Re: [PATCH 03/10] btrfs: split btrfs_submit_data_bio

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

On 2022/4/25 19:09, Christoph Hellwig wrote:
> On Mon, Apr 25, 2022 at 05:37:40PM +0800, Qu Wenruo wrote:
>> Oh, please don't completely get rid of btrfs_bio, even just for writes.
>>
>> The btrfs_bio::iter is pretty important for us to grab the original
>> logical bytenr of a bio.
>> As bio::bi_iter can be modified by lower level (does dm modifies it
>> too?), or btrfs itself.
>>
>> In fact, my incoming updated btrfs repair repair code heavily rely on
>> btrfs_bio::iter, both read and write, to grab the original logical
>> bytenr of the bio.
>
> Then it's doing the wrong thing.  I actually have a series to remove
> it entirely.

I'm wondering how would you iterate the bvec of a cloned bio then.

Regular bio_for_each_segment_all() will just trigger warning on cloned bio.
If you go something like chained bio, then any error would mark the
whole range error, and in fact my repair work is going to make
read-repair work with chained bio, thus I have to directly iterate
cloned bio anyway.

Just bio_for_each_segment()? That bi_iter is no longer reliable, just
btrfs_map_block() can modify it.

Anyway, what I really need is just a proper way to:

- Iterate bvecs of a clone bio
- Grab the original logical bytenr from a bio

If you can do that with extra members, I'm fine with alternative ways.

Thanks,
Qu