Fix: tap array subscript has type char warning

On architectures where "char" is signed, it should be cast to unsigned
char before being passed as parameter to isdigit or isspace. Based on
their man page:

       These  functions  check  whether  c,  which  must  have the value of an
       unsigned char or EOF, falls into a certain character class according to
       the  specified  locale.

Passing a signed char as parameter is invalid if the values fall into
the negative range of the signed char.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Version 0.12.0

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

tls-compat.h: introduce DEFINE_URCU_TLS_INIT

DEFINE_URCU_TLS() cannot be used in this way for !CONFIG_RCU_TLS:

  DEFINE_URCU_TLS(ino_t, cached_net_ns) = NS_INO_UNINITIALIZED;

Introduce a new DEFINE_URCU_TLS_INIT which allows expressing a nonzero
intial value:

  DEFINE_URCU_TLS_INIT(ino_t, cached_net_ns, NS_INO_UNINITIALIZED);

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use _umtx_op for futex on FreeBSD

Tested with "make check", "make regtest" and a knotd slave.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add FreeBSD, DragonFly to syscall-compat.h

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

urcu-bp: perform thread registration on urcu_bp_register_thread

Some real-time use-cases (e.g. Xenomai) require to perform urcu-bp
thread registration early in the thread life-time before it starts
performing real-time tasks.

Currently, this can be achieved by issuing a urcu_bp_read_lock() and
urcu_bp_read_unlock() pair, or by using urcu_bp_read_ongoing(), while in
the initialization phrase.

However, it seems natural to expect that calling urcu_bp_register_thread()
would have the side effect to perform the lazy thread registration
immediately rather than being a no-op.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Require automake >= 1.12

The test suite LOG_DRIVER statement requires that automake >= 1.12 be used
during bootstrap.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

cds_lfht_is_node_deleted parameter can be marked const

Mark the cds_lfht_node pointer parameter of cds_lfht_is_node_deleted
as const. The fact that this parameter is mutable makes it harder to
use liburcu in const-correct code.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: provide errno as argument to urcu_die()

commit 1a990de3add "Fix: rculfhash worker needs to unblock to SIGRCU"
provides "ret" (-1) as argument to urcu_die(), but should rather provide
errno.

Reported by Coverity:

** CID 1405700:  Error handling issues  (NEGATIVE_RETURNS) /src/rculfhash.c: 2171 in cds_lfht_worker_init()

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: rculfhash worker needs to unblock to SIGRCU

In urcu-signal flavor, call_rcu_thread calls synchronize_rcu which
will send SIGRCU signal to all registed threads, and then loops to
wait need_mb to be cleared. However, the registed workqueue_thread
does not process the SIGRCU signal, and never clear the need_mb.
Based on above, call_rcu_thread and workqueue_thread will wait
forever for completion of the grace period: call_rcu_thread which holds
the rcu_registry_lock, waits for workqueue_thread to do cmm_smp_mb.
While workqueue thread never does cmm_smp_mb because of signal blocking,
and it will eventually wait to get rcu_registry_lock in do_resize_cb.

The phenomenon is as follows, which is easy to be triggered:

(gdb) t 2
[Switching to thread 2 (Thread 0xffff83c3b080 (LWP 27116))]
0  0x0000ffff845296c4 in poll () from /lib64/libc.so.6
(gdb) bt
0 0x0000ffff845296c4 in poll () from /lib64/libc.so.6
1 0x0000ffff8461b93c in force_mb_all_readers () at urcu.c:241
2 0x0000ffff8461c748 in smp_mb_master () at urcu.c:249
3 urcu_signal_synchronize_rcu () at urcu.c:445
4 0x0000ffff8461d004 in call_rcu_thread  at urcu-call-rcu-impl.h:364
5 0x0000ffff845eb8bc in start_thread () from /lib64/libpthread.so.0
6 0x0000ffff845335cc in thread_start () from /lib64/libc.so.6
(gdb) t 3
[Switching to thread 3 (Thread 0xffff8443c080 (LWP 27191))]
0 0x0000ffff845f51c4 in __lll_lock_wait () from /lib64/libpthread.so.0
(gdb) bt
0 0x0000ffff845f51c4 in __lll_lock_wait () from /lib64/libpthread.so.0
1 0x0000ffff845ee048 in pthread_mutex_lock () from /lib64/libpthread.so.0
2 0x0000ffff8461b814 in mutex_lock ( <rcu_registry_lock>) at urcu.c:157
3 0x0000ffff8461b9e4 in urcu_signal_unregister_thread () at urcu.c:564
4 0x0000ffff8463e62c in do_resize_cb (work=0x11e2e790) at rculfhash.c:2042
5 0x0000ffff8463c940 in workqueue_thread (arg=0x11e1d260) at workqueue.c:228
6 0x0000ffff845eb8bc in start_thread () from /lib64/libpthread.so.0
7 0x0000ffff845335cc in thread_start () from /lib64/libc.so.6

