This "quiescent state based reclamation" module provides support for
the qp-trie module in dns/qp. It is a replacement for liburcu, written
without reference to the urcu source code, and in fact it works in a
significantly different way.
A few specifics of BIND make this variant of QSBR somewhat simpler:
* We can require that wait-free access to a qp-trie only happens in
an isc_loop callback. The loop provides a natural quiescent state,
after the callbacks are done, when no qp-trie access occurs.
* We can dispense with any API like rcu_synchronize(). In practice,
it takes far too long to wait for a grace period to elapse for each
write to a data structure.
* We use the idea of "phases" (aka epochs or eras) from EBR to
reduce the amount of bookkeeping needed to track memory that is no
longer needed, knowing that the qp-trie does most of that work
already.
I considered hazard pointers for safe memory reclamation. They have
more read-side overhead (updating the hazard pointers) and it wasn't
clear to me how to nicely schedule the cleanup work. Another
alternative, epoch-based reclamation, is designed for fine-grained
lock-free updates, so it needs some rethinking to work well with the
heavily read-biased design of the qp-trie. QSBR has the fastest read
side of the basic SMR algorithms (with no barriers), and fits well
into a libuv loop. More recent hybrid SMR algorithms do not appear to
have enough benefits to justify the extra complexity.
the isc_glob module was originally needed to support posix-style glob
processing on Windows, but is now just an unnecessary wrapper around
glob(3). this commit removes it.
the parser could crash when "include" specified an empty string in place
of the filename. this has been fixed by returning ISC_R_FILENOTFOUND
when the string length is 0.
Previously, the async job queue would use a locked-list (ISC_LIST).
With introduction of atomic stack (that has to be drained at once), we
could use it to remove some contention between the threads and simplify
the async queue.
Fortunately, the reverse order still works for us - instead of append
and tail/prev operation on the list, we are now using prepend and
head/next operation on the atomic stack.
Add a singly-linked stack that supports lock-free prepend and drain (to
empty the list and clean up its elements). Intended for use with QSBR
to collect objects that need safe memory reclamation, or any other user
that works with adding objects to the stack and then draining them in
one go like various work queues.
In <isc/atomic.h>, add an `atomic_ptr()` macro to make type
declarations a little less abominable, and clean up a duplicate
definition of `atomic_compare_exchange_strong_acq_rel()`
removed references in code comments, doc/dev documentation, etc, to
isc_task, isc_timer_reset(), and isc_timertype_inactive. also removed a
coccinelle patch related to isc_timer_reset() that was no longer needed.
move all dns_sdb code into bin/named/builtin.c, which is the
only place from which it's called.
(note this is temporary: later we'll refactor builtin so that it's a
standalone dns_db implementation on its own instead of using SDB
as a wrapper.)
move database attach/detach functions to db.c, instead of
requiring them to be implemented for every database type.
instead, they must implement a 'destroy' function that is
called when references go to zero.
this enables us to use ISC_REFCOUNT_IMPL for databases,
with detailed tracing enabled by setting DNS_DB_TRACE to 1.
SDB is currently (and foreseeably) only used by the named
builtin databases, so it only needs as much of its API as
those databases use.
- removed three flags defined for the SDB API that were always
set the same by builtin databases.
- there were two different types of lookup functions defined for
SDB, using slightly different function signatures. since backward
compatibility is no longer a concern, we can eliminate the 'lookup'
entry point and rename 'lookup2' to 'lookup'.
- removed the 'allnodes' entry point and all database iterator
implementation code
- removed dns_sdb_putnamedrr() and dns_sdb_putnamedrdata() since
they were never used.
initialize dns_dbmethods, dns_sdbmethods and dns_rdatasetmethods
using explicit struct member names, so we don't have to keep track
of NULLs for unimplemented functions any longer.
some dns_db functions would have crashed if the DB implementation failed
to implement them, requiring the implementations to add functions that
did nothing but return ISC_R_NOTIMPLEMENTED or some obvious default
value. we can just have the dns_db wrapper functions themselves return
those values, and clean up the implementations accordingly.
make the private isc__rdatalist_* functions public dns_rdatalist
functions so that all the rdatalist primitives can be used by
callers to libdns. (this will be needed later for moving SDB and
SDLZ out of libdns.)
This was causing 'CID 436299: Null pointer dereferences (REVERSE_INULL)'
in Coverity. Also removed an 'INSIST(fctx != NULL);' that should
no longer be needed.
this function was just a front-end for gethostname(). it was
needed when we supported windows, which has a different function
for looking up the hostname; it's not needed any longer.
as every validator function is loop-synchronized, it should no longer be
necessary to use a validator lock.
calling dns_validator_send(), dns_validator_cancel() or
dns_validator_destroy() from a thread other than the one on which the
validator is running will now cause an assertion failure; this should be
fine since the validator and resolver are tightly coupled, and the fetch
contexts and validators run in the same loops.
refactor validator so that the validation status object (previously
called dns_valstatus_t, which was derived from dns_validatorevent_t), is
now part of the dns_validator object. when calling validator callbacks,
the validator itself is now sent as the argument.
