Refactor a bit of `apply_configuration` by extracting (into respective dedicated function) the logic to build the keystores list, the KASP list as well as creating the view/zones and configuring those. This is the next step of MR !10895 and !10901
While the code is extracted, some global variables has been changed into a function parameters which enable to have a clear view of the dependency of the function, typically, to know if it depends on local configuration object or runtime "production" object. The end goal (not in this MR, but later on) is to move as much as possible initialization logic outside of the exclusive mode.
As a first step, latest commits move the keystores list, KASP list and view/zones creation outside of the exclusive mode. (The view/zone configuration remain in exclusive mode for now, because of a dependency to the runtime "cachelist". This is the target of a next MR.
For the record; while moving the keystores list, KASP list and view/zone creation doesn't have a significant impact on the time the exclusive mode is taken (from my experiment on a 1M small zones instance); moving `configure_views` did have a _massive_ impact (basically, the time spend in the exclusive mode is then non calculable). Configuring views outside the exclusive mode needs more work, which will be done in future MRs.
See #4673
Merge branch 'colin/refactor-applyconfig' into 'main'
See merge request isc-projects/bind9!10910
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.
Replace the locked hashmap with the lock-free hashtable from the RCU
library and protect the fetch contexts against reuse by replacing the
libisc reference counting with urcu_ref that can soft-fail in situation
where the reference count is already zero. This allows us to easily
skip re-using the fetch context if it is already in process of being
destroyed.
Merge branch 'ondrej/use-urcu-lfht-for-resolver-tables' into 'main'
See merge request isc-projects/bind9!10653
Previously, the fetch contexts were stored inside rwlocked hashmap
table. This was one of the most contended places for the resolver,
especially in the cold cache situation.
Replace the locked hashmap with the lock-free hashtable from the RCU
library and protect the fetch contexts against reuse by replacing the
libisc reference counting with urcu_ref that can soft-fail in situation
where the reference count is already zero. This allows us to easily
skip re-using the fetch context if it is already in process of being
destroyed.
Previously, the slabtops for "type" and its signature was only loosely
coupled and the headers could expire at different time (both TTL and LRU
based expiry). Add a .related member to the slabtop that allows us to
expire the headers in both related headers and also optimize the lookups
because now both slabtops are looked up at the same time.
Closes#3396
Merge branch '3396-bind-rrsigs-to-records' into 'main'
See merge request isc-projects/bind9!10985
Previously, the slabtops for "type" and its signature was only loosely
coupled and the headers could expire at different time (both TTL and LRU
based expiry). This commit expires the headers in both related
headers.
Co-authored-by: Matthijs Mekking <matthijs@isc.org>
Previously, the slabtops for "type" and its signature was only loosely
coupled. Add a .related member to the slabtop that allows us to
optimize the lookups because now both slabtops are looked up at the
same time.
Co-authored-by: Matthijs Mekking <matthijs@isc.org>
There was a pattern where first the header was checked for NULL
and then for being stale. In both cases the code path is the same
so it makes sense to put them in a separate function.
Co-authored-by: Matthijs Mekking <matthijs@isc.org>
This is the first MR in series that aims to reduce the node locking
by replacing the single-linked list of slabtop(s) and slabheader(s)
with CDS linked list. This commit doesn't do anything else beyond
replacing .next and .down links with the cds_list_head. The RCU
semantics will be added later.
Merge branch 'ondrej/use-rcu-list-for-slabtop' into 'main'
See merge request isc-projects/bind9!10944
This is the second commit in series that aims to reduce the node locking
by replacing the single-linked list of slabheader(s) with CDS linked list.
This commit doesn't do anything else beyond replacing .next link with
the cds_list_head. RCU semantics is going to be added in the subsequent
commits.
This is the first commit in series that aims to reduce the node locking
by replacing the single-linked list of slabtop(s) with CDS linked list.
This commit doesn't do anything else beyond replacing .next link with
the cds_list_head. RCU semantics is going to be added in the subsequent
commits.
There was a data race where new fetch response could be added to the
fetch context after we unlock the fetch context and before we shut it
down. This could cause assertion failure when fctx__done() was called
with ISC_R_SUCCESS because there was originally no fetch response, but
new fetch response without associated dataset was added before we had a
chance to shutdown the fetch context. This manifested in the
validated() callback, where cache_rrset() now returns ISC_R_SUCCESS
instead of DNS_R_UNCHANGED when cache was not changed. However the data
race was wrong on a general level.
Add new argument to fctx__done() that allows to call it with fctx->lock
already acquired to prevent these data races.
Closes#5507
Merge branch '5507-dont-release-fctx-lock-on-done' into 'main'
See merge request isc-projects/bind9!10961
There was a data race where new fetch response could be added to the
fetch context after we unlock the fetch context and before we shut it
down. This could cause assertion failure when fctx__done() was called
with ISC_R_SUCCESS because there was originally no fetch response, but
new fetch response without associated dataset was added before we had a
chance to shutdown the fetch context. This manifested in the
validated() callback, where cache_rrset() now returns ISC_R_SUCCESS
instead of DNS_R_UNCHANGED when cache was not changed. However the data
race was wrong on a general level.
When the fctx__done() is called with ISC_R_SUCCESS as result is expects
the fctx->lock to be already acquired to prevent these data races.
The split will allow us to call fctx__done() with fctx->lock acquired
when it is called with ISC_R_SUCESS to prevent data races when finishing
the fetch context.
Trigger selected CI jobs on MR automatically only if there are related
code changes. Otherwise, offer an option to run the jobs manually in
MRs. For other sources, like schedules, tags etc., execute the jobs as
usual.
Merge branch 'nicki/ci-restrict-rules-changes' into 'main'
See merge request isc-projects/bind9!10987
Trigger selected CI jobs on MR automatically only if there are related
code changes. Otherwise, offer an option to run the jobs manually in
MRs. For other sources, like schedules, tags etc., execute the jobs as
usual.
Fixes an issue where failing to reconfigure/reload the server would prevent to preserved the views caches on the subsequent server reconfiguration/reload.
Closes#5523
Merge branch 'colin/fix-cache-revert' into 'main'
See merge request isc-projects/bind9!10984
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.
A named bug scrap the cache on a second reload after an initial reload
failure. Adds a test checking that the cache is preserved between server
reconfiguration/reloads even if it fails at some point (after attempting
to re-use the cache) and the server is re-loaded later.
The dns_qpcache already had all the namespace changes needed to put the
normal data and auxiliary NSEC data into a single tree. Remove the
extra nsec QP trie and use the single QP trie for all the cache data.
Merge branch 'ondrej/use-qp-namespace-in-cache' into 'main'
See merge request isc-projects/bind9!10975
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).
The dns_qpcache already had all the namespace changes needed to put the
normal data and auxiliary NSEC data into a single tree. Remove the
extra nsec QP trie and use the single QP trie for all the cache data.
The dbiterator_{last,prev} functions are not used in the cache, and the
implementation would get quite complicated when we squash the main and
nsec trees together. It's easier to just not implement these.
We would prefer if explicit $ORIGIN is used only for root zone and
nothing else, solely to avoid zone files named "..db". For all other
zones the file name should match zone name.
The dns_qpmulti_memusage() causes assertion failure when called on
freshly created qpmulti instance because the qp->usage hasn't been
allocated yet.
Merge branch 'ondrej/fix-qpmulti_memusage' into 'main'
See merge request isc-projects/bind9!10977
Slight simplification of the logic to define .nchildren listening UDP
socket.
Merge branch 'colin/simplify-socket-nchildren-count' into 'main'
See merge request isc-projects/bind9!10978
