On 2020/03/28 2:26, Christoph Hellwig wrote: > On Sat, Mar 28, 2020 at 01:50:10AM +0900, Johannes Thumshirn wrote: >> From: Damien Le Moal <damien.lemoal@xxxxxxx> >> >> Support REQ_OP_ZONE_APPEND requests for zone mode null_blk devices. >> Use the internally tracked zone write pointer position as the actual >> write position, which is returned using the command request __sector >> field in the case of an mq device and using the command BIO sector in >> the case of a BIO device. Since the write position is used for data copy >> in the case of a memory backed device, reverse the order in which >> null_handle_zoned() and null_handle_memory_backed() are called to ensure >> that null_handle_memory_backed() sees the correct write position for >> REQ_OP_ZONE_APPEND operations. > > I think moving null_zone_write earlier actually is a bug-fixd as is > as we should not touch memory if the zone condition or write pointer > isn't valid for a write. I'd suggest splitting that out as a bug fix > and move it to the start of the series so that Jens can pick it up > ASAP. OK. Will do that. Johannes, If you agree, I will send a patch separately for the move of null_handle_zoned() before the memcopy. While at it, I think I could also take patch 7 from this series and send it together with the reset all cleanup using req flag. That will make a mini series for cleaning & fixing null blk. > > Otherwise this looks good: > > Reviewed-by: Christoph Hellwig <hch@xxxxxx> > -- Damien Le Moal Western Digital Research
