Repository navigation
nutdrv_hashx: add opt-in USB monitoring for Keen 2000 - #3737
khalilyamoun wants to merge 3 commits into
Conversation
Add bounded HashX status transport for explicitly enabled USB monitoring. One Keen 2000 AC-mode handoff matched nearby vendor readings; cold startup, reconnect, battery operation and shutdown remain unvalidated or unavailable. Signed-off-by: Khalil Yamoun <khalil.yamoun@gmail.com>
|
Preparing a ZIP file with standard source tarball and another tarball with pre-built docs for commit e7c9c99 ... |
jimklimov
added
enhancement
USB
serial port
Incorrect or missing readings
|
Thanks, got a quick reply from CI: it seems the spelling dictionary and/or checked text should be updated. To clarify, when you mention "the HashX status tunnel" - does it use additional information from an additional physical link, or for this device everything can go over USB and the |
| USBFreeExactMatcher(hashx_reopen_matcher); | ||
| USBFreeRegexMatcher(hashx_regex_matcher); | ||
| hashx_usb_free_device(); | ||
| return; |
There was a problem hiding this comment.
Related to the question whether it is strictly USB-or-Serial vs. possibly USB-and-Serial (in the latter case, it should not return here and fall through to ser_close instead).
There was a problem hiding this comment.
Yes, it is USB or serial for each driver instance. The USB branch returns from upsdrv_initups() before ser_open(), so no serial descriptor is opened in that mode. That is why cleanup returns after releasing the USB resources.
|
Thanks! Everything for this monitoring mode goes over USB; there is no second physical link or serial port. By "tunnel" I meant the HashX I also found the spelling failures: |
…and dictionary [networkupstools#3737] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>
|
✅ Build nut 2.8.5.5918-master completed (commit b72f5c02ab by @jimklimov)
|
|
✅ Build nut 2.8.5.5918-master completed (commit b72f5c02ab by @jimklimov) |
Adds experimental USB monitoring to
nutdrv_hashxfor the HashX status tunnel observed on an nJoy Keen 2000 (0764:0601). It requires bothport=autoandusb_monitor. The USB ID is shared by other devices, so a complete valid protocol response is required before publishing telemetry.The USB path sends only the
Bstatus query, drains queued input with a bounded limit, reassembles and validates length-prefixed replies, and marks failed or malformed exchanges stale. Complete zero-load replies are preserved. It registers no USB instant commands and sends no session, settings, battery-test, beeper or shutdown commands. The patch includes protocol/transaction tests, the manual, NEWS and an explicitly experimental one-star HCL entry.One controlled AC-mode USB handoff was tested on one unit after an existing PowerMaster+ session. The driver exited successfully with no status-query errors. Six readings agreed with a nearby passive vendor frame within the comparison tolerances: input 241 V, output 243 V, load 6%, charge 100%, battery 27.2 V and runtime 2400 s. The nearby frame reported 27.1 V battery voltage; the other five values were identical. Samples were not synchronous. NUT reported
OL HBand charger stateresting.PowerMaster+ polling was verified restored afterward. An immediate eight-second observation found no new polls; a later 30-second passive audit captured 15 complete replies and 15 exact successful status queries, with no frame errors or kernel drops. Both vendor services were active and current vendor readings matched the latest passive frame. This establishes current monitoring availability, without proving uninterrupted protection during startup. The experimental driver was not left deployed.
Validation against upstream base
1a8369f:make checkpassed all 14 tests, including 551 checks in the new protocol suite.make distcheck-light, clean-base patch application andgit diff --checkpassed.distcheck-lightpermits placeholder prebuilt manuals; the changed manual was separately rendered witha2x.Cold startup, reconnect/timeout recovery and battery/low-battery status semantics remain unvalidated. Replies have no transaction ID, so a response arriving after the queue drain cannot be conclusively correlated with the new query. USB shutdown is unavailable. Historical intermittent zero-load readings with
usbhid-upswere not reproduced; this provides an alternative telemetry path and does not establish a defect or fix in the current HID decoder.Related report: nJoy Keen 2000 compatibility, #2650.
Contribution preparation used AI assistance from OpenAI Codex (GPT-6; the exact runtime variant is not exposed in this session).