Re: [PATCH RFC 4/7] md/raid10: convert read/write to use bio_submit_split()

[John Garry <john.g.garry@oracle.com>]

On 27/08/2025 08:31, Christoph Hellwig wrote:
> On Tue, Aug 26, 2025 at 06:35:10PM +0100, anthony wrote:
>> On 26/08/2025 10:14, Yu Kuai wrote:
>>>> Umm, that's actually a red flag.  If a device guarantees atomic behavior
>>>> it can't just fail it.  So I think REQ_ATOMIC should be disallowed
>>>> for md raid with bad block tracking.
>>>>
>>>
>>> I agree that do not look good, however, John explained while adding this
>>> that user should retry and fallback without REQ_ATOMIC to make things
>>> work as usual.
>>
>> Whether a device promises atomic write is orthogonal to whether that write
>> succeeds - it could fail for a whole host of reasons, so why can't "this is
>> too big to be atomic" just be another reason for failing?
> 
> Too big to be atomic is a valid failure reason.  But the limit needs
> to be documented in the queue limits beforehand.
> 
> 

What exactly could need to be documented?

We just report -EIO in this case (when we try to write to a bad blocks 
region with REQ_ATOMIC). In general, for RWF_ATOMIC, we report -EINVAL 
for too large/small a size.

BTW, do we realistically expect atomic writes HW support and bad blocks 
ever to meet?

Thanks,
John