ComputerSurge/bind9
3
0
Fork 0
Code Pull Requests 1 Activity
30,267 Commits 589 Branches 805 Tags
3f736b43740ceec2016ec72049a89b6673949a78
Commit Graph
30267 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Evan HuntandWitold Kręcicki 3f736b4374 fixup! - Get rid of obsolete fields in ns_client structure - WiP - local address detection - WiP - XFR cancel 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki de05f22fd5 Add retry loop for tests immediately after server restart 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki da5da14f36 fixup! - Get rid of obsolete fields in ns_client structure - WiP - local address detection - WiP - XFR cancel 2019-11-05 12:54:45 +01:00
Witold Kręcicki d44a0385f8 fixup! - Get rid of obsolete fields in ns_client structure - WiP - local address detection - WiP - XFR cancel 2019-11-05 12:54:45 +01:00
Witold Kręcicki 28f35351ad - Get rid of obsolete fields in ns_client structure 
- WiP - local address detection
- WiP - XFR cancel
 2019-11-05 12:54:45 +01:00
Witold Kręcicki 39104f766d Initialize atomics properly in isc_nm_start 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki f78d234ce3 remove isc_nm_paused(), it's no longer needed 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki fb160b071c fixup! Fix a race between isc_nm_start and nm_thread 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 469e7e5a3e fixup! fixup! Define a new 'interlocked' status of netmgr - if it's active that means there's an operation that depends on many workers. No two such operation can be launched at once without risking a deadlock, so other ones must either wait or requeue. 2019-11-05 12:54:45 +01:00
Witold Kręcicki 28a0e1e36f Add tactical post-reload delay to some system tests 2019-11-05 12:54:45 +01:00
Witold Kręcicki a8fda869a2 fixup! Define a new 'interlocked' status of netmgr - if it's active that means there's an operation that depends on many workers. No two such operation can be launched at once without risking a deadlock, so other ones must either wait or requeue. 2019-11-05 12:54:45 +01:00
Witold Kręcicki d90aa3a0e1 Disable leak detection in ASAN - libuv has some problems with that 2019-11-05 12:54:45 +01:00
Witold Kręcicki b29b2bfa2e Define a new 'interlocked' status of netmgr - if it's active 
that means there's an operation that depends on many workers.
No two such operation can be launched at once without risking
a deadlock, so other ones must either wait or requeue.
 2019-11-05 12:54:45 +01:00
Witold Kręcicki fdeb8c4f48 fixup! netmgr: don't try to send to a closed tcpdns socket 2019-11-05 12:54:45 +01:00
Witold Kręcicki e3ab21352a Fix a race between isc_nm_start and nm_thread 2019-11-05 12:54:45 +01:00
Witold Kręcicki 27593715b3 netmgr: empty the async event queue right before destroying it 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki c32c4c4b3b silence a clang warning 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 5febb7e002 fixup! - Always close listening sockets asynchronously, even if we're in the specific thread - Export some lib/ns/client stuff for unit tests - Fix lib/ns unit tests 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 97da9703cd add missing atomic_exchange macro in mutexatomic.h 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 28d4fa2003 fixup! - Always close listening sockets asynchronously, even if we're in the specific thread - Export some lib/ns/client stuff for unit tests - Fix lib/ns unit tests 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 1ac082941c make library wrapping work correctly in the libns unit tests 
when LD_WRAP is useable, use it; otherwise, use a "libwrap"
shared library that's linked with '-z interpose' and replaces
isc_nmhandle_unref().
 2019-11-05 12:54:45 +01:00
Witold Kręcicki 6dc0edad86 - Free all the pending async calls when the worker is exiting 
- Add a hack to fix libuv not sending uv_close callback if nothing
  was in the workqueue
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 395b9b50a3 re-enqueue stop listening events if network manager is pausing 
it was still possible to deadlock between stop-listening and pause
events. this is now addressed by testing to see whether the manager is
in the process of pausing, and re-enqueueing stop-listening events
for later. the uv_run() loop is now terminated immediately after
uv_stop() is called, so that the re-enqueued stop-listeneing event
won't be processed until after the manager is unpaused.