So we should not block SIGRCU in workqueue thread to avoid blocking
forever in the grace period awaiting on the worker thread when using
urcu-signal flavor.

Signed-off-by: hewenliang <hewenliang4@huawei.com>
Co-developed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

Cleanup: test_perthreadlock_timing: handle pthread mutex errors

Found by Coverity:
CID 1357055 (#1 of 1): Unchecked return value (CHECKED_RETURN)
4. check_return: Calling pthread_mutex_lock without checking return value
(as is done elsewhere 44 out of 48 times).

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: SONAME bump to 6.1.0

In commit d6c78161aed9b2d550ce201b0a8cd5b3ee515ac8 we bumped the 'age'
part of the library version with the intention of keeping the same major
SONAME because we only introduced new symbols. However by bumping the
'age' and not the 'current' we substracted 1 to the major SONAME which
we did not intend. Seems like we missed this in testing.

Fix it by bumping the 'current' to end up with an SONAME of 6.1.0 which
is what we originally intended.

From the libtool manual for reference :

  Programs using the previous version may use the new version as drop-in
  replacement, but programs using the new version may use APIs not present
  in the previous one. In other words, a program linking against the new
  version may fail with “unresolved symbols” if linking against the old
  version at runtime: set revision to 0, bump current and age.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: remove unused variable from configure.ac check

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: urcu/futex.h: users of struct timespec should include time.h

Fixes: #1187
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Version 0.11.0

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Bump library version to 6:0:1

Public symbol were added between 0.10 and 0.11.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: update code layout to fix old gcc warning

Some CI jobs show:

urcu-pointer.o
13:46:22 In file included from urcu.c:49:0:
13:46:22 urcu-wait.h:70:9: warning: missing initializer for field 'lock' of 'struct cds_wfs_stack' [-Wmissing-field-initializers]
13:46:22   struct urcu_wait_queue name = URCU_WAIT_QUEUE_HEAD_INIT(name)
13:46:22          ^
13:46:22 urcu.c:150:8: note: in expansion of macro 'DEFINE_URCU_WAIT_QUEUE'
13:46:22  static DEFINE_URCU_WAIT_QUEUE(gp_waiters);
13:46:22         ^
13:46:22 In file included from urcu-wait.h:27:0,
13:46:22                  from urcu.c:49:
13:46:22 ../include/urcu/wfstack.h:92:18: note: 'lock' declared here
13:46:22   pthread_mutex_t lock;
13:46:22

Change code layout so not to confuse old gcc.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: typo CPPLAGS in examples Makefile

Overriding CPPFLAGS throught the environment was ignored for the
examples.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Harmonize pprint macro across projects

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Check for TLS support after CC detection

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update macros from the autotools archive

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

tap-driver.sh: flush stdout after each test result

This is useful in a CI system where stdout is fully buffered and you
look at the console output to see which test is hanging.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update dead link in lgpl-relicensing.txt

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add multiflavor compat identifiers

This will maintain buildtime compatibility with softwares like lttng-ust
that used the prior undocumented multiflavor API.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: missing sign compare fixes

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: enable signed/unsigned compare compiler warning

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: compiler warning on 32-bit architectures

Replace a runtime test for 64-bit arch by a preprocessor test to
fix this compiler warning on 32-bit archs :

  comparison is always true due to limited range of data type [-Wtype-limits]

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

config.h.in: rename CONFIG_RCU_MULTIFLAVOR to CONFIG_RCU_HAVE_MULTIFLAVOR

Considering that this config option cannot be selected by
configure, and is just exposing the capabilities of liburcu,
use the "HAVE" semantic.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

rculfhash: implement iterator debugging config option

Building liburcu with --enable-cds-lfht-iter-debug and rebuilding
application to match the ABI change allows finding cases where the
hash table iterator is re-purposed to be used on a different hash
table while still being used to iterate on a hash table.

