[PATCH v3 06/12] of: overlay: detect cases where device tree may become corrupt

[Frank Rowand]

On 10/19/17 12:04, Alan Tull wrote:
> On Tue, Oct 17, 2017 at 6:36 PM,  <frowand.list at gmail.com> wrote:
> 
>>  static int overlay_notify(struct overlay_changeset *ovcs,
>>                 enum of_overlay_notify_action action)
>>  {
>> @@ -86,8 +109,14 @@ static int overlay_notify(struct overlay_changeset *ovcs,
>>
>>                 ret = blocking_notifier_call_chain(&overlay_notify_chain,
>>                                                    action, &nd);
>> -               if (ret)
>> -                       return notifier_to_errno(ret);
>> +               if (ret == NOTIFY_STOP)
>> +                       return 0;
>> +               if (ret) {
>> +                       ret = notifier_to_errno(ret);
>> +                       pr_err("overlay changeset %s notifier error %d, target: %pOF\n",
>> +                              of_overlay_action_name[action], ret, nd.target);
>> +                       return ret;
>> +
> 
> Hi Frank,
> 
> This will spew lots of "error 0" messages for every notifier that
> returns NOTIFY_OK.
> 
> rmdir /sys/kernel/config/device-tree/overlays/1
> [  131.972505] OF: overlay: overlay changeset pre-remove notifier
> error 0, target: /soc/base_fpga_region/fpga_pr_region0
> [  131.987879] OF: overlay: overlay changeset post-remove notifier
> error 0, target: /soc/base_fpga_region/fpga_pr_region0

Thanks for finding that.  I'll send a patch so that NOTIFY_OK does
not spew an error.  (And overlay_notify() should also not be
returning an error in that case.)

-Frank


> 
> I could change fpga-region.c to return NOTIFY_STOP if it is accepting
> the overlay, but it will still want to return NOTIFY_OK for every case
> where it doesn't have an opinion.
> 
> Alan
> 
>>         }
>>
>>         return 0;
>