Commit Graph
29 Commits
Author SHA1 Message Date
Vitaliy Filippov 32721046b8 WIP Switch back to TCP when RDMA send fails (does not work) 2026-05-01 22:26:41 +00:00
Vitaliy Filippov f8de201625 Close TCP connections for RDMA clients 2026-05-01 22:13:29 +00:00
Vitaliy Filippov d0a4ac914b Implement OSD TLS support 2026-04-27 15:24:44 +03:00
Vitaliy Filippov e42b812323 Allow to skip checksums for headers 2026-04-27 15:24:44 +03:00
Vitaliy Filippov ac5c753735 Implement protocol-level checksums (xxhash3) 2026-04-27 15:24:44 +03:00
Vitaliy Filippov d7f854ab3e Batch handle_immediate_ops more 2026-04-27 15:24:44 +03:00
Vitaliy Filippov 713b17e574 Add basic AES-XTS client-side encryption support 2026-04-27 15:24:44 +03:00
Vitaliy Filippov 4efa48c9ca Rework msgr send/receive to allow encryption support 2026-04-27 15:24:44 +03:00
Vitaliy Filippov bcc6419760 Auto-select the only RDMA device/port if there is only one 2026-04-08 15:46:10 +03:00
Vitaliy Filippov 07915c2881 Use surrogate peer IDs instead of FDs 2026-03-30 02:06:10 +03:00
Vitaliy Filippov 8f8172db99 clear_immediate_ops is now done by stop_client 2026-03-26 01:08:23 +03:00
Vitaliy Filippov 44c895dc30 Fix clients possibly not destroyed after being switched to RDMA
It could in theory lead to hung operations on RDMA client disconnections.

Workflow (probable):
* OSD switches an outbound connection to PEER_RDMA and removes the FD from epoll
* But a receive request in io_uring is still active
* Receive response is never handled, refs remains > 0
* The connection is dropped, but osd_client_t is not destroyed because refs > 0
* Outbound ops are not canceled, suboperations hang, primary operations hang too
2026-03-17 01:04:17 +03:00
Vitaliy Filippov 2ed6760447 Remove ODP support 2026-02-17 02:35:46 +03:00
Vitaliy Filippov 9ec18f0aa1 Fix double ibv_close_device on RDMA init failure 2026-02-11 00:32:12 +03:00
Vitaliy Filippov fc5a183959 Allow routed RoCE 2025-12-12 15:06:49 +00:00
Vitaliy Filippov 55de37e58a Use a single memory region per client instead of separate mrs per each malloc 2025-12-12 15:05:33 +00:00
Vitaliy Filippov aacfdf0dec Rename try_recv_rdma to init_recv_rdma 2025-12-07 21:06:36 +03:00
Vitaliy Filippov 9fa8ae5384 Reset OSD ping state on receiving data from it via RDMA 2025-04-26 14:17:09 +03:00
Vitaliy Filippov 80e5f8ba76 Add missing WITH_RDMACM defines 2025-04-12 19:06:27 +03:00
Vitaliy Filippov 8508e78288 Add an alternative RDMA implementation via RDMA-CM
Required for non-RoCE cards: iWARP and, possibly, Infiniband
2025-03-31 21:01:25 +03:00
Vitaliy Filippov f32dea02bf Support multiple RDMA networks 2025-03-31 21:01:25 +03:00
Vitaliy Filippov a103065d12 Support multiple OSD networks and separate OSD cluster network 2025-03-31 21:01:15 +03:00
Vitaliy Filippov 8b0389b4e8 Log RDMA ibv_modify_qp() errors 2025-03-22 15:58:13 +03:00
Vitaliy Filippov 63b85b6bfb Fix clang warnings/errors 2024-12-19 15:30:31 +03:00
Vitaliy Filippov 1f6a061283 Move ibv_query_gid under #ifdef to only build it with libibverbs 32+ 2024-11-21 23:47:57 +03:00
Vitaliy Filippov ddea31d86d Auto-select first RDMA device only if RoCE is not found, add rocev2->rocev1->ib priority 2024-11-19 01:54:00 +03:00
Vitaliy Filippov 7de38250ad Auto-select RDMA device based on osd_network 2024-11-16 18:38:57 +03:00
Vitaliy Filippov c312557ace Do not execute remaining operations if the client is stopped during read 2024-11-10 16:44:13 +03:00
Vitaliy Filippov c79b38bd26 Move all sources to subdirs 2024-05-15 11:06:01 +03:00