Re: [PATCH 2/3] of: Support hashtable lookups for phandles

[Rob Herring]

On Tue, May 10, 2016 at 8:45 AM, Pantelis Antoniou
<pantelis.antoniou@xxxxxxxxxxxx> wrote:
> Hi Rob,
>
>> On May 10, 2016, at 00:11 , Rob Herring <robherring2@xxxxxxxxx> wrote:
>>
>> On Mon, May 9, 2016 at 3:38 PM, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
>>> Hi Pantelis,
>>>
>>> On Mon, May 9, 2016 at 8:11 PM, Pantelis Antoniou
>>> <pantelis.antoniou@xxxxxxxxxxxx> wrote:
>>>> --- a/drivers/of/base.c
>>>> +++ b/drivers/of/base.c
>>>
>>>> @@ -1073,9 +1097,14 @@ struct device_node *of_find_node_by_phandle(phandle handle)
>>>>                return NULL;
>>>>
>>>>        raw_spin_lock_irqsave(&devtree_lock, flags);
>>>> -       for_each_of_allnodes(np)
>>>> -               if (np->phandle == handle)
>>>> -                       break;
>>>> +       /* when we're ready use the hash table */
>>>> +       if (of_phandle_ht_available() && !in_interrupt())
>>>
>>> I guess the !in_interrupt() test is because of the locking inside
>>> rhashtable_lookup_fast()?
>>
>> Not a use we should support. Just warn for anyone parsing DT in
>> interrupt context.
>>
>
> Thatâs not about users calling in interrupt context. Itâs when weâre
> very early in the boot sequence weâre under interrupt context and
> calls to the hash methods cannot be made.

I don't understand. When exactly are we in interrupt context?

Rob