Subject: Re: device attachement detection from userland
To: None <>
From: Michael van Elst <>
List: tech-kern
Date: 10/14/2006 10:53:48 (Emmanuel Dreyfus) writes:

>IMO, amd badly sucks. You have to actually access the filesystem to get
>it mounted,


>and the first access can get painfuly delayed.

I wouldn't see this as a problem.

>It can even
>just hang if there is a problem.

Now, that happens with any attempt to mount a device with "a problem".

>That problem is especially sensible
>with slow devices such as floppy or CD-ROM. 
>Moreover, removable devices partitions are not always at the same
>disklabel letter.

I consider this a problem of the disklabel. There is simply no
information to tell what partition to access. Named partition
labels could help.

Simply mounting all partitions on the device isn't really a solution.

>Some USB key feature DOS partitions with a MBR (and
>pop up as sd0e), some without (and pop up as sd0d or sd0c). I just don't
>know how to handle that properly with amd.

You can make amd run a script that does the decision and the mounting.

>> The tricky part however is: how do you manage removal of the device?

>You can mount with synchronous I/O, and umount -f when the device gets
>removed. Won't that work? 

It never did for me but I didn't try with 3.x or later and it doesn't
solve the problem of removing a device without prior umounting.

Especially the latter is something amd avoids as it doesn't mount
everything in the first place and umounts it quickly when possible.

                                Michael van Elst
                                "A potential Snark may lurk in every tree."