[PATCH] Fix bug #9130 - Certain xattrs cause Windows error 0x800700FF

[Jeremy Allison]

On Thu, Mar 28, 2013 at 07:42:01PM +0100, David Disseldorp wrote:
> On Thu, 28 Mar 2013 10:55:05 -0700
> Jeremy Allison <jra at samba.org> wrote:
> 
> > On Thu, Mar 28, 2013 at 10:42:55AM -0700, Jeremy Allison wrote:
> > > 
> > > New version that fixes these issues (and adds one code
> > > clarification change Andreas mailed to me privately).
> > > 
> > > Please review and push if you're happy.
> > 
> > Scratch that - a request from David crossed in the email :-).
> > 
> > Here's a version with both David and Andreas's suggestions
> > included.
> 
> Looks good Jeremy. One variation for ntvfs:
> 
> diff --git a/source4/ntvfs/posix/pvfs_qfileinfo.c b/source4/ntvfs/posix/pvfs_qfileinfo.c
> index 4309f3b..f3bd791 100644
> --- a/source4/ntvfs/posix/pvfs_qfileinfo.c
> +++ b/source4/ntvfs/posix/pvfs_qfileinfo.c
> @@ -96,16 +96,17 @@ NTSTATUS pvfs_query_ea_list(struct pvfs_state *pvfs, TALLOC_CTX *mem_ctx,
>         eas->num_eas = 0;
>         for (i=0;i<num_names;i++) {
>                 int j;
> -               eas->eas[i].flags = 0;
> -               eas->eas[i].name.s = names[i].name.s;
> -               eas->eas[i].value = data_blob(NULL, 0);
> +               eas->eas[eas->num_eas].flags = 0;
> +               eas->eas[eas->num_eas].name.s = names[i].name.s;
> +               eas->eas[eas->num_eas].value = data_blob(NULL, 0);
>                 for (j=0;j<ealist->num_eas;j++) {
> -                       if (strcasecmp_m(eas->eas[i].name.s, 
> +                       if (strcasecmp_m(eas->eas[eas->num_eas].name.s,
>                                        ealist->eas[j].name) == 0) {
> -                               if (ealist->eas[i].value.length == 0) {
> +                               if (ealist->eas[j].value.length == 0) {
>                                         continue;
>                                 }
> -                               eas->eas[i].value = ealist->eas[j].value;
> +                               eas->eas[eas->num_eas].value
> +                                               = ealist->eas[j].value;
>                                 eas->num_eas++;
>                                 break;
>                         }
> 
> What do you think?

No, that's actually incorrect. smbtorture raw.eas actually
tests for asking for an EA that dosn't exist by name using
an SMB1 EA list query, and Windows returns an EA entry containing
the requested name, with a zero length. So the current code
in both smbd and ntvfs for that case is correct.

We could expand the test to ask for 2 EA's in a list, one
that exists and one that doesn't, but my guess is it does
return an entry for each EA name requested, with a zero
length showing that value doesn't exist (if you want to
add that test feel free :-).

Jeremy.