Commit Graph

106 Commits

Author SHA1 Message Date
Alexey Rybalchenko
1597999aed fix(shmem): don't cache nullptr in GetRegionFromCache
A failed region lookup was inserted into the thread-local cache as
nullptr, making the failure permanent for the lifetime of the cache
generation - retrying never healed because the fast path would return
nullptr without calling GetRegion again. Skip the cache insert on
failure so subsequent calls retry the slow path.
2026-06-10 18:51:04 +02:00
Alexey Rybalchenko
215c31428b feat(shmem): expose side-channel metadata API for unsent messages
Add two public entry points needed by the ALICE use case where shmem
messages are allocated via a transport but never sent — their metadata
is instead serialised into Arrow tables and delivered over a separate
channel, allowing consumer devices to resolve the payload pointer
without taking ownership.

shmem::Message::GetMeta() returns the MetaHeader of the message,
mirroring the existing positional-init pattern already used in Socket.h.

shmem::GetDataAddressFromHandle(TransportFactory&, const MetaHeader&)
is a free function declared in Common.h and defined in Manager.cxx.
Keeping it out of the TransportFactory class body means callers only
need to include Common.h (available transitively via Message.h) and do
not drag in Socket.h or zmq.h. The implementation handles both managed
segments and unmanaged regions, and throws SharedMemoryError with a
typed message on a bad segment or region id. TransportFactory also
gains a same-named member for callers that already have the concrete
type. Lifetime of the returned pointer is the caller's responsibility;
the cache device is expected to hold the messages alive.

A SideChannel test covers the GetMeta/GetDataAddressFromHandle
round-trip for both standard and expanded-metadata configurations.
2026-06-10 18:51:04 +02:00
Dennis Klein
d5e0c29ced perf: emplace elements instead of inserting temporaries
- construct container elements in place instead of inserting a temporary
- clang-tidy modernize-use-emplace
  https://clang.llvm.org/extra/clang-tidy/checks/modernize/use-emplace.html
2026-06-09 23:00:58 +02:00
Alexey Rybalchenko
faf577086a shm: fix initialization of rc segment when region is created externally 2023-11-24 14:19:21 +01:00
Alexey Rybalchenko
8d28824489 Shm: Use MakeShmName to construct shm object names 2023-09-29 11:18:24 +02:00
Alexey Rybalchenko
7bd31f8ff0 apply readability-else-after-return 2023-09-29 11:18:24 +02:00
Alexey Rybalchenko
1a0ab3a4e2 shm: Ref counting for unmanaged regions in a dedicated segment 2023-09-29 11:18:24 +02:00
Alexey Rybalchenko
cacf69d5f6 Replace boost::variant with std::variant 2023-09-29 11:18:24 +02:00
Alexey Rybalchenko
68038c4693 shm: Move ShmHeader into Common.h 2023-09-29 11:18:24 +02:00
Dennis Klein
f278e7e312 feat: Add new tunable --shm-metadata-msg-size
The shm metadata msg will be right-padded to the given size. This
tunable may be used to saturate the kernel msg buffers more quickly with
the effect that the ZeroMQ message queue size - on which the FairMQ
shmem transport relies upon - behaves more accurately for very small
queue sizes.

This introduces a change for the meta msg format in the multipart case:
old: | MetaHeader 1 | ... | MetaHeader n |
new: | n | MetaHeader 1 | ... | MetaHeader n | padded to fMetadataMsgSize |
where `n` is a `size_t` and contains the number of following meta headers.
Previously, this number was infered from the msg buffer size itself which is
no longer possible due to the potential padding.