This is a common programming mistake that happens often enough
to justify creating a debugging mode to track this automatically.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: examples silent rules on Solaris

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add missing fall through annotations

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: symbol aliases with TLS compat

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Port: no symbols aliases on MacOS

There is no equivalent to symbols aliases on MacOS, this will
unfortunatly break the ABI for SONAME(6) and will require a rebuild of
client applications.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add -Wextra to CFLAGS

Edited by Mathieu Desnoyers:

Use /* fall through */ rather than __attribute__((fallthrough)) to
stay compatible with clang and gcc < 7. The fallthrough attribute
was introduced in gcc 7.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add silent mode to examples Makefiles

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

doc: update examples to API changes

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

test multiflavor single compile unit

Test multiple liburcu flavors within a compile unit.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update README following API changes

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use new header locations for includes from urcu code

This also moves urcu/static/urcu-pointer.h to urcu/static/pointer.h.
Considering that it is not meant to be included directly by library
users, it should not cause any problem.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update call-rcu.h and defer.h comments and include guards

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

rculfqueue.h: do not include urcu-call-rcu.h

urcu-call-rcu.h is included by rculfhqueue.h only for struct rcu_head
forward declaration, but as a result the urcu flavor needs to be chosen
beforehand, and therefore prevents using rculfqueue.h with multiple
urcu flavors in a given compile unit.

Remove that include and do a forward declaration of struct rcu_head
instead.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

rculfhash: support use with multiple flavors per compile unit

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

rculfhash: do not include urcu-call-rcu.h from public API

This include is not needed, and because of it rculfhash needs to be
included after the RCU flavor is chosen.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Refactor liburcu to support many flavors per compile unit

This refactoring keeps the prior use of liburcu "map" APIs unchanged.
However, it introduces the following new APIs:

Each urcu flavor is now available as its own header:

include/urcu/urcu-memb.h
include/urcu/urcu-mb.h
include/urcu/urcu-signal.h
include/urcu/urcu-bp.h
include/urcu/urcu-qsbr.h

The installed urcu headers that were not under the urcu/ subdirectory
are moved there:

include/urcu-call-rcu.h -> include/urcu/call-rcu.h
include/urcu-defer.h -> include/urcu/defer.h
include/urcu-flavor.h -> include/urcu/flavor.h
include/urcu-pointer.h -> include/urcu/pointer.h
include/urcu-bp.h -> include/urcu/urcu-bp.h
include/urcu.h -> include/urcu/urcu.h
include/urcu-qsbr.h -> include/urcu/urcu-qsbr.h

The liburcu "map" API is now only available for use when URCU_API_MAP is
defined before including the liburcu flavor headers.

The old headers are now placeholders defining URCU_API_MAP and including
the new headers for backward compatibility:

include/urcu-bp.h
include/urcu-call-rcu.h
include/urcu-defer.h
include/urcu-flavor.h
include/urcu-pointer.h
include/urcu-qsbr.h
include/urcu.h

The header include/urcu/urcu.h now includes the right header between the
memb, signal, or mb flavors based on the compiler defines.

The symbol names of liburcu flavors are cleaned up, favoring the
following hierarchy:

urcu_<flavor name>_...

This is an ABI-breaking change, however the previous symbols name were
kept as aliases to maintain backward compatibility. They will be removed
when the next SONAME bump occurs.

The new liburcu-memb.so shared object is introduced, properly
namespacing this flavor. It is a duplicate of the previous liburcu.so,
which is kept around for backward compatibility.

The new URCU_API_MAP macro is introduced, controlling whether the
urcu API "mapping" should stay defined after inclusion of the flavor
headers. Users wishing to use the prior urcu API should either
explicitly define URCU_API_MAP before including the urcu/urcu*.h flavor
headers, or include the flavor header files from the include toplevel
directory, which are placeholders for backward compatibility. Use of
many urcu flavors within the same _LGPL_SOURCE compile unit should not
use the "map" APIs.

