Re: nosh version 1.31

From: Jonathan de Boyne Pollard <>
Date: Mon, 23 Jan 2017 22:36:59 +0000

> The problem seems to be those whose corresponding array of iovecs
> doesn't have a 'from' element,

I hit this with cgroupfs and tmpfs in version 1.28 and thought that it
was just specific to those filesystem types. The reason that I never
hit it with procfs, devtmpfs, and sysfs is actually tied in to your
other problem.

> And the catch-all logger reports this:

Reading the logs is always a good thing, I have found. In this case, it
meant that I knew what was going on in just a couple of minutes.
Otherwise I'd be wondering why this worked in my system when I
bootstrapped into the older Debian 8 kernel to test what happens when
only v1 control groups are available, but it didn't work for you.

Debian's initramfs mounts /proc, /sys, /run, and a whole lot of other
stuff before /sbin/init is run. Yours does not. I see far more
messages in my log than you do. So:

1. There's a chicken-and-egg situation with having to read
/proc/filesystems in order to determine what API filesystems can be
mounted. Good old Linux, eh?

2. The procfs, devtmpfs, and sysfs problems did exist, but I never hit
them because on Debian they had already been mounted with a "from" name.

I've already fixed the "from" problem ready for 1.32. I'll have a go at
fixing the other one, too. There are simply going to have to be two
phases of API filesystem mounting. Three on Debian. (-:
Received on Mon Jan 23 2017 - 22:36:59 UTC

This archive was generated by hypermail 2.3.0 : Sun May 09 2021 - 19:44:19 UTC