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 2136e7568e fixup! - Always close listening sockets asynchronously, even if we're in the specific thread - Export some lib/ns/client stuff for unit tests - Fix lib/ns unit tests 2019-11-05 12:54:45 +01:00
Witold Kręcicki 4e1bdb2e63 - Always close listening sockets asynchronously, even if we're in the specific thread 
- Export some lib/ns/client stuff for unit tests
- Fix lib/ns unit tests
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 26b9212bff unlink from recursing-clients and reset state when resuming from recursion 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 4075fb013c release recursion quota in fetch_callback() 
this was previously only done in ns_client_endrequest(), which is
called by client_reset_cb() when a client is released. this could
make the system report more recursive clients than were actually
recursing, and caused intermittent test failures in 'fetchlimit'.
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki b8e54a6ca4 don't assert in isc_task_pause() if the task is 'running' 
there is a window of time after a task finishes running an event
before the task state is reset from 'running' to 'ready' or 'idle'.
previously, if the network manager ran ns__client_request() during
that time, isc_task_pause() would assert.
 2019-11-05 12:54:45 +01:00
Witold Kręcicki d395a23335 fixup! Add atomic_exchange operations to stdatomics.h 2019-11-05 12:54:45 +01:00
Witold Kręcicki 26b19951ab fixup! Add atomic_exchange operations to stdatomics.h 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 026efd2cc8 fully reset the client object before reusing it 
not doing so caused unpredictable failures, for example, when
client->rcode_override was set by a previous query and then the
client was reused.
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 549c3a0111 use family AF_INET6 when creating UDP sockets for IPv6 addresses 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 31fe4bbe7c lock manager when stopping workers, to prevent a possible shutdown deadlock 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki a0410678de Don't assert if there's left-over data after reading a DNS TCP message. 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 6900793dba set the base pointer correctly when gluing TCP reads into a single message 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 638a289ca9 Implement "maxudp" for netmgr so that the legacy test will pass. 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 1102ec73a7 remove obsolete references to '-T clienttest' 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 67211aa288 acquire manager lock before sending 'stoplisten' events to workers 
this prevents a potential race between isc_nm_udp_stoplistening()
and isc_nm_pause() which could cause a server to deadlock.
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 7bc1004f00 always reference the client->handle before ns_query_recurse() 
this supercedes the experiment in commit eab6c2bff
("attach to nmhandle during zerottl refetch"), and ensures that
the client can't be reset before fetch_callback() is reached.
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 44e7fec8a7 make "stoplisten" events asynchronous so we can issue them while paused 
when reconfiguring a server we're in task exclusive mode, which
means the netmgr is paused. previously, when we issued "stoplistening"
events, we would wait for the worker thread to process the event
before proceeding. since that can't happen while paused, we could
hit a deadlock during reconfig if the listen-on addresses changed.
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki a5ca545f39 more fixes for handle reference counting 
- ref handle in tcp_send_direct and tcp_send_cb
- unref handle when xfrout fails
- remove some unneeded unrefs in ns__client_request
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 0e2154f980 style nits 2019-11-05 12:54:45 +01:00
Witold Kręcicki 7a677338ec Fix compilation issues in mutexatomics mode 2019-11-05 12:54:45 +01:00
Witold Kręcicki 334d9aa90b Use pkg-config to detect libuv 2019-11-05 12:54:45 +01:00
Witold Kręcicki 3584565c4b netmgr: fixes for Windows: 
- MSVC doesn't like empty {} initializers
 - Emulate stdalign.h on Windows with __declspec(align())
 - There's no usleep on Windows
 - Add missing symbols to libns.def and libisc.def
 - Proper use of SO_REUSE****
 - libuv detection in Configure
 - Minor fixes to the atomics code
 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 1b4bbebd11 eliminate isc_nm_shutdown in favor of isc_nm_detach 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki e457feb1e3 continue client/handle reference counting changes 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 021848bca9 experiment: eliminate client reference counting; use nmhandler only 2019-11-05 12:54:45 +01:00
Evan HuntandWitold Kręcicki 669f659fe3 partial fix: improve socket and metmgr cleanup for orderly shudown 
- when we attempt to shut down a parent socket, but its children
  still have active handles, the destruction is delayed, and previously
  it was never resumed. now, when the last handle for a child socket is
  detached, we check again whether the parent can be destroyed.
- the netmgr is not shut down until all references to it are detached.
 2019-11-05 12:54:45 +01:00
Witold Kręcicki e3b597ffa7 Add atomic_exchange operations to stdatomics.h 2019-11-05 12:54:45 +01:00