Re: [PATCH] [RFC] UBI: Implement Fastmap support

[Richard Weinberger <richard@nod.at>]

On 24.05.2012 10:22, Artem Bityutskiy wrote:
> On Tue, 2012-05-22 at 18:55 +0200, Richard Weinberger wrote:
>>>> +     * has never seen any PEB used by the original fastmap.
>>>> +     */
>>>> +    if (!e) {
>>>> +            ubi_assert(ubi->old_fm);
>>>> +            e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_ATOMIC);
>>>
>>> Must it be GFP_ATOMIC?
>>
>> Yes. This function is called under a spinlock.
>
> I did not look close, but this sounds bad.
>
> You need to have a much better justification than "I allocate it under a
> spinlock". You need to tell "... because there is no way or very
> difficult to pre-allocate it while I do not have the spinlock held,
> because ... (explanation)".
>

True, in v7 I've reworked ubi_wl_put_fm_peb(). Now it does no longer 
need GFP_ATOMIC.

Thanks,
//richard