On 3/5/19 5:32 PM, Jason Gunthorpe wrote:
> On Wed, Mar 06, 2019 at 03:02:36AM +0200, Artemy Kovalyov wrote:
>>
>>
>> On 04/03/2019 00:37, John Hubbard wrote:
>>> On 3/3/19 1:52 AM, Artemy Kovalyov wrote:
>>>>
>>>>
>>>> On 02/03/2019 21:44, Ira Weiny wrote:
>>>>>
>>>>> On Sat, Mar 02, 2019 at 12:24:35PM -0800, john.hubb...@gmail.com wrote:
>>>>>> From: John Hubbard <jhubb...@nvidia.com>
>>>>>>
>>>>>> ...
>>>
>>> OK, thanks for explaining! Artemy, while you're here, any thoughts about the
>>> release_pages, and the change of the starting point, from the other part of
>>> the
>>> patch:
>>>
>>> @@ -684,9 +677,11 @@ int ib_umem_odp_map_dma_pages(struct ib_umem_odp
>>> *umem_odp,
>>> u64 user_virt,
>>> mutex_unlock(&umem_odp->umem_mutex);
>>>
>>> if (ret < 0) {
>>> - /* Release left over pages when handling errors. */
>>> - for (++j; j < npages; ++j)
>> release_pages() is an optimized batch put_page() so it's ok.
>> but! release starting from page next to one cause failure in
>> ib_umem_odp_map_dma_single_page() is correct because failure flow of this
>> functions already called put_page().
>> So release_pages(&local_page_list[j+1], npages - j-1) would be correct.
>
> Someone send a fixup patch please...
>
> Jason
Yeah, I'm on it. Just need to double-check that this is the case. But Jason,
you're confirming it already, so that helps too.
Patch coming shortly.
thanks,
--
John Hubbard
NVIDIA
