[PATCH] nvdimm/of_pmem: Provide a unique name for bus provider

[Dan Williams]

On Tue, Aug 6, 2019 at 11:00 PM Aneesh Kumar K.V
<aneesh.kumar at linux.ibm.com> wrote:
>
> On 8/7/19 10:22 AM, Dan Williams wrote:
> > On Tue, Aug 6, 2019 at 9:17 PM Aneesh Kumar K.V
> > <aneesh.kumar at linux.ibm.com> wrote:
> >>
> >> On 8/7/19 9:43 AM, Dan Williams wrote:
> >>> On Tue, Aug 6, 2019 at 9:00 PM Aneesh Kumar K.V
> >>> <aneesh.kumar at linux.ibm.com> wrote:
> >>>>
> >>>> ndctl utility requires the ndbus to have unique names. If not while
> >>>> enumerating the bus in userspace it drops bus with similar names.
> >>>> This results in us not listing devices beneath the bus.
> >>>
> >>> It does?
> >>>
> >>>>
> >>>> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar at linux.ibm.com>
> >>>> ---
> >>>>    drivers/nvdimm/of_pmem.c | 2 +-
> >>>>    1 file changed, 1 insertion(+), 1 deletion(-)
> >>>>
> >>>> diff --git a/drivers/nvdimm/of_pmem.c b/drivers/nvdimm/of_pmem.c
> >>>> index a0c8dcfa0bf9..97187d6c0bdb 100644
> >>>> --- a/drivers/nvdimm/of_pmem.c
> >>>> +++ b/drivers/nvdimm/of_pmem.c
> >>>> @@ -42,7 +42,7 @@ static int of_pmem_region_probe(struct platform_device *pdev)
> >>>>                   return -ENOMEM;
> >>>>
> >>>>           priv->bus_desc.attr_groups = bus_attr_groups;
> >>>> -       priv->bus_desc.provider_name = "of_pmem";
> >>>> +       priv->bus_desc.provider_name = kstrdup(pdev->name, GFP_KERNEL);
> >>>
> >>> This looks ok to me to address support for older ndctl binaries, but
> >>> I'd like to also fix the ndctl bug that makes non-unique provider
> >>> names fail.
> >>>
> >>
> >> 0462269ab121d323a016874ebdd42217f2911ee7 (ndctl: provide a method to
> >> invalidate the bus list)
> >>
> >> This hunk does the filtering.
> >>
> >> @@ -928,6 +929,14 @@ static int add_bus(void *parent, int id, const char
> >> *ctl_base)
> >>                  goto err_read;
> >>          bus->buf_len = strlen(bus->bus_path) + 50;
> >>
> >> +       ndctl_bus_foreach(ctx, bus_dup)
> >> +               if (strcmp(ndctl_bus_get_provider(bus_dup),
> >> +                                       ndctl_bus_get_provider(bus)) == 0) {
> >> +                       free_bus(bus, NULL);
> >> +                       free(path);
> >> +                       return 1;
> >> +               }
> >> +
> >
> > Yup, that's broken, does this incremental fix work?
> >
> > diff --git a/ndctl/lib/libndctl.c b/ndctl/lib/libndctl.c
> > index 4d9cc7e29c6b..6596f94edef8 100644
> > --- a/ndctl/lib/libndctl.c
> > +++ b/ndctl/lib/libndctl.c
> > @@ -889,7 +889,9 @@ static void *add_bus(void *parent, int id, const
> > char *ctl_base)
> >
> >          ndctl_bus_foreach(ctx, bus_dup)
> >                  if (strcmp(ndctl_bus_get_provider(bus_dup),
> > -                                       ndctl_bus_get_provider(bus)) == 0) {
> > +                                       ndctl_bus_get_provider(bus)) == 0
> > +                               && strcmp(ndctl_bus_get_devname(bus_dup),
> > +                                       ndctl_bus_get_devname(bus)) == 0) {
> >                          free_bus(bus, NULL);
> >                          free(path);
> >                          return bus_dup;
> >
>
> That worked.

Great. I'll make a formal patch, and I'll amend the changelog of the
proposed kernel change to say "older ndctl binaries mistakenly
require"
>
> -aneesh