xfs
[Top] [All Lists]

Re: Possible bug in open_by_handle

To: "William L. Jones" <jones@xxxxxxxxxxxxxxxxxx>
Subject: Re: Possible bug in open_by_handle
From: "Andi Kleen" <ak@xxxxxxx>
Date: Wed, 16 Aug 2000 01:29:43 +0200
Cc: "Andi Kleen" <ak@xxxxxxx>, Steve Lord <lord@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <4.2.0.58.20000815171608.016a0c20@127.0.0.1>; from jones@tacc.cc.utexas.edu on Tue, Aug 15, 2000 at 05:36:37PM -0500
References: <200008152007.PAA01650@jen.americas.sgi.com> <jones@tacc.cc.utexas.edu> <200008152007.PAA01650@jen.americas.sgi.com> <20000815222734.A13055@gruyere.muc.suse.de> <4.2.0.58.20000815171608.016a0c20@127.0.0.1>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Tue, Aug 15, 2000 at 05:36:37PM -0500, William L. Jones wrote:
> 
> 
> I just looked at the nfsfh.c code.   It looks like it just calls
> d_alloc_root if it cannot find a non anonymous indoe in the dcache.  Is this
> what you mean by an anonymous dentry?

Yes.

> 
> Why is the nfsfs.c code so carefully to look for a non anonymous dcache entry.
> Is it just trying to save space in the dcache?

For some access checking the nfsfh code needs to know the full path name
of the file. For that it needs to notice when the filehandle is deleted,
which is easiest when the dentry is connected to the normal dentry tree
(otherwise it has to search the directory) 

For the "virtual hardlinks" the xfs file handle code uses this is not
a problem, because they don't need file names and cannot be deleted.


-Andi



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