Skip to content

Release 0.1.0-alpha.10.2

A small release, for the XGS 107w of issue 890, whose driver refuses or drops one received frame in five: npuep looks again, for up to 10 us, at a receive slot that cannot be a frame before refusing it, counts what the second look found, and keeps the last slot still wrong - its descriptor and its buffer's shape, without an address - in dev.npuep.0.giu.rx_refused_last. On an XGS 107 that tells frames that land late from slots written in another form. Measured on the XGS 136 (vendor version 21): no second look in 584,006 frames over its LAN and WAN; not run on an XGS 107 - there is none here. With it: what the coprocessor's console prints is kept, its kernel's state is asked for the moment its fast path stops, and the recovery waits for its new boot (measured on the first XGS 3300, vendor version 23); and verify.sh is installed on the XGS 2100, 2300, 3100 and 3300 too (run by the installer on the first XGS 3300, and from the source copy by an XGS 3100's owner).

Improvements and additions

  • Coprocessor stops: what its console prints kept, its kernel's state at the stop, a recovery that waits for this boot (#819) (#909)
  • verify.sh installed on both families (#914)
  • npuep: a second look at a slot that cannot be a frame (#915)

Documentation

  • docs: the fans follow the CPU's temperature; the 1506 struck where it still stood (#907)
  • Improve documentation wording (#910)

The release, with its files: v0.1.0-alpha.10.2 on GitHub.

Discussion

The discussion of this page is held in the project's GitHub Discussions. Showing it here loads giscus from giscus.app and GitHub, and writing in it needs a GitHub account - what that means.