[Top] [All Lists]

Re: [xfs-masters] [Bug 48651] [TRIVIAl]'xfs_uuid_table' allocated in xfs

To: nickkrause@xxxxxxxxxxxx, sandeen@xxxxxxxxxx
Subject: Re: [xfs-masters] [Bug 48651] [TRIVIAl]'xfs_uuid_table' allocated in xfs_uuid_mount() is never freed
From: Christoph Hellwig <hch@xxxxxx>
Date: Tue, 20 May 2014 16:40:46 +0200
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <bug-48651-470-8kNPGrGVTO@xxxxxxxxxxxxxxxxxxxxxxxxx/>
References: <bug-48651-470@xxxxxxxxxxxxxxxxxxxxxxxxx/> <bug-48651-470-8kNPGrGVTO@xxxxxxxxxxxxxxxxxxxxxxxxx/>
User-agent: Mutt/1.5.17 (2007-11-01)
On Tue, May 20, 2014 at 04:01:20AM +0000, bugzilla-daemon@xxxxxxxxxxxxxxxxxxx 
> https://bugzilla.kernel.org/show_bug.cgi?id=48651
> --- Comment #2 from Eric Sandeen <sandeen@xxxxxxxxxx> ---
> Freeing it when the module is unloaded should be sufficient.
> In normal use, unmounting a filesystem creates a hole in the table, which the
> next mount can fill.
> If you mount a lot of unique UUID filesystems without unmounting any in
> between, the table will grow accordingly, and unmounting won't shrink it - 
> i.e.
> the table only grows as the number of concurrently mounted filesystem grows,
> but that doesn't seem like a particularly critical problem to fix.

If anyoneîreally cares about this he or she should just kill the uuid
table entirely.  We can simple walk a list of mounted xfs filesystems
and check their uuids.  The VFS even keeps a list already, which we
probably could use if we are careful.

Bonus points for implementing this in a way so that other filesystems
could share the code.

<Prev in Thread] Current Thread [Next in Thread>
  • Re: [xfs-masters] [Bug 48651] [TRIVIAl]'xfs_uuid_table' allocated in xfs_uuid_mount() is never freed, Christoph Hellwig <=