Implements #432
2023-06-13 21:24:40 +02:00
Dennis Klein
8960ce9416 fix: Use std::chrono consistently 2023-03-06 15:32:48 +01:00
Dennis Klein
05b734ee0d feat!: Migrate to std::filesystem consistently 2023-03-06 15:32:48 +01:00
Piotr Konopka
9389030835 Warn when going to try allocate a shmem message more times 2022-11-18 09:36:52 +01:00
Alexey Rybalchenko
ed364a4857 shm: error on duplicate region IDs 2022-10-05 09:13:37 +02:00
Alexey Rybalchenko
9a25c4d28a shm: refactor UnamangedRegion: rename fRemote to fController 2022-10-05 09:13:37 +02:00
Alexey Rybalchenko
19e40bd32e shm: Improve debug output a bit 2022-10-05 09:13:37 +02:00
Alexey Rybalchenko
de09018198 Shm: fix region init with external regions 2022-09-09 15:11:12 +02:00
Alexey Rybalchenko
75e68e3e4d fix(shm): Fix incorrect parameters when mapping regions 2022-09-06 07:57:45 +02:00
Alexey Rybalchenko
4587af2eb4 fix(shm): race/deadlock in region locks 2022-08-19 12:24:35 +02:00
Alexey Rybalchenko
29827f0426 Shm: bring back thread-safety for fRegions (intra-process) 2022-02-16 23:17:04 +01:00
Alexey Rybalchenko
8efe7adf0e Shm: fix number of region events 2022-02-16 23:17:04 +01:00
Alexey Rybalchenko
1a75141fc4 shm: allow monitor::ResetContent to cleanup after a crash 2022-02-02 10:49:00 +01:00
Alexey Rybalchenko
ddf9bc7272 shm: keep mng segment around when skipping cleanup 2022-01-12 19:54:49 +01:00
Alexey Rybalchenko
f79a0714b4 shm: fix double unlock() 2022-01-12 19:54:49 +01:00
Alexey Rybalchenko
c04958e2a4 shm: reduce contention on region events 2022-01-10 19:42:08 +01:00
Alexey Rybalchenko
692576a5b1 shm: add APIs for implementing keep-alive process 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
eb4620b1ec shm: always open_or_create segment 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
08ba068791 shm: remove unused member 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
1839f7e8c0 shm: integrate mtx and cv into management segment 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
80ed45df63 extend region config 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
eef42d2dea simplify region cleanup 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
d630fbb1e4 consolidate UnmanagedRegion options 2021-12-16 16:27:07 +01:00
Alexey Rybalchenko
d7fb01908c fix(shm): fix regression in debug mode data 2021-10-08 09:42:28 +02:00
Alexey Rybalchenko
1449166d44 feat: add options to control allocation attempts 2021-10-08 09:42:28 +02:00
Alexey Rybalchenko
153dcfab94 Apply modernize-use-override 2021-09-27 12:04:07 +02:00
Alexey Rybalchenko
ad824b4de1 Define copy/move ctors and assignment ops
Delete special member functions where they are not used.
(as part of applying suggestions from cppcoreguidelines-special-member-functions)

These classes don't need to be copyable/movable:
  # copy/move not used:
  zmq:: TransportFactory, Socket, Message, UnmanagedRegion, Poller, Context
  shm:: TransportFactory, Socket, Message, UnmanagedRegion, Poller
  ofi:: TransportFactory, Socket, Message, Context
  shm:: ZMsg, Region, Monitor, TerminalConfig, Manager
  plugins:: Config, Control, TerminalConfig
  fairmq::StateQueue, StateMachine, ProgOptions, PluginServices, PluginManager, Plugin, Device, StateSubscription
  TestData, BadDevice, TestDevice (test suite heplers)

  # used via ptr interface:
  fairmq::UnmanagedRegion, TransportFactory, Socket, Poller, Message

These classes need to be movable/copyable:
 MyClass (test suite helper), fairmq::Channel, fairmq::Parts
2021-09-27 12:04:07 +02:00
Alexey Rybalchenko
597d88277b Apply clang-analyzer-optin.performance.Padding 2021-09-27 12:04:07 +02:00
Alexey Rybalchenko
e484bf4578 Shm: Fix SetUsedSize() 2021-09-20 13:29:28 +02:00
Alexey Rybalchenko
bce380d871 Implement shmem msg zero-copy 2021-09-07 20:53:16 +02:00
Alexey Rybalchenko
815b2f1d76 shm: reimplement alignment 2021-09-07 20:53:16 +02:00
Dennis Klein
1007de8e49 feat(ofi): Require asiofi 0.5
* Modernize some ofi transport code along the way
  * Replace Boost.Container with `<memory_resource>`
  * Introduce namespaced headers
    * `<fairmq/Channel.h>`
    * `<fairmq/Message.h>`
    * `<fairmq/Poller.h>`
    * `<fairmq/Socket.h>`
    * `<fairmq/TransportFactory.h>`
    * `<fairmq/UnmanagedRegion.h>`
* Compile-firewall Boost.Process in `shmem::Manager` because it conflicts
with standalone asio
2021-07-16 12:31:17 +02:00
Alexey Rybalchenko
03ba9eb558 Add --shm-zero-segment-on-creation option 2021-07-16 09:40:17 +02:00
David Rohr
a6193a380d Add --shm-mlock-segment-on-creation option 2021-07-16 09:40:17 +02:00
Alexey Rybalchenko
a8bdb91165 shm: throw TransportError if could not lock region 2021-06-15 11:54:29 +02:00
Alexey Rybalchenko
37c059177f shm: improve exception handling 2021-06-15 11:54:29 +02:00
Alexey Rybalchenko
28a887a457 shm: optimize monitor heartbeats 2021-06-15 11:54:29 +02:00
Alexey Rybalchenko
9444de5868 Apply readability-redundant-member-init 2021-05-28 13:14:51 +02:00
Alexey Rybalchenko
9bf908fb52 shm: revert some changes from c85d6e0 that introduced a race 2021-05-20 00:40:58 +02:00
Alexey Rybalchenko
091d0824d1 ofi: fix Events() signature 2021-05-07 21:33:15 +02:00
Alexey Rybalchenko
857aa84fa3 add mlock/zero options to unmanaged region 2021-05-07 21:33:15 +02:00