Subject: Re: Problem accessing Windows 98 partition
To: None <LetsGoCanes@webmail.bellsouth.net, martin@rumolt.teuto.de>
From: Wolfgang Solfrank <ws@tools.de>
List: port-i386
Date: 07/03/2000 16:49:43
Hi,
> > Next free cluster in FSInfo block (2) not free
>
> This is realy scary! (Background: the filesystem claims to have free space
> [in allocation units called "clusters"], but the list of free clusters
> starts with one that is not marked as free.)
Actually it's not. What happens is this:
With FAT32 filesytems, there is an extra information block at the start
of the filesystem. According to the FAT32 spec, this information block
contains (as one part) a pointer to some free cluster. The NetBSD kernel
code actually does maintain this pointer, and fsck_msdos checks it, too.
It turns out however that M$ software doesn't give a damn about this
(contrary to their own spec). Net result is that this pointer might
point to some non-free cluster, after you happen to write something to
your FAT32 partition with W9x.
(As an aside, I'm not even sure what they tried to achieve with this.)
> While the fsck_msdos implementation is not idetnical to the in-kernel msodsfs
> implemation, I wouldn't dare to mount it writeable at this point.
For now you can ignore the message. It doesn't even change the exit
status of fsck_msdos, independent of whether you let it fix this or
not (of course, trying to fix it can lead to an i/o error which will
be reflected in the exit status).
> This might be an errror in the fsck_msdos implementation, though.
Hmm, probably the message should more clearly state that this is
non-fatal. Or we should drop maintenance of this field altogether.
Ciao,
Wolfgang
--
ws@TooLs.DE Wolfgang Solfrank, TooLs GmbH +49-228-985800