Hi,currently, the BIND 9 design involves multiple threads with per-thread event loop from libuv (uv_loop).We are using the default rcu variant, but it seems like an ideal place to switch to QSBR as we can call quiescent state periodically before polling on the loop.This comes with one trouble thoug
> On 10. 7. 2025, at 16:05, Mathieu Desnoyers
> wrote:
>
> On 2025-07-09 14:57, Ondřej Surý via lttng-dev wrote:
>> Hi,
>> as the answer to this might be useful to more people, I am asking here:
>> The cds_lfht_new documentation specifies 3 sizes.
>> *
Hi,
as the answer to this might be useful to more people, I am asking here:
The cds_lfht_new documentation specifies 3 sizes.
* cds_lfht_new - allocate a hash table.
* @init_size: number of buckets to allocate initially. Must be power of two.
* @min_nr_alloc_buckets: the minimum number of all
Hey Mathieu,
I’m actually thinking about using librseq for some more lightweight stuff first
- like statistic counters (especially those related to memory allocations), and
my first obvious question would be - how portable is the librseq memory
allocator?
I am not keen on having a code with if
Hi,
looks like my git-send-email configuration is not correct and my mailserver ate
the patch,
so here's one created by git-format-patch...
Nothing important, but it bite me today...
0001-pointer.h-Fix-the-rcu_cmpxchg_pointer-documentation.patch
Description: Binary data
Ondrej
--
Ondřej Sur
Hi,
we are just exploring the field, so asking here is probably good idea.
We are currently using bucketed heaps for TTL-based expiration of
DNS records in BIND 9. This doesn't scale that well and has other
problems, so I was thinking that we might be able to replace this with
something better.
Hi,
I've updated to the latest HEAD and compilation with clang-16 and clang-17 and
./configure --enable-compiler-atomic-builtins --disable-legacy-mb
It fails with:
rculfhash.c:1189:2: error: address argument to atomic operation must be a
pointer to integer ('typeof (node_next)' (aka 'struct cd
Hi,
we are in process of adding USDT probes to the BIND 9 probes and the obvious
question is whether it’s possible to use the probes with the LTTng tools.
Alternatively, if there’s an easy way how to use one or another. The USDT
probes are provided by SystemTap on Linux and by DTrace on BSDs an
(Sorry, I missed closing brackets in both macros, so resending fixed patch...)
The cds_lfht_for_each_entry and cds_lfht_for_each_entry_duplicate macros
would call caa_container_of() macro on NULL pointer. This is not a
problem under normal circumstances as the check in the for loop fails
and the
The cds_lfht_for_each_entry and cds_lfht_for_each_entry_duplicate macros
would call caa_container_of() macro on NULL pointer. This is not a
problem under normal circumstances as the check in the for loop fails
and the loop-statement is not called with invalid (pos) value.
However AddressSanitizer
Olivier,
is this somewhere in Gerrit again, please? It’s much easier for us to pull the
changes from git than cherry-pick patches from the mailing list.
Ondřej
--
Ondřej Surý (He/Him)
> On 7. 6. 2023, at 20:54, Olivier Dion wrote:
>
> This patch set adds support for TSAN in liburcu.
>
> *
> On 26. 5. 2023, at 8:08, Dmitry Vyukov wrote:
>
> On Fri, 26 May 2023 at 07:33, Ondřej Surý wrote:
>>
>> A little bit related question/problem - some of our system test jobs got
>> significantly slower (3x-5x) with **clang-16** TSAN after we integrated
>> liburcu to BIND 9. The GCC (13.1.1
A little bit related question/problem - some of our system test jobs got
significantly slower (3x-5x) with **clang-16** TSAN after we integrated liburcu
to BIND 9. The GCC (13.1.1) TSAN doesn’t exhibit this behavior.
The liburcu integration in BIND 9 is very light at the moment - we use liburcu
Hi,
we (ISC) have no intention of using signal based URCU in BIND 9. Our plan
relies on using the QSBR flavor as BIND 9 is internally event-loop based with
natural quiescent states.
Ondrej
--
Ondřej Surý (He/Him)
> On 10. 5. 2023, at 23:10, Olivier Dion via lttng-dev
> wrote:
>
> Hi all,
Hi,
the documentation is pretty silent on this, and asking here is probably going
to be faster
than me trying to use the source to figure this out.
Is it legal to call_rcu() from within the call_rcu() callback?
What about the other RCU (and CDS) API calls?
