summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2026-11-02Prepare a versioning mechanism for the tracker that works from tarballsDirk Engling
2026-11-01Live logging networks was defunc for a while, does not work with v6, so we remove itDirk Engling
2026-11-01Add some tests for the actual trackerlogic, something I put off for too longDirk Engling
2026-11-01Add tests for the ot_mutex code, it didn't have any until nowDirk Engling
2026-11-01Remove some obsolete commentsDirk Engling
2026-11-01Generalize the worker setup/runloop/teardown logic by adding a smol ot_worker helperDirk Engling
2026-10-31Make the dynamic accesslist readers release their line buffer and properly exitDirk Engling
2026-10-31Refactor ot_workqueue out of ot_mutex, no behaviour changeDirk Engling
2026-10-31Make the mainloop behave like every worker by claiming and releasing a socket being worked onDirk Engling
2026-10-31Make tcp worker multithreaded, tooDirk Engling
This shamelessly steals the model established in the UDP workers, i.e. if the kernel is able to bind multiple listening sockets to the same endpoint, we will share them across threads. If it can't, all threads will accept on the same socket. The fast path handles most of the connections, only the rare case of no data yet on the socket is offloaded to the slow path (which used to be the mainloop) for completion. This scales nicely, as the higher the load is, the higher chances are that an accept happens only after the network roundtrip to the peer completes, leading to more connections that already have data waiting for us. We also consolidate the workstruct init and deinit code, preparation for later unifying all thread state structs for all protocols (mca, udc and tcp). The config file gains a per tcp connection worker count and some docs explaining what this knob buys and costs, nudging the user to either keep it 0 or dial it up to at least 2.
2026-10-31Refactor all tcp specific code into the ot_tcp module to prepare for multiple concurrent workersDirk Engling
2026-10-31Add test for new config paramsDirk Engling
2026-10-31Add config parsers for multiple tcp socket readers and a helper to see if we can offer it on the platformDirk Engling
2026-10-30Introduce a fast path assuming that a tcp connection's payload already has arrived at accept timeDirk Engling
2026-10-30Start documenting design decisions and measurmentsDirk Engling
2026-10-30Remove references that you need libowfat to build opentrackerDirk Engling
2026-10-30git ignore the poll flavor of the runner, tooDirk Engling
2026-10-30Fix idle tests in presence of an accept filterDirk Engling
2026-10-30Remove stale xcode build fileDirk Engling
2026-10-30Fix expiry tests under the new accept filtersDirk Engling
2026-10-30Use accept filter dataready instead of httpready on FreeBSDDirk Engling
2026-10-30Defer accepts until some data actually is thereDirk Engling
2026-10-30Enforce the invariant that each socket is only ever expected to be read from or written to, never bothDirk Engling
2026-10-30Don't treat a full send buffer as an error when sending http. Retry laterDirk Engling
2026-10-30Reduce syscalls and add helper to count themDirk Engling
2026-10-30Add an http load generatorDirk Engling
2026-10-30Reduce number of per connection sys callsDirk Engling
2026-10-30Check for free port in the udp integration test, also speed it up by a bitDirk Engling
2026-10-30Consolidate all tcp/http tear down into a single http_close function to prevent a rare UAF for expiring connectionsDirk Engling
2026-10-29Build everything in one sweep, without intermediate object filesDirk Engling
2026-10-29add wake-up tester that helped figuring out how to tickle hanging udp readersDirk Engling
2026-10-29Revert last wake-up experiment. Having each udp worker wake up once a second to check if the tracker stops is wastefulDirk Engling
2026-10-28Fix a test that would over-report success rate when losing the race for cycles to the test targetDirk Engling
2026-10-28Livesync now also gets proper deinit codeDirk Engling
2026-10-28With more workers blockingly waiting on udp sockets, we need to keep track of all sockets to shut them down on exitDirk Engling
2026-10-28Default udp workers to 1, as it is the most sane choice everywhere, also reflect the new udp parallelism in the man pagesDirk Engling
2026-10-28Let the load tester log losses to better gauge performanceDirk Engling
2026-10-28Add load tests to sanitize check the trackerlogic, based on udp requestsDirk Engling
2026-10-28Performance improvement, where it works: Let different udp worker threads have their own socketDirk Engling
2026-10-27The main event loop(s) now also catch an EINTR and return instead of blockingDirk Engling
2026-10-27Add concurrency testDirk Engling
2026-10-27Make global torrent counter atomic and fix the setter and getterDirk Engling
2026-10-27Plug in the new per-thread stats counter, also make sure that the ot_clean sleep does not stall tracker exitDirk Engling
2026-10-27Second half of moving the stat enum defines to trackerlogicDirk Engling
2026-10-27Prepare having each thread keep their own stats to avoid data racesDirk Engling
2026-10-27Turn malloc+memset into callocDirk Engling
2026-10-27Some more g_now_seconds occurrencesDirk Engling
2026-10-27Forgot some g_now_seconds occurencesDirk Engling
2026-10-27All OSes now do have a cheap monotonic clock, so no caching thread is needed. Also add some relative time helpersDirk Engling
2026-10-27While we're at it with the dynamic analyzers, add a tsanDirk Engling