#include <hallo.h> * Thomas Hood [Sun, Jan 15 2006, 10:51:34AM]: > That 'umount -a' excludes procfs _can_ be regarded as a precedent for making > that command also exclude sysfs. > > However, I would nevertheless suggest that this request not be implemented. > It is disruptive to change the established behavior of programs, and I don't
Heh? > see a strong reason for changing the behavior of "-a", given that users can > already exclude any filesystem types they choose by means of the '-t' option > and the 'no' prefix. > > Another point is that "-a" in combination with "-t no..." gives the user > complete > control over which types of filesystems are unmounted (except for procfs); > but if > sysfs were excluded by default then there would be no way of overriding this > and > including it again SFAICS. I am not talking about some "control" virtue. I am talking about making simple things complicated for no good reason. Read: without /sys, a lot of things go wrong nowadays, especially udev. Yes, this can be avoided by adding a dozen of magic options to each command and thinking an hour before you enter every chear on your keyboard, but... jeez, what is this -a option good for? You have the same amount of "control" by reading /etc/mtab yourself, great, isn't? > In any case, I can't see how this issue rises above severity: wishlist. Hahaha. Just coming back to your "established behaviour of programs" - a suspend script on the system here just fscked up the uptime because hardware aliases were missing after suspend2 cycle. Why? Not obvious to see, even udevmonitor did not give a hint. Reason: /sys was umounted because a "harmless" umount -a command. Feel free to call that a "chain of incidents immediately leading to a failure". Unexpected umounting of /sys was the trigger. Eduard. -- <meebey> köstlich * meebey schlürft gerade selbstgemachten Swimming Pool Cocktail <Moermel> meebey: Wasser+Harnstoff?
