case it then
discards it (by detaching) later.
Instead, call dns_catz_new_zones() only when it is needed, i.e. when
there is no existing "previous" view with an existing 'catzs', that
is going to be reused.
* Change 'dns_catz_new_zones()' function's prototype (the order of the
arguments) to synchronize it with the similar function in rpz.c.
* Rename 'refs' to 'references' in preparation of ISC_REFCOUNT_*
macros usage for reference tracking.
* Unify dns_catz_zone_t naming to catz, and dns_catz_zones_t naming to
catzs, following the logic of similar changes in rpz.c.
* Use C compound literals for structure initialization.
* Synchronize the "new zone version came too soon" log message with the
one in rpz.c.
* Use more of 'sizeof(*ptr)' style instead of the 'sizeof(type_t)' style
expressions when allocating or freeing memory for 'ptr'.
`libirs` used to be a reference implementation of `getaddrinfo` and
related modern resolver APIs. It was stripped down in BIND 9.18
leaving only the `irs_resconf` module, which parses
`/etc/resolv.conf`. I have kept its include path and namespace prefix,
so it remains a little fragment of libirs now embedded in libdns.
Add new SonarCloud GitHub Action and configuration; something (maybe
the way the builds were submitted) has apparently changed and the
project got deleted and the analysis wasn't working.
when a message arrives over a TCP connection matching an expected
QID, the dispatch is updated so it no longer expects that QID,
but continues reading. subsequent messages with the same QID are
ignored, unless the dispatch entry has called dns_dispatch_getnext()
or dns_dispatch_resume().
however, a coding error caused those functions to have no effect
when the dispatch was reading, so streams of messages with the same
QID could not be received over a single TCP connection, breaking *XFR.
this has been corrected by changing the order of operations in
tcp_dispatch_getnext() so that disp->reading isn't checked until
after the dispatch entry has been reactivated.