Internally, the "map" header files are split into one header per
flavor. The include guards are removed, so their effect can be
applied more than once. A new include/urcu/map/clear.h header is
introduced, which undefines the mappings at the end of the flavor
header if URCU_API_MAP is not set.

The new APIs namespaced for each urcu flavor is the recommended way to
use liburcu. We can expect the prior APIs to eventually become
deprecated over time.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: only wait if work queue is empty in real-time mode

Unconditionally waiting for 10 ms after the completion of every batch
of jobs of the work queue in real-time mode appears to be a behaviour
inherited from the call-rcu thread.

While this is a fair trade-off in the context of call-rcu, it is less
evident that it is desirable in the context of a general-purpose
work queue. Waiting when work is available artificially degrades the
latency characteristics of the work queue.

If a workqueue user even need the explicit delay for batching (e.g. if
a call-rcu implementation would ever use the workqueue worker thread),
it can add it within the worker_before_wait_fct callback received as
argument from workqueue creation.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: don't wait after completion of a work queue job batch

As indicated in the previous patch of this series, waiting on
completion of a job batch from the work queue artificially increases
the latency of the work queue.

The previous patch removed the wait that is performed when the
work queue is observed to be empty and was observed as the cause of a
performance problem.

It is likely that waiting when the queue is observed to be non-empty
is similarly unintended. Note that I have not observed such a problem
myself.

If a workqueue user even need the explicit delay for batching (e.g. if
a call-rcu implementation would ever use the workqueue worker thread),
it can add it within the worker_before_wait_fct callback received as
argument from workqueue creation.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: don't wait after completion of job batch if work queue is empty

On completion of a batch of jobs from the work queue, a wait of 10
ms (using poll()) is performed if there is no work present in the
work queue before waiting on its futex.

The work queue thread's structure is inspired by the call-rcu thread.
In the context of the call-rcu thread, my understanding is that the
intention is to ensure that the thread does not continuously wake-up
to process a single queued item. This is fine as an application should
not wait for a call-rcu job to be executed (or at least I don't see a
use-case for that).

In the context of the work queue, waiting for more work to be available
artificially slows down the execution of work on which an application
may wait.

I have observed a case where LTTng's session daemon's shutdown is
takes around 4 seconds as a large number of cds_lfht objects are
destroyed. Removing the wait reduces the duration of this phase of the
shut-down to almost ~10ms.

If a workqueue user even need the explicit delay for batching (e.g. if
a call-rcu implementation would ever use the workqueue worker thread),
it can add it within the worker_before_wait_fct callback received as
argument from workqueue creation.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: workqueue: struct urcu_work vs rcu_head mixup

The workqueue code was derived from call-rcu, and its API
expects a struct urcu_work for work items, but it internally iterates
over struct rcu_head.

This is not an issue at runtime because both structures have the
exact same layout and content, but it is a type mixup nevertheless.

Use the right type in the implementation.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: workqueue: update comments referring to call-rcu

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: mixup between URCU_WORKQUEUE_RT and URCU_CALL_RCU_RT

The work queue implementation is derived from the call-rcu thread. A
number of references seem to have been left in place when adapting the
code for its new purpose.

The URCU_CALL_RCU_RT flag is used by wake_worker_thread() while the
rest of the workqueue.c code uses URCU_WORKQUEUE_RT to determine if
the work queue was configured in real-time mode. Both flags are defined
to the same value (0x1) and the current internal user of the
work queue (lfht) never specifies any flags.

In practice, this does not cause any problem, but this mixup should
be fixed nevertheless.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

test_rwlock: Add per-thread count to verbose output

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add *.exe to gitignore for Cygwin

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: pthread_rwlock initialization on Cygwin

On Cygwin the PTHREAD_RWLOCK_INITIALIZER macro is not
sufficient to get a properly initialized pthread_rwlock_t
struct. Use the pthread_rwlock_init function instead which
should work on all platforms.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: compat_futex_noasync on Cygwin

The futex_noasync compat code uses a weak symbol to share state across
different shared object which is not possible on Windows with the
Portable Executable format. Use the async compat code for both cases.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

wfcqueue: allow defining CDS_WFCQ_WAIT_SLEEP to override `poll'

Users may want to use alternative sleeping behavior instead of
`poll'.  Make CDS_WFCQ_WAIT_SLEEP a macro which may be defined
before including wfcqueue.h.

