From: Reco <recovery...@enotuniq.net> Date: Wed, 25 Nov 2020 10:42:59 +0300 > No. "auto" means (taken directly from interfaces(5)): > ...
OK, thanks. Several years since I last read the man page. Can't say it's inspiring prose; absolutely no offense to authors; just my honest impression. > The main difference between "auto" and "allow-hotplug" is, well, hotplug > processing. And judging from that "predictable" interface name, you're > using an USB dongle, so hotplug is important here. Righto. From a naive English language perspective, hot-plug processing is a form of automation. So "hot-plug" is a subset of "auto". But, as you say, not really. I'll speculate that these directives came from evolution rather than comprehensive design. "In the beginning" there was coax Ethernet. When a programmer was sufficiently tired of starting links manually, auto was invented. Later USB came along. When a programmer was sufficiently tired of raising and lowering a link, after plugging or unplugging an adapter, allow-hotplug was invented. With twisted pair I've been in the habit of "setting and forgetting" both directives. With an access point which shuts down unpredictably I've omitted both directives and use ifup and ifdown interactively. If review and overhaul of the software is announced in the next decade, that won't be too shocking. > The kernel has no free RAM to queue a packet or that Tp-Link device > you're is using low-quality kernel module. Happens with Tp-Link, but > there's a bright side - it could've been Broadcom. > > Try increasing a value of vm.min_free_kbytes, it may help. Righto, thanks. The TP-Link TL-WN722N adapters were recommended here a few years back. If someone has better advice now, I'm interested. Blue-sky idea. FPGAs are routinely available. Is anyone working on a FPGA wifi adapter? That would eliminate the specs. mystery. Regards, ... P. -- Tel: +1 604 670 0140 Bcc: peter at easthope. ca
