Re: [Xen-devel] [PATCH 0 of 3] libxl: leak in uuid to string conversions

[Gianni Tedesco <gianni.tedesco@xxxxxxxxxx>]

On Mon, 2010-08-16 at 15:05 +0100, Stefano Stabellini wrote:
> On Mon, 16 Aug 2010, Ian Campbell wrote:
> > On Mon, 2010-08-16 at 14:51 +0100, Stefano Stabellini wrote:
> > > On Mon, 16 Aug 2010, Ian Campbell wrote:
> > > > On Mon, 2010-08-16 at 13:51 +0100, Gianni Tedesco wrote:
> > > > > On Fri, 2010-08-13 at 16:16 +0100, Ian Jackson wrote:
> > > > > > Ian Campbell writes ("[Xen-devel] [PATCH 0 of 3] libxl: leak in 
> > > > > > uuid to string conversions"):
> > > > > > > Fix a few memory leaks in the functions which convert uuids into
> > > > > > > strings.
> > > > > > 
> > > > > > Acked-by: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
> > > > > > 
> > > > > > (all three)
> > > > > > 
> > > > > > Ian.
> > > > > 
> > > > > I'm not sure I agree with this patch, especially 3 of 3. The callers 
> > > > > are
> > > > > inconsistent and also pointless. There's no need to allocate a uuid
> > > > > string when all that's really needed is a few printf macros as in
> > > > > "[PATCH,v2] xl: make libxl_uuid2string internal to libxenlight" - Also
> > > > > UUID_FMT remains duplicated.
> > > > > 
> > > > > The first two patches are probably fine but then we should just nuke
> > > > > libxl_uuid2string all together. Such a function makes sense for libxl
> > > > > where it's going to be constructing xenstore paths so may need to keep
> > > > > such things around but xl has no use for this as far as I can see.
> > > > 
> > > > FWIW the reason I didn't followup further with this thread (other than
> > > > it being a Friday afternoon) was that I agreed with your v2 patch which
> > > > added the macros and made the libxl_uuid2string fn internal to the
> > > > library. AFAICT your patch also solved all the leaks I was seeing.
> > > > 
> > > 
> > > I also think Gianni's libuuid series should be applied
> > 
> > I think that's orthogonal to the particular patch we are discussing here
> > though, that patch concerns the generation of UUIDs while this one
> > concerns leaking strings when formatting them...
>  
> sorry, I meant libuuid2string

Well, regarding the libuuid thing, I need to re-write that incorporating
Egger's suggestions about portability so it shouldn't be applied as is.

My conclusion on the mini-flame-war regarding mac addresses is that that
stuff is orthogonal to the part where we actually use a proper random
source.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel