Fix potential data loss and corruption due to Incorrect BIO Chain Handling

[Gao Xiang]

Hi Christoph,

On 2025/11/27 15:14, Christoph Hellwig wrote:
> On Thu, Nov 27, 2025 at 03:05:29PM +0800, Stephen Zhang wrote:
>> No, they are not using bcache.
> 
> Then please figure out how bio_chain_endio even gets called in this
> setup.  I think for mainline the approach should be to fix bcache
> and eorfs to not call into ->bi_end_io and add a BUG_ON() to

Just saw this.

For erofs, let me fix this directly to use bio_endio() instead
and go through the erofs (although it doesn't matter in practice
since no chain i/os for erofs and bio interfaces are unique and
friendly to operate bvecs for both block or non-block I/Os
compared to awkward bvec interfaces) and I will Cc you, Ming
and Stephen then.

Thanks,
Gao Xiang