On 3/22/23 5:26?PM, Andrew Morton wrote:
> On Wed, 22 Mar 2023 17:15:48 -0600 Jens Axboe <axboe@xxxxxxxxx> wrote:
>
>> On 3/22/23 5:13?PM, David Howells wrote:
>>> Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>>>
>>>> + if (unlikely(iov_iter_is_pipe(i))) {
>>>> + copied = copy_page_to_iter_pipe(page, offset, bytes, i);
>>>> + goto out;
>>>> + }
>>>
>>> This bit would need to be removed from copy_page_to_iter_atomic() as the two
>>> functions it calls should be removed by the patch in the block tree.
>>
>> Maybe it'd be better to consolidate rather than split the changes over
>> two trees?
>
> fyi, Lorenzo has sent out v7 of this series. I'll be pushing this in
> an hour or so. After which I suggest Stephen removes those (now) two
> lines and sends out one of his "build fix" emails which can be the
> basis for Linus's resolution.
>
> Or I can just steal "iov_iter: Kill ITER_PIPE"...
Or how about we just make sure to queue up items that touch them same
stuff in the same tree? I've already had this queued for a week, and
seems pretty silly to shuffle things around because some thing got
posted in 4 iterations today and then queued up right after.
Plus the build is now broken, so maybe a bit more diligence would be
required here than the drive-by-merging.
--
Jens Axboe