This alternative behavior could include logging, performing
low-priority cleanup work, sleeping a shorter/longer interval
or any combination of that.

This will also make integration into glibc easier, as `poll'
linkage causes conformance test failures even when relegated
to an impossible code path:
https://public-inbox.org/libc-alpha/20180801092626.jrwyrojfye4avcis@whir/

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update documentation for call_rcu before/after fork

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add support for the RISC-V architecture

Tested in QEMU 2.12.0-rc0, requires --disable-compiler-tls to go
through the benchmarks reliably.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Tests: Add tap-driver.sh for automake < 1.12

Add the tap-driver.sh script in the aux directory to enable
systems with an automake version prior to 1.12 to bootstrap
the source tree from git.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Tests: Replace prove by autotools tap runner

This patch removes the dependency on the prove perl script
to run the TAP test suite. It replaces it with the autotools
shell TAP driver that only requires a shell and awk.

Custom arguments can be passed to the test runner with
env variables as follow:

  env LOG_DRIVER_FLAGS='--comments --ignore-exit' \
      TESTS='foo.test baz.test' make -e check

This tap driver also creates a log file for each test that
can then be used by another system to build a test report.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liburcu-bp: Use membarrier private expedited when available

For the liburcu-bp flavor, use the membarrier private expedited command
when available. It is faster than the shared expedited command, but has
only been introduced in 4.14 Linux kernels.

When configured with --disable-sys-membarrier-fallback, liburcu-bp
will abort if running on a kernel that do not provide the membarrier
private expedited command (e.g. CONFIG_MEMBARRIER=n or kernel version
below 4.14).

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

liburcu: Use membarrier private expedited when available

For the liburcu flavor, use the membarrier private expedited
command when available. It is faster than the shared expedited
command, but has only been introduced in 4.14 Linux kernels.

When configured with --disable-sys-membarrier-fallback, liburcu
will abort if running on a kernel that provide neither the shared
nor the private expedited membarrier commands. This is the case
if running on a CONFIG_MEMBARRIER=n kernel, or a kernel version
below 4.3.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

rculfhash: improve error handling of mmap backend

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: don't use overlapping mmap mappings on Cygwin

The allocation scheme used by the mmap based RCU hash table is to make a
large unaccessible mapping to reserve memory without allocating it.
Then smaller chunks are allocated by overlapping read/write mappings which
do allocate memory. Deallocation is done by an overlapping unaccessible
mapping.

This scheme was tested on Linux, macOS and Solaris. However, on Cygwin the
mmap wrapper is based on the Windows NtMapViewOfSection API which doesn't
support overlapping mappings.

An alternative to the overlapping mappings is to use mprotect to change the
protection on chunks of the large mapping, read/write to allocate and none
to deallocate. This works perfecty on Cygwin and Solaris but on Linux a
call to madvise is also required to deallocate and it just doesn't work on
macOS.

For this reason, we keep to original scheme on all platforms except Cygwin.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Tests fix: errors in shell scripts

Fix all shellcheck errors in the test scripts, switch to posix
compatible syntax. Remove duplicated code already included in common.sh.
Call the tap.sh cleanup code from our exit trap instead of overriding
it.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Revert "Use initial-exec tls model"

This reverts commit 6fd172f599e8d798e68974a786dd930d876f182e.

The initial-exec model seems to behave differently than global-dynamic
with respect to lazy initialization, causing locks to be taken then
first time each thread touch the TLS. This introduces deadlocks
with library constructors waiting on other threads.

This will require further investigation.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use initial-exec tls model

The initial-exec tls model removes requirement on performing memory
allocation the first time a tls variable is touched by any given thread.
This is needed to ensure usage of the TLS from a signal handler works
fine.

Given that the link-editor figures out the right model to use at
runtime, we can change the tls model without changing the soname major
version.

This also brings interesting speedups over the GD model. This does not
affects TLS accesses performed by executables, but does affect TLS
accesses performed by libraries.

