On Tue, May 16, 2017 at 10:42 AM, Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
> On Tue, May 16, 2017 at 9:27 AM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
...
>>
>> Fix looks ok, but I wonder: upper fs with no xattr support is really a crippled
>> case we only need to support for backward compat. right?
>
> Yes, xattr is needed for opaque dir too, where we error out because
> there's no sane fallback. In fact there's a bit of a back-incompat
> thing due to 97c684cc9110 ("ovl: create directories inside merged
> parent opaque") trying to add the opaque flag even when not strictly
> needed and failing if xattr isn't supported, whereas before the patch
> that case wouldn't fail.
>
Right... but it can fail only in rename (into merge dir) case and before v4.10
there was no dir rename, so the back-incompat change is from -EXDEV
to -ENOTSUPP in this case. I can fix it up.
And now I see that 02cac32 (ovl: mark upper dir with type origin
entries "impure")
has the same problem. I'll send a fix for both.
>>
>> How about pre testing the noxattr condition on mount time (on workdir)
>> and issue a warning at mount time along the same lines of:
>> pr_warn("overlayfs: upper fs needs to support d_type.\n");
>> pr_warn("overlayfs: upper fs does not support tmpfile.\n");
>> pr_warn("overlayfs: upper fs does not support xattr.\n");
>
> We could. Theoretically we'd need to test for each xattr separately,
> but it's highly unlikely that a filesystem would support only a subset
> of the trusted.overlay.* namespace.
>
So I'll add a set xattr test on mount in addition to the existing tests.
Cheers,
Amir.
--
To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