(note: this necessitates caution in the callback functions that are
internal to validator.c validators spawn other validators, and it can be
confusing at times whether we need to be looking at val, val->subvalidator,
or val->parent.)
There was a code flow error that would remove the expired ADB entry from
the LRU list and then a check in the expire_entry() would cause
assertion error because it expect the ADB entry to be linked.
Additionally, the expire mechanism would loop for cases when we would
held only a read rwlock; in such case we need to upgrade the lock and
try again, not just try again.
as there is no further use of isc_task in BIND, this commit removes
it, along with isc_taskmgr, isc_event, and all other related types.
functions that accepted taskmgr as a parameter have been cleaned up.
as a result of this change, some functions can no longer fail, so
they've been changed to type void, and their callers have been
updated accordingly.
the tasks table has been removed from the statistics channel and
the stats version has been updated. dns_dyndbctx has been changed
to reference the loopmgr instead of taskmgr, and DNS_DYNDB_VERSION
has been udpated as well.
change functions using isc_taskmgr_beginexclusive() to use
isc_loopmgr_pause() instead.
also, removed an unnecessary use of exclusive mode in
named_server_tcptimeouts().
most functions that were implemented as task events because they needed
to be running in a task to use exclusive mode have now been changed
into loop callbacks instead. (the exception is catz, which is being
changed in a separate commit because it's a particularly complex change.)
callback events from dns_resolver_createfetch() are now posted
using isc_async_run.
other modules which called the resolver and maintained task/taskmgr
objects for this purpose have been cleaned up.
The callbacks from dns_abd_createfind() are now posted using
isc_async_run() instead of isc_task_send(). ADB event types
have been replaced with a new dns_adbstatus_t type which is
included as find->status.
(The ADB still uses a task for dns_resolver_createfetch().)
dns_request_create() and _createraw() now take a 'loop' parameter
and run the callback event on the specified loop.
as the task manager is no longer used, it has been removed from
the dns_requestmgr structure. the dns_resolver_taskmgr() function
is also no longer used and has been removed.
Replace the isc_mutex in the dns_adb unit with isc_rwlock for better
performance. Both ADB names and ADB entries hashtables and LRU are now
using isc_rwlock.
This changes the internal isc_rwlock implementation to:
Irina Calciu, Dave Dice, Yossi Lev, Victor Luchangco, Virendra
J. Marathe, and Nir Shavit. 2013. NUMA-aware reader-writer locks.
SIGPLAN Not. 48, 8 (August 2013), 157–166.
DOI:https://doi.org/10.1145/2517327.24425
(The full article available from:
http://mcg.cs.tau.ac.il/papers/ppopp2013-rwlocks.pdf)
The implementation is based on the The Writer-Preference Lock (C-RW-WP)
variant (see the 3.4 section of the paper for the rationale).
The implemented algorithm has been modified for simplicity and for usage
patterns in rbtdb.c.
The changes compared to the original algorithm:
* We haven't implemented the cohort locks because that would require a
knowledge of NUMA nodes, instead a simple atomic_bool is used as
synchronization point for writer lock.
* The per-thread reader counters are not being used - this would
require the internal thread id (isc_tid_v) to be always initialized,
even in the utilities; the change has a slight performance penalty,
so we might revisit this change in the future. However, this change
also saves a lot of memory, because cache-line aligned counters were
used, so on 32-core machine, the rwlock would be 4096+ bytes big.
* The readers use a writer_barrier that will raise after a while when
readers lock can't be acquired to prevent readers starvation.
* Separate ingress and egress readers counters queues to reduce both
inter and intra-thread contention.
The dns_badcache was pulling the <isc/atomic.h> header only indirectly
via <isc/rwlock.h>, add the direct include as the <isc/rwlock.h> no
longer pulls the header when pthread_rwlock is used.
I misunderstood the purpose of the `heap_index` rdataset header
member; I thought it identified which heap to use, and could therefore
be smaller, the same size as `locknum` indexes. But in fact it is a
position within a heap, so it needs to be able to count up to the
total number of rdatasets in the rbtdb.
So this changes `heap_index` from `uint16_t` back to `unsigned int`.
To avoid re-embiggening the rdatasetheader, shrink the `count` member
from `uint32` to `uint16`. The `count` is used to rotate RRsets in
`dns_rdataset_towiresorted()`, so 16 bits is more than large enough.
This change also means we no longer need to avoid colliding with
`DNS_RDATASET_COUNT_UNDEFINED` i.e. UINT32_MAX.
Closes#3862
