devfs
[Top] [All Lists]

Re: modules.devfsd needs /dev/scsi/*/part* handling

To: Richard Gooch <rgooch@xxxxxxxxxxxxxxx>
Subject: Re: modules.devfsd needs /dev/scsi/*/part* handling
From: Thierry Vignaud <tvignaud@xxxxxxxxxxxxxxxx>
Date: Tue, 22 Jan 2002 14:51:39 +0100
Cc: Borsenkow Andrej <Andrej.Borsenkow@xxxxxxxxxxxxxx>, devfs@xxxxxxxxxxx
In-reply-to: <200201210108.g0L18uQ17789@xxxxxxxxxxxxxxxxxxxxxxxx> (Richard Gooch's message of "Sun, 20 Jan 2002 18:08:56 -0700")
Organization: MandrakeSoft
References: <1008885491.6464.4.camel@xxxxxxxxxxxxxxxxxxxxx> <200201150628.g0F6SQf06569@xxxxxxxxxxxxxxxxxxxxxxxx> <m21ygrurjk.fsf@xxxxxxxxxxxxxxxxxxxxxx> <200201152319.g0FNJ5S20314@xxxxxxxxxxxxxxxxxxxxxxxx> <m2g057tbsz.fsf@xxxxxxxxxxxxxxxxxxxxxx> <200201152353.g0FNr5j21148@xxxxxxxxxxxxxxxxxxxxxxxx> <1011209242.4046.5.camel@xxxxxxxxxxxxxxxxxxxxx> <200201210108.g0L18uQ17789@xxxxxxxxxxxxxxxxxxxxxxxx>
Sender: owner-devfs@xxxxxxxxxxx
User-agent: Gnus/5.090004 (Oort Gnus v0.04) Emacs/21.1
Richard Gooch <rgooch@xxxxxxxxxxxxxxx> writes:

>> Attached patch is against current Mandrake devfsd.conf. It works just
>> fine here for SCSI; I'd appreciate anybody testing it for IDE. It is
>> great because it means we can continue to use compatibility names by
>> default without endless "hey, why it says no such device!"=20
>> 
>> blockdev is garanteed to exist on Mandrake that is currently my concern
>> :-)
>> 
>> many thanks for idea (yes, sometimes you must really read TFM).
>
> Well, thanks, but since blockdev isn't available everywhere, I can't
> really use it in a generic configuration file. However, I wonder if
> blockdev is actually required. IIRC, merely opening the block device
> should suffice to trigger media revalidation. Is this not the case?
> What is blockdev doing that is special?
>
> Can you please try to use dd instead in your devfsd.conf file and tell
> me if that works? An action like this:
>       EXECUTE dd if=$mntpnt/\1 of=/dev/null count=1

this would add a dependancy on fileutils. as for mandrake, it's required by
basesystem package, so it's ok for us. should be the same for rh & debian, but a
check wouldn't hurt.

> should suffice. It will probably generate some messages about 1+0
> blocks read and so forth, but ignore that for now.


<Prev in Thread] Current Thread [Next in Thread>