On 08/12/2020 01:50, Ming Lei wrote:
> On Mon, Dec 07, 2020 at 06:07:39PM +0000, Pavel Begunkov wrote:
>> On 01/12/2020 12:06, Ming Lei wrote:
>>> Pavel reported that iov_iter_npages is a bit heavy in case of bvec
>>> iter.
>>>
>>> Turns out it isn't necessary to iterate every page in the bvec iter,
>>> and we call iov_iter_npages() just for figuring out how many bio
>>> vecs need to be allocated. And we can simply map each vector in bvec iter
>>> to bio's vec, so just return iter->nr_segs from bio_iov_iter_nvecs() for
>>> bvec iter.
>>>
>>> Also rename local variable 'nr_pages' as 'nr_vecs' which exactly matches its
>>> real usage.
>>>
>>> This patch is based on Mathew's post:
>>
>> Tried this, the system didn't boot + discovered a filesystem blowned after
>> booting with a stable kernel. That's on top of 4498a8536c816 ("block: use
>> an xarray for disk->part_tbl"), which works fine. Ideas?
>
> I guess it is caused by Christoph's "store a pointer to the block_device in struct bio (again)"
> which has been reverted in for-5.11/block.
>
> I'd suggest to run your test against the latest for-5.11/block one more time.
Ah, now it works, thanks for the idea
Unfortunately, I haven't got any logs, it died pretty early
--
Pavel Begunkov