* Executable (no change)

./test_urcu 1 0 10
SUMMARY /media/truecrypt1/compudj/doc/userspace-rcu/tests/benchmark/.libs/test_urcu testdur   10 nr_readers   1 rdur      0 wdur      0 nr_writers   0 wdelay      0 nr_reads   4420328692 nr_writes            0 nr_ops   4420328692

(with initial-exec)
./test_urcu 1 0 10
SUMMARY /media/truecrypt1/compudj/doc/userspace-rcu/tests/benchmark/.libs/test_urcu testdur   10 nr_readers   1 rdur      0 wdur      0 nr_writers   0 wdelay      0 nr_reads   4424925864 nr_writes            0 nr_ops   4424925864

* Library

(with global-dynamic)
./test_urcu_dynamic_link 1 0 10
SUMMARY /media/truecrypt1/compudj/doc/userspace-rcu/tests/benchmark/.libs/test_urcu_dynamic_link testdur   10 nr_readers   1 rdur      0 wdur      0 nr_writers   0 wdelay      0 nr_reads    573209491 nr_writes            0 nr_ops    573209491

(with initial-exec)
./test_urcu_dynamic_link 1 0 10
SUMMARY /media/truecrypt1/compudj/doc/userspace-rcu/tests/benchmark/.libs/test_urcu_dynamic_link testdur   10 nr_readers   1 rdur      0 wdur      0 nr_writers   0 wdelay      0 nr_reads   1088836185 nr_writes            0 nr_ops   1088836185

Link: https://docs.oracle.com/cd/E23824_01/html/819-0690/chapter8-20.html
Link: https://gcc.gnu.org/onlinedocs/gcc/Common-Variable-Attributes.html#Common-Variable-Attributes
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: don't use membarrier SHARED syscall command in liburcu-bp

One main user of liburcu-bp (lttng-ust) invokes synchronize_rcu()
repeatedly, without batching (does not use call_rcu).

Those delays introduced by sys_membarrier SHARED command significantly
impacts application startup time. Therefore, revert to not using the
membarrier SHARED command.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Tests fix: add missing Cygwin thread id

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: assignment from incompatible pointer type warnings

On some platforms, mmap returns a caddr_t pointer which generates
compiler warnings, cast to the proper pointer type to eliminate them.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Tests fix: unused variable warnings

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: add missing m68k headers to dist

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Bump version to 0.11-pre

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Version 0.10.0

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Bump library soname due to urcu flavor structure change

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: use mutex_lock() wrapper in rculfhash

Fixes this coverity issue:

*** CID 1375952:  Error handling issues  (CHECKED_RETURN)
/src/rculfhash.c: 1950 in cds_lfht_resize()
1944     }
1945
1946     void cds_lfht_resize(struct cds_lfht *ht, unsigned long new_size)
1947     {
1948             resize_target_update_count(ht, new_size);
1949             CMM_STORE_SHARED(ht->resize_initiated, 1);
>>>     CID 1375952:  Error handling issues  (CHECKED_RETURN)
>>>     Calling "pthread_mutex_lock" without checking return value (as is done elsewhere 44 out of 50 times).
1950             pthread_mutex_lock(&ht->resize_mutex);
1951             _do_cds_lfht_resize(ht);
1952             pthread_mutex_unlock(&ht->resize_mutex);
1953     }
1954
1955     static

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use workqueue in rculfhash

The RCU lock-free hash table currently requires that the destroy
function should not be called from within RCU read-side critical
sections. This is caused by the lazy resize, which uses the call_rcu
worker thread, even though all it really needs is a workqueue/worker
thread scheme.

Use the new internal workqueue API instead of call_rcu in rculfhash to
overcome this limitation.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Implement urcu workqueues internal API

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add support for m68k architecture

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Tested-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Set -Wall globally in AM_CFLAGS

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: remove double use of PTHREAD_CFLAGS

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Re-add PTHREAD_CFLAGS to global CFLAGS

Fixes a build failure on Solaris and maybe other esoteric architectures.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: Don't override user variables within the build system

Instead use the appropriatly prefixed AM_* variables as to not interfere
when a user variable is passed to a make command. The proper use of flag
variables is documented at :

https://www.gnu.org/software/automake/manual/automake.html#Flag-Variables-Ordering

Fixes #1095

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add report at the end of configure

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

uatomic-api docs: use the third-person singular

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add --enable-rcu-debug to configure

When used CONFIG_RCU_DEBUG is defined in urcu/config.h, thus the
debugging self-test are used at all time. This enables a permanent
built-in debugging behaviour.

