mounting a loopback file from a link fails with busybox 1.2.1

Yann Le Doaré yledoare at linuxconsole.org
Tue Aug 8 20:35:14 UTC 2006


Yes, it fixes it.

Good job Rob !

Thank you.

Yann. 

Le lundi 07 août 2006 à 20:57 -0400, Rob Landley a écrit :
> On Monday 07 August 2006 2:26 pm, Yann Le Doaré wrote:
> > Hi, I have a problem with busybox 1.2.1 (1.0.1 relase is OK for that) :
> > When I want to mount a lookback file from a link, I have the "Block
> > device required" error.
> 
> Sigh.  Darn it.  It's using lstat() instead of stat.  Why was I doing that...? 
> In case of a dead symlink, which is treated as a file not found.  But that's 
> probably the behavior we _want_, so...
> 
> Try svn 15788?
> http://busybox.net/downloads/patches/svn-15788.patch
> 
> > ./busybox mount /tmp/testfslnk /mnt/floppy/ -o loop
> 
> You don't need to say "-o loop" anymore, that's treated as a NOP by the 
> current busybox mount command.  It'll autodetect when it needs to do a 
> loopback mount (or a bind mount for that matter).  This bug is a case of the 
> autodetect code failing.
> 
> Rob
> 
> P.S.  Sigh, I'm now ~65 messages behind on the list, blocked untangling 
> the "add CIFS and shared subtree support to mount" patch Vladimir Dronnikov 
> sent me off-list, which is hard because the shared subtree part is something 
> I want a CONFIG_NITPICK option for, but it turns out that the config 
> infrastructure we lifted from the linux kernel switches off everything that 
> depends on a symbol that is switched off.  It doesn't revert invisible 
> symbols to their default value, it switches them _OFF_, and there's no 
> obvious way to control visibility independently of selection...
> 
> And _this_ is why you shouldn't send one patch that does two things.  The 
> shared subtree support is trivial compared to CIFS support, but CIFS is 
> blocked on shared subtree because they were submitted as one patch...
> 
> Rob




More information about the busybox mailing list