Get the latest tech news

Spoiling Linux Kernel with "sanctioned" code

…sending patches which can't be accepted, ever? Last year I've fixed a bug in Linux kernel OHCI (USB 1.1) stack, which rendered my old printer unreliable: OHCI adds artificial delay of 1ms for each USB transfer of send-read pattern, while my very old, host-based printer expects very tight timings and constant latency, like some kind of CnC machine. This artificial delay was introduced in pre-git era (March 2004) as a race condition fix for the USB OHCI host controllers of the time: Endpoint Descriptors (EDs) are unlinked as soon as they become idle, but only after receiving next Start of Frame (SOF) interrupt.

None

Get the Android app