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

ARM32: use dmb ish (inner shareable domain) for smp barriers

Based on https://gcc.gnu.org/ml/gcc-patches/2014-11/msg01272.html:

From: Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>:

"Hans Boehm pointed out that we were using dmb sy instead of dmb ish.
Given that the ARM-ARM says that the inner shareability domain is really
the one that contains all PE's controlled by a single hypervisor or
operating system, it would be safe to replace all dmb sy's with dmb
ish's. "

Keep full system barriers for cmm_mb()/cmm_rmb()/cmm_wmb().

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: remove cmm_wmb() from rcu_xchg_pointer and rcu_cmpxchg_pointer

Both rcu_xchg_pointer() and rcu_cmpxchg_pointer() imply both release and
acquire barriers. Therefore, the extra cmm_wmb() is redundant and can be
removed.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: uatomic arm32: add missing release barrier before uatomic_xchg

__sync_lock_test_and_set() only imply a release barrier, but
uatomic_xchg() guarantees both acquire and release barrier semantics.
Therefore, add the missing release barrier.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Tests: Add verbose support to test script

Add support for the standard "V=1" to make the test runner script
verbose.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: add missing CONFIG_RCU_FORCE_SYS_MEMBARRIER to urcu/config.h.in

Expose this configuration define in the installed header.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Allow forcing the use of sys membarrier

When using the default (liburcu.so) and bulletproof (liburcu-bp.so)
flavours of Userspace RCU, kernel support for sys-membarrier is detected
dynamically and stored in the rcu_has_sys_membarrier_memb and
urcu_bp_has_sys_membarrier global variables.

Checking the value of these variables adds a small but measurable overhead
to smp_mb_slave. On systems which support sys-membarrier, it would be
nice to have a way of avoiding that overhead.

Here is the proposed approach: if CONFIG_RCU_FORCE_SYS_MEMBARRIER is
defined then rcu_has_sys_membarrier_memb/urcu_bp_has_sys_membarrier are
replaced with the constant 1, eliminating the overhead in smp_mb_slave.
As a sanity check, support for sys-membarrier is still detected at
startup. A program using liburcu or liburcu-bp compiled with this option
aborts in the library constructor if the membarrier system call is not
supported by the operating system.

Suggested-by: Duncan Sands <duncan.sands@deepbluecap.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: rcutorture: work-around signal issue on mac os x

Our MacOS X test machine with the following config:

15.6.0 Darwin Kernel Version 15.6.0
root:xnu-3248.60.10~1/RELEASE_X86_64

appears to have issues with liburcu-signal signal being delivered on top
of pthread_cond_wait. It seems to make the thread continue, and
therefore corrupt the rcu_head. Work around this issue by unregistering
the RCU read-side thread immediately after call_rcu (call_rcu needs us
to be registered RCU readers).

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: rcutorture should register thread using call_rcu

From rcu-api.txt:

`call_rcu` should be called from registered RCU read-side threads.
For the QSBR flavor, the caller should be online.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: add missing backslash in Makefile.am

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: Do not use wildcards in include/Makefile.am

Wildcards are not officially supported by autotools
in Makefiles since it needs to know the exact list
of files it has to work with.

Using an absolute path was a hack that worked as long
as the path to the header files from the top source dir
was the same as the install path of those files, which
is not the case anymore.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Bump version to 0.10-pre

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: check for rand_r() in compat-rand.h

The rand_r() function exists in newer Android versions,
detect it instead of relying on the __ANDROID__ define.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: Move rand-compat to private src dir

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix: remove AC_FUNC_MALLOC from configure.ac

AC_FUNC_MALLOC fails cross-compile builds and is
only used to detect non-standard glibc behavior where
malloc(0) does not return a null pointer.

We don't depend on that behavior since we would have
to ship a compat implementation of malloc() for this
macro to be of any use.

Keep it commented because autoscan will report it
as missing and it might get re-introduced.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: Re-organise source dir

Re-organise the sources, add a top level "src" and "include" dir and
move relevant files.

Disable autotools automated includes and define them manually. This
fixes problems with collision of header names with system headers.

Include the autoconf config.h in the default includes and remove it
where it's explicitely included. Remove _GNU_SOURCE defines since
it's detected at configure for platforms that requires it and added
to the config.h.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>