How does that interact with create_ca
> On 22. 3. 2023, at 9:02, Ondřej Surý via lttng-dev
> wrote:
>
> That's pretty much weird because the "Write" happens on stack local variable,
> while the "Previous write" happens after futex, which lead me to the fact that
> ThreadSanitizer doesn
Hi,
this happens with all the patches fully applied and doesn't seem to be caused
by anything I am doing :)
WARNING: ThreadSanitizer: data race (pid=3995296)
Write of size 8 at 0x7fb51e5fd048 by thread T296:
#0 __tsan_memset (badcache_test+0x49257d) (BuildId:
166dea93b2dca28264fc85c79b56
> On 21. 3. 2023, at 15:49, Mathieu Desnoyers
> wrote:
>
> Agreed, let's see how it holds up to testing under TSAN. :)
The make check is OK, but any of the make regtest, short_bench or long_bench
produces
lot of TSAN warnings like these below.
Looks like they mostly come from the tests itself
When adding REMOVED_FLAG to the pointers in the rculfhash
implementation, retype the generic pointer to unsigned long to fix the
following compiler error:
rculfhash.c:1201:2: error: address argument to atomic operation must be a
pointer to integer ('struct cds_lfht_node **' invalid)
uatomi
> On 21. 3. 2023, at 15:46, Mathieu Desnoyers
> wrote:
>
> On 2023-03-21 06:21, Ondřej Surý wrote:
>>> On 20. 3. 2023, at 19:37, Mathieu Desnoyers
>>> wrote:
>>>
>>> On 2023-03-17 17:37, Ondřej Surý via lttng-dev wrote:
>>>> FIXME:
Add autoconf checks for all __atomic builtins that urcu require, and
adjust the gcc and clang versions in the README.md.
Signed-off-by: Ondřej Surý
---
README.md| 33 +
configure.ac | 15 +++
2 files changed, 24 insertions(+), 24 deletions(-)
diff
Instead of using CMM_ACCESS_ONCE() with memory barriers, use __atomic
builtins with relaxed memory ordering to implement CMM_LOAD_SHARED() and
CMM_STORE_SHARED().
Signed-off-by: Ondřej Surý
---
include/urcu/system.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/include/
Instead of a custom code, use the __atomic_thread_fence() builtin to
implement the cmm_mb(), cmm_rmb(), cmm_wmb(), cmm_smp_mb(),
cmm_smp_rmb(), and cmm_smp_wmb() on all architectures, and
cmm_read_barrier_depends() on alpha (otherwise it's still no-op).
family of functions
Signed-off-by: Ondřej S
When adding REMOVED_FLAG to the pointers in the rculfhash
implementation, retype the generic pointer to uintptr_t to fix the
compiler error.
Signed-off-by: Ondřej Surý
---
src/rculfhash.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/rculfhash.c b/src/rculfhash.c
in
Use __atomic_signal_fence(__ATOMIC_SEQ_CST) for cmm_barrier(), so
ThreadSanitizer can understand the memory synchronization.
Signed-off-by: Ondřej Surý
---
include/urcu/compiler.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/include/urcu/compiler.h b/include/urcu/compiler.
Replace the custom assembly code in include/urcu/uatomic/ with __atomic
builtins provided by C11-compatible compiler.
Signed-off-by: Ondřej Surý
---
include/Makefile.am| 16 -
include/urcu/uatomic.h | 84 +++--
include/urcu/uatomic/aarch64.h | 41 ---
include/urcu/uatomic/
This is a second round of the patches after implementing the requested changes
from the first round.
Ondrej
[PATCH 1/7] Require __atomic builtins to build
- no changes
[PATCH 2/7] Use gcc __atomic builtis for
- the non return macros are now __ATOMIC_RELAXED
- the return macros are now __ATOMIC_
Instead of custom code, use the __atomic builtins to implement the
rcu_dereference(), rcu_cmpxchg_pointer(), rcu_xchg_pointer() and
rcu_assign_pointer().
The rcu_dereference() now relies on CONSUME memory order.
Signed-off-by: Ondřej Surý
---
include/urcu/static/pointer.h | 77 +++--
> On 20. 3. 2023, at 19:37, Mathieu Desnoyers
> wrote:
>
> On 2023-03-17 17:37, Ondřej Surý via lttng-dev wrote:
>> FIXME: This is experiment that adds explicit memory barrier in the
>> free_completion in the workqueue.c, so ThreadSanitizer knows it's ok to
>>
> On 20. 3. 2023, at 19:31, Mathieu Desnoyers
> wrote:
>
> On 2023-03-17 17:37, Ondřej Surý via lttng-dev wrote:
>> When adding REMOVED_FLAG to the pointers in the rculfhash
>> implementation, retype the generic pointer to uintptr_t to fix the
>> compiler error.
&
> On 20. 3. 2023, at 19:03, Mathieu Desnoyers
> wrote:
>
> In doc/uatomic-api.md, we document:
>
> "```c
> type uatomic_cmpxchg(type *addr, type old, type new);
> ```
>
> An atomic read-modify-write operation that performs this
> sequence of operations atomically: check if `addr` contains `ol
Instead of using CMM_ACCESS_ONCE() with memory barriers, use __atomic
builtins with relaxed memory ordering to implement CMM_LOAD_SHARED() and
CMM_STORE_SHARED().
Signed-off-by: Ondřej Surý
---
include/urcu/system.h | 7 +++
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/inclu
FIXME: This is experiment that adds explicit memory barrier in the
free_completion in the workqueue.c, so ThreadSanitizer knows it's ok to
free the resources.
Signed-off-by: Ondřej Surý
---
src/workqueue.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/src/workqueue.c b/src/workqueue.c
inde
When adding REMOVED_FLAG to the pointers in the rculfhash
implementation, retype the generic pointer to uintptr_t to fix the
compiler error.
Signed-off-by: Ondřej Surý
---
src/rculfhash.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/rculfhash.c b/src/rculfhash.c
in
Instead of custom code, use the __atomic builtins to implement the
rcu_dereference(), rcu_cmpxchg_pointer(), rcu_xchg_pointer() and
rcu_assign_pointer().
Signed-off-by: Ondřej Surý
---
include/urcu/arch.h | 20 +
include/urcu/arch/alpha.h | 6 +++
include/urcu/arch/arm.h
Use __atomic_thread_fence(__ATOMIC_ACQ_REL) for cmm_barrier(), so
ThreadSanitizer can understand the memory synchronization.
FIXME: What should be the correct memory ordering here?
Signed-off-by: Ondřej Surý
---
include/urcu/compiler.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
di
Replace the custom assembly code in include/urcu/uatomic/ with __atomic
builtins provided by C11-compatible compiler.
Signed-off-by: Ondřej Surý
---
include/Makefile.am| 16 -
include/urcu/uatomic.h | 84 +++--
include/urcu/uatomic/aarch64.h | 41 ---
include/urcu/uatomic/
Add autoconf checks for all __atomic builtins that urcu require, and
adjust the gcc and clang versions in the README.md.
Signed-off-by: Ondřej Surý
---
README.md| 33 +
configure.ac | 15 +++
2 files changed, 24 insertions(+), 24 deletions(-)
diff
Hi,
(this is my first time using git send-email, so I apologise in advance if
anything breaks).
Here's my attempt to convert the Userspace RCU to use __atomic builtins whenever
possible instead of custom assembly.
The __atomic builtins were first introduced in gcc 4.7.0 and clang 3.1.0.
Apart f
> On 17. 3. 2023, at 14:44, Mathieu Desnoyers
> wrote:
>
> I would indeed like to remove all the custom atomics assembly code from
> liburcu now that there are good atomics support in the major compilers (gcc
> and clang).
Here's very preliminary implementation:
https://gitlab.isc.org/isc-p
> On 17. 3. 2023, at 14:44, Mathieu Desnoyers
> wrote:
>
> Sure, can you please submit the patch as a separate email with subject/commit
> message/signed-off-by tag ?
https://gitlab.isc.org/isc-projects/userspace-rcu/-/merge_requests/1.patch
Would this work for you?
Or do you need to have t
Hey,
we use ThreadSanitizer in BIND 9 CI quite extensively and with userspace-rcu
it's lit like an American house on Saturnalia ;).
I have two questions:
1. I've tried to help TSAN by replacing the custom atomics with __atomic gcc
primitives - that seems to work pretty well. Note that using C
3. 3. 2023, at 15:29, Mathieu Desnoyers
>>> wrote:
>>>
>>> On 2023-03-11 01:04, Ondřej Surý via lttng-dev wrote:
>>>> Hey,
>>>> so, we are integrating userspace-rcu to BIND 9 (yay!) and as experiment,
>>>> I am rewriting the internal a
Hi Matthieu,
I spent some more time with the userspace-rcu on Friday and over weekend and
now I am in much better place.
> On 13. 3. 2023, at 15:29, Mathieu Desnoyers
> wrote:
>
> On 2023-03-11 01:04, Ondřej Surý via lttng-dev wrote:
>> Hey,
>> so, we are integrating
Hey,
so, we are integrating userspace-rcu to BIND 9 (yay!) and as experiment,
I am rewriting the internal address database (keeps the infrastructure
information about names and addresses).
There's a hashtable to keep the entries and there's associated LRU list.
For the hashtable the cds_lfht see
45 matches
Mail